Computer With

Existirtg research provides little insight into how social influence affects the adoption and diffusion of competing innovative artifacts and how the experiences of organizational members who have worked with particular innovations in their previous employers affect their current organizations' adoption decision. We adapt and extend the heterogeneous diffusion model from sociology and examine the conditions under which prior adopters of competing open source software (OSS) licenses socially influence how a new OSS project chooses among such licenses and how the experiences of the project manager of a new OSS project with particular licenses affects its susceptibility to this social influence. We test our predictions using a sample of 5,307 open source projects hosted at SourceForge. Our results suggest the most important factor determining a new project's license choice is the type of license chosen by existing projects that are socially closer to it in its inter-project social network. Moreover, we find that prior adopters of a particular license are more infectious in their influence on the license choice of a new project as their size and performance rankings increase. We also find that managers of new projects who have been members of more successful prior OSS projects and who have greater depth and diversity of experience in the OSS community are less susceptible to social influence. Finally, we find a project manager is more likely to adopt a particular license type when his or her project occupies a similar social role as other projects that have adopted the same license. These results have implications for research on innovation adoption and diffusion, open source software licensing, and the governance of economic exchange.

Key words: open source software license; social networks; innovation adoption and diffusion; social influence

History: Shaila Miranda, Senior Editor; Ramnath Chellappa, Associate Editor. This paper was received on July 17, 2009, and was with the authors 23 months for 3 revisions. Published online in Articles in Advance November 8, 2012.

1. Introduction

The adoption of innovative artifacts--such as new ideas, products, technologies, processes and practices--by members of a social system and their diffusion through such systems has fascinated scholars from a variety of disciplines and research traditions for decades (Rogers 2003, Phelps et al 2012). Information Systems (IS) scholars, in particular, have sought to explain the adoption and diffusion of a variety of information technologies (IT) and IT-related practices (e.g., Hardgrave et al. 2003, Jeyaraj et al. 2006, Susarla et al. 2012). Although the vast majority of IT adoption research and much of the broader adoption and diffusion literature employs a rational decision-theoretic framework and models adoption as a straightforward cost-benefit analysis (Fichman 2004, Hall 2004, Rogers 2003), substantial research shows adoption decisions often involve profound uncertainty about costs and benefits (Rogers 2003). Facing such uncertainty, potential adopters typically turn to prior adopters as socially influential referents for guidance in determining the appropriate adoption choice (e.g., DiMaggio and Powell 1983).

An emerging theoretical and estimation framework that incorporates spatial and temporal heterogeneity in modeling the social influence of prior adopters is the heterogeneous diffusion model (Strang and Tuma 1993). This framework decomposes social influence in terms of three classes of factors: the susceptibility of a potential adopter to social information about the innovation, the infectiousness of information about the innovation from prior adopters, and the social proximity between prior and potential adopters (Greve 2005). Although research has used this framework to explain, inter alia, the adoption and diffusion of particular corporate governance practices (Davis and Greve 1997), prescription drugs (Strang and Tuma 1993), and business strategies (Greve 1998), its use by IS scholars is rare, despite calls for doing so (Fichman 2004). This research suggests the extent to which prior adopters socially influence potential adopters depends on the number and characteristics of prior adopters as sources of information about an innovation, characteristics of potential adopters as recipients and interpreters of this information, and the social proximity between them. Social influence also varies over time as greater or fewer actors adopt the innovation and become more or less socially proximate to potential adopters (Greve 2005).

Despite its contributions to an understanding of how and why innovative artifacts are adopted and diffused, research employing a heterogeneous diffusion framework, as well as the broader innovation adoption and diffusion literature, is limited in two important respects. First, the vast majority of research restricts its attention to a single innovation (Strang and Soule 1998). However, individuals and organizations are often confronted with multiple, substitutable innovations that compete for adoption. This menu of options increases the complexity of the adoption decision and is therefore likely to increase the uncertainty a potential adopter faces, potentially amplifying the role social influence plays in adoption and diffusion. Extant research provides little insight into how, when, or why social influence affects the adoption and diffusion of competing artifacts (Strang and Soule 1998). Moreover, the heterogeneous diffusion model has yet to be applied to explain the adoption and diffusion of multiple, competing innovations.

Second, existing adoption and diffusion research typically assumes that potential organizational adopters are exposed to an innovation through mass media and direct and indirect social ties to prior adopters rather than direct experience with the innovation. However, because organizations are collections of individuals, who often move between organizations as they change jobs, employees can gain experience with and knowledge about particular innovations adopted by their current organization and transfer this experience to other organizations when they move (Song et al. 2003). As a result, the nature of the experience employees have had with innovations adopted by their previous employers may influence how their new organizations evaluate the innovations, making them more or less sensitive to the social influence of prior adopters. Although interfirm employee mobility is a primary mechanism by which organizations learn from and influence one another (Song et al. 2003), research does not consider when or how the experiences of organizational members who have worked with particular innovations in their previous employers affect their current organizations' adoption of such innovations. Little research has examined the characteristics of potential organizational adopters, such as their employees' prior exposure to particular innovations, which affect their sensitivity to the choices of prior adopters (Weinert 2002).

This study seeks to address these important limitations of innovation adoption research employing a heterogeneous diffusion framework and the broader adoption and diffusion literature. We do so by investigating the conditions under which prior adopters of competing innovations socially influence how a potential adopter chooses among such innovations and how the experiences of members of a potential organizational adopter with particular competing innovations affects its susceptibility to this social influence. Of particular importance to IS research, we investigate this question by studying the adoption of particular types of open source software (OSS) licenses by new open source projects.

The OSS context is an ideal setting to investigate our research question for several reasons. First, the OSS licensing framework was an innovative departure from previous legal mechanisms to promote cooperation and beneficial exchange among actors (i.e., software developers and users) with divergent incentives (Demil and Lecocq 2006). Within the broad OSS framework, there is a variety of specific licenses. Because managers of new OSS projects choose a license at the inception of the project, specific types of OSS licenses represent discrete, competing innovative licensing practices that are at risk of being adopted by new OSS projects. Second, project managers face substantial uncertainty in choosing an appropriate license because of the novelty and large number of licenses (Rosenberg 2000) and the challenge of predicting how their choice of a particular license will affect developers' incentives to join their projects (Shah 2006). This uncertainty is exacerbated by the fact that the choice of an OSS license is a one-shot, largely irreversible decision made by licensors who are typically software developers with little or no legal expertise (McGowan 2001). Given this uncertainty, managers of new OSS projects are likely to consider existing projects as social referents that guide their license choice (DiMaggio and Powell 1983). Third, new OSS projects are often initiated by developers who have previously worked on other OSS projects (Hahn et al. 2008), allowing them to gain experience with the licenses used by these projects. These developers carry this experience with them to their new projects, which may influence how susceptible their new organizations are to the social influence of the license choices of extant projects. Moreover, in the OSS setting the project manager is the sole member of the project when the license choice is made, allowing for a clear, direct link between employee prior experience and organizational adoption choice. A final reason to study OSS license choice is that although research shows the choice of license by OSS projects can affect their performance (Comino et al. 2007, Stewart et al. 2006), OSS licensing research typically examines the choice between open and closed source licenses rather than the choice among OSS licenses. Only two studies have examined the determinants of OSS project license choice (Lerner and Tirole 2005b, Sen et al. 2009). Both studies, however, ignore the potential social influence of prior adopters of particular licenses on a project manager's license choice and how this influence may vary over time and by a project manager's social proximity to established OSS projects, characteristics of these projects, and the manager's previous OSS experience. This is a surprising and substantive limitation of this research given the substantial uncertainty managers of new OSS projects face in choosing a license, which suggests they may be socially influenced by the license choices of existing projects.

We adapt and extend the heterogeneous diffusion model to accommodate multiple, competing innovations. To do so, we investigate each dimension of this model--infectiousness, social proximity, and susceptibility--and derive predictions related to each dimension. We incorporate a novel source of influence on a potential adopter's susceptibility to the social influence of prior adopters--namely, the experiences of organizational members who have worked with particular innovations in other organizations.

We test our predictions in a sample of 5,307 OSS projects hosted at SourceForge. After controlling for factors shown to affect OSS license choice (Lerner and Tirole 2005b), our results suggest the most important factor determining a new project's license choice is the type of license chosen by existing projects that are socially closer to it in its inter-project social network. Moreover, we find that prior adopters of a particular license are more infectious in their influence on the license choice of a new project as their size and performance rankings increase. We also find that managers of new projects who have been members of more successful prior OSS projects and who have greater depth and diversity of experience in the OSS community are less susceptible to social influence. Finally, we find a project manager is more likely to adopt a particular license type when his or her project occupies a similar social role as other projects that have adopted the same license.

This study contributes to the innovation adoption and diffusion literature by addressing important limitations of the heterogeneous diffusion framework and to the literature on open source software by being the first study to explore when and how social influence from existing OSS projects affects a new project's license choice. This study also has substantive implications for understanding the origins and influence of the social institutions that govern economic exchange.

2. Open Source Software

2.1. Open Source Software Development Process

All OSS projects follow a similar process. An "initiating developer" begins a project by working on an idea and then hosts the source code and invites other developers to contribute. Developers volunteer to perform specific tasks and collaborate as a team, incorporating their individual creations into a single body of source code. Once an executable version of the software is developed, it is released for testing and feedback. The software evolves as new features are added, existing features are modified, and bugs get fixed. The process involves the sharing of ideas and joint problem solving that fosters social bonds among collaborators. Given the small number of developers typically involved in OSS projects (Krishnamurthy 2002) and the frequency and intensity of their interactions over time, the social ties among them can be quite strong (Hahn et al. 2008, Singh et al. 2011a, Singh and Tan 2010, Singh 2010). Because OSS projects stimulate the formation of social ties among teams of developers and because developers often work on multiple projects, a social network is produced that directly and indirectly connects developers participating in the broader OSS community. Although projects create ties among developers, projects also become connected by sharing common developers (Grewal et al. 2006). We examine the influence of this latter inter-project social network in this study.

2.2. Open Source Software Licenses

To be characterized as "open source," software must be offered under a license that satisfies several conditions. (1) Both the Free Software Foundation and the Open Source Initiative approve OSS licenses. OSS licenses differ in the extent to which they restrict how users may use and modify the software. At one extreme are highly restrictive licenses, such as the GNU general public license (GPL), and at the other extreme are permissive licenses, such as the Berkeley Software Distribution (BSD) license. Highly restrictive licenses differ from permissive licenses in two key ways (de Laat 2005):

(1) They require that, when modified versions of the program are distributed, the source code must be made generally available. This provision is called the "copyleft" clause.

(2) They prohibit the software to be mingled with other software that does not use the same license. This provision is called the "viral" or the "reciprocal" clause.

The highly restrictive licenses were the first free software licenses. The most famous of these, GPL, was authored by Richard Stallman, an early proponent of OSS and the initiating developer of a free operating system, GNU. The copyleft and viral clauses were designed to protect the software from being hijacked by proprietary software developers. These clauses require that any modification or derivation of the software has to be offered under GPL, making GPL'd software less attractive to commercial actors. The viral clause restricts the software from exploiting complementarities with other software, reducing its appeal to both contributors and users. Although the BSD license makes software attractive for commercial use because it allows modified or derivative works to be kept private, it does not protect the software from being hijacked. Stallman authored the Lesser General Public License (LGPL) as a compromise between the highly restrictive GPL and the permissive BSD license (Stallman et al. 2002). The LGPL includes the copyleft but not the viral provision. Lerner and Tirole (2005b) refer to LGPL-type licenses as restrictive. Any OSS license can be categorized as permissive, restrictive, or highly restrictive.

3. Prior Research on the Choice of Open Source License

Few studies have examined the determinants of OSS project license choice (Lerner and Tirole 2005b, Sen et al. 2009). Lerner and Tirole (2005b) model the licensor's problem as an optimizing balancing act between choosing a more restrictive license to attract more contributing developers and adopting a permissive license to preserve her ability to commercialize the software. Conditional on the exogenous characteristics of the software development project, the manager chooses the license that maximizes her expected benefits from the project given her evaluation of the expected response by potential developers. Consistent with their expectations, Lerner and Tirole (2005b) found that characteristics of an OSS project--such as its intended audience (e.g., end users), application genre (e.g., gaming), operating system (e.g., Linux), and user interface (e.g., GUI)--determine its choice of license (Lerner and Tirole 2005b). In contrast to the focus on project characteristics, Sen et al (2009) examine how the choice of OSS license type can be explained by the intrinsic and extrinsic motivations and the attitudes of the project manager. They showed that project managers who were motivated by the problem-solving challenges of OSS projects preferred moderately restrictive licenses, whereas those motivated by peer recognition preferred unrestrictive licenses. They also found that, when choosing a license, project managers are more concerned with the ideological principle that all OSS should be able to be freely redistributed than they are with end-users' rights and that licensors prefer licenses that are aligned with these attitudes.

Despite the insights this research provides into understanding OSS license choice, neither study considers the potential social influence of prior adopters of particular licenses on a new project's license Choice. This represents an important limitation of this research because a manager of a new OSS project is likely to be socially influenced by existing projects' licensing choices given the substantial uncertainty he or she faces in choosing a license (as described above). Indeed, we visited many online forums in which developers queried one another about which license to choose for their projects and observed substantial confusion and uncertainty on this topic. The two postings below are representative of what we observed:

The number of licenses is INSANE, I may as well lust write my own and add it to the list (joking ... mostly) .I have a few open source apps and think they are probably not licensed as to what I want, but I can't seem to find a good source to get a suggestion, and I don't want to read all of them. (2)

I wish there was just a spreadsheet of "This license has this feature."

I'm on the verge of releasing something and it's very hard to pick the right license. (3)

To explore the role social influence plays in the choice of open source license, we build on and extend the heterogeneous diffusion model (Strang and Tuma 1993).

4. Heterogeneous Diffusion Model

The heterogeneous diffusion model was developed to explain how social context influences an actor's adoption of an innovation and its diffusion within a population of actors (Strang and Tuma 1993). This model allows for spatial heterogeneity in the influence of prior adopters on individual adoption behavior and accommodates temporal heterogeneity by allowing the influence of prior adopters and social proximity to vary over time (Strang and Tuma 1993). In this model, the extent to which an actor's social context influences its adoption behavior depends on three categories of specific explanatory factors: infectiousness, social proximity, and susceptibility (Strang and Tuma 1993). The infectiousness of a prior adopter of an innovation refers to how influential information generated about its actions is for potential adopters, which is a function of characteristics of the prior adopter such as its size, performance, or status (Greve 2005). The social proximity of the source to a potential adopter describes how easily information is transmitted between them, based on their social distance from each other (Greve 2005). The susceptibility of a potential adopter refers to how open, receptive, or sensitive it is to being influenced by information available about the innovation and depends on inherent attributes that affect its motivation and ability to adopt the particular innovation (Greve 2005). The mechanism by which prior adopters influence potential adopters is the transmission of information between them, either through direct and indirect social ties, or through potential adopters observing prior adopters (Strang and Soule 1998). The model also specifies that an actor has intrinsic propensities to adopt an innovation based on its own inherent characteristics, independent of social context. We consider intrinsic characteristics affecting adoption as control variables that provide a baseline model.

Because the OSS licensing framework represents a novel set of contractual practices for protecting intellectual property (Demil and Lecocq 2006), we characterize types of OSS licenses as discrete and substitutable innovative practices that are at risk of being adopted by new OSS projects. We focus on the adoption of specific categories of licenses rather than individual OSS licenses. Although there are over 40 OSS licenses, they are small variations on two underlying themes--whether they contain copyleft and/or viral clauses (de Laat 2005, Lerner and Tirole 2005b, Stewart et al. 2006)--and thus represent three categories (types) of OSS licenses: unrestrictive (neither clause is present), restrictive (only the copyleft clause is present), and highly restrictive (both clauses are present) (Lerner and Tirole 2005b). Because these categories capture the essential differences among OSS licenses (de Laat 2005), each type represents a prototypical OSS license. Given the uncertainty surrounding these complex artifacts, potential licensors will tend to simplify their licensing decisions by focusing on prototypical licenses because these represent the essential and salient differences among the many individual licenses (Kahneman et al. 1982). Indeed, online tools (such as OSS Watch and Three.org) and articles (e.g., Niiranen 2009) designed to help developers choose OSS licenses typically explain and prescribe the licenses in terms of these prototypical characteristics rather than peripheral features. When an innovative practice is complex and uncertain, its prototypical characteristics rather than peripheral features drive its adoption (Ansari et al. 2010). Next, we define the type of social network in the OSS setting through which we expect social influence to operate. We then develop hypotheses linking variables associated with each dimension of the heterogeneous diffusion model to the likelihood a new OSS project will adopt a particular license type.

4.1. Inter-Project Social Network

The nature of the development process in OSS leads to the emergence of affiliation networks (Grewal et al. 2006). An affiliation network is a two-mode network because it consists of actors connected by their participation in common events and events that are connected by common actors (Wasserman and Faust 1994). An affiliation network therefore represents two different types of one-mode networks: an interactor network and an inter-event network (Wasserman and Faust 1994). In the OSS setting, the actors are individual developers and the events are projects. Developers have social ties with one another as a result of working together on the same project and projects are connected to one another as a result of sharing common developers (Grewal et al. 2006). Rather than focus on the inter-developer social networks that result from OSS affiliation networks, we focus on inter-project networks as the appropriate social network. An OSS project administrator initiates a project and is the sole developer on the project at the time of project registration and license choice. Thus, at the time of license choice, the project and administrator (i.e., licensor) are one and the same. Our theory suggests an administrator is influenced by his or her social proximity to, and characteristics of, other projects (rather than individual developers). Thus, an administrator's relevant social network at the time of license choice consists of the social ties the administrator-as-project has to previously established OSS projects.

4.2. Social Proximity

Social proximity refers to the social distance between two actors in a social network and determines how easily information is transmitted between them and the relevance of this information (Coleman et al. 1966). A socially proximate actor provides an influential frame of reference by which a focal actor evaluates and interprets information (Leenders 2002). Two approaches to conceptualizing social proximity in a network exist, each with its own causal mechanism linking proximity with social influence. The first approach--social cohesion--defines proximity in terms of the number, length, and strength of the paths that connect actors in a network (Marsden and Friedkin 1993). The second approach--equivalence--defines proximity in terms of the similarity of two actors' profiles of network relations (Marsden and Friedkin 1993).

4.2.1. Social Cohesion. The social cohesion approach defines social proximity in terms of the number, length, and strength of the paths that connect actors in a network (Marsden and Friedkin 1993). We focus on social distance (i.e., path length) as the primary dimension of social cohesion. The simplest form of cohesion is when two actors, such as a potential and prior adopter of an innovation, share a direct social tie. Directly connected actors communicate and share information with each other more frequently and with greater fidelity than indirectly connected actors (Burt 1982). Direct ties are conduits for the communication of rich, personalized information, which tends to be more influential than impersonal information sources (Rogers and Kincaid 1981). The volume and fidelity of information decays as the number of links indirectly connecting actors increases (Shannon 1949), making indirectly connected, actors less socially influential on a potential adopter than direct contacts (Burt 1982). Research in social psychology suggests involvement in shared activities provides opportunities for social cohesion to develop and that shared attitudes develop from social cohesion (Homans 1961). Faced with an uncertain situation, such as the adoption of an innovation, individuals discuss it with their proximate peers and develop a consensual normative understanding of the associated costs and benefits (Rogers 2003). Social ties provide detailed, personalized and more persuasive information on costs and benefits of adoption than general information sources (Rogers 2003). Discussions with prior adopters of an innovation build social pressures on the potential adopter to adopt the innovation when faced with an opportunity to do so (Rogers and Kincaid 1981). Social pressure increases with social cohesion and hence a potential adopter is more likely to adopt an innovation that has been adopted by his or her most proximate peers. Research has found that shared attitudes and behavior develop among people or organizations that are connected through direct communication channels (Coleman et al. 1966, Davis and Greve 1997, Haunschild 1994).

In the OSS context, projects on which a licensor has worked in the past provide greater opportunities for communication and thus for social cohesion. Through her discussions with developers on prior projects, a focal licensor develops a shared understanding of the costs and benefits associated with the license type chosen for those projects. Although direct involvement in a project provides opportunities to observe the consequences of adopting a license, a licensor may also receive useful and persuasive information from projects with which she is not involved but has social ties with developers who are. These ties can provide the licensor with detailed, personalized, and persuasive information on the costs and benefits of the particular license types adopted by these projects. Prior adopters have experience with a particular license type and thus understand it better than a potential adopter and may communicate their preferences persuasively via social ties with the focal project administrator. Hence, socially cohesive prior adopters exert social pressure on a potential licensor to adopt the same license type.

HYPOTHESIS 1 (H1). A licensor is more likely to choose a license type that was adopted by other projects to which he or she is more closely socially connected (i.e., socially cohesive).

4.2.2. Role Equivalence. An alternative conceptualization defines social proximity in terms of equivalence--the similarity of two actors' profiles of network relations (Marsden and Friedkin 1993). The social network literature initially conceptualized equivalence as structural equivalence. Two actors are structurally equivalent to the extent they have ties to the exact same other actors (Burt 1976). Research findings on the influence of structural equivalence on actor behavior can, however be interpreted in two ways (Mizruchi 1993). First, competition between actors over the same resources provided by the same alters triggers imitation between socially substitutable actors (Burt 1987). Second, social cohesion because of direct ties with the same set of third parties induces similar behaviors.

Because of these alternative interpretations, we employ the concept of role equivalence to capture how competition among actors, such as prior and potential adopters, results in imitative behavior. Two actors are role equivalent to the extent they engage in the same kinds of relationships with third parties (Mizruchi 1993). For example, two organizations are more role equivalent when they produce or trade in the same products, and therefore have similar types of upstream and downstream relationships, although not necessarily with the same suppliers or buyers (Guler et al. 2002, Winship and Mandel 1983). In this vein, equivalence among organizations has been defined as the extent to which they produce similar products or produce products using similar technologies and therefore have similar kinds of vertical relationships (Bothner 2003, Davis and Greve 1997, Flingstein 1985). Role equivalence captures the degree to which two actors occupy similar social roles and thus serve as common referents for one another, regardless of whether or not they share direct ties or are structurally equivalent (Mizruchi 1993).

Because role equivalence does not depend on the presence (or absence) of a tie between the actors being compared, the social influence effect of role equivalence on actor behavior is different from the influence associated with direct ties. The causal mechanism linking equivalence and the similarity of actors' behavior is social influence via a process of observational social comparison. Although role equivalent actors do not necessarily share relations with the same third parties, they compete with one another to retain their existing ties because third parties view such actors as substitutable objects of interaction (Guler et al. 2002). Competition among two equivalent actors increases their incentives to monitor and compare behaviors to ensure neither has an advantage or falls behind (Burt 1987). The more equivalent, the more one actor is likely to adopt an innovation previously adopted by another because it may make the other more attractive as the object or subject of relationships, resulting in some alters abandoning the non-adopter in favor of the actor that adopted (Guler et al. 2002). Nonadopters monitor and imitate the adoption behavior of role equivalent adopters in order to retain existing relationships and the benefits they provide (Guler et al. 2002). Actors look to equivalent others to identify appropriate behavior, particularly in contexts characterized by substantial uncertainty about how to behave (Burt 1987). The equivalence model of social proximity highlights symbolic communication among social substitutes rather than direct communication among contacts, which a social cohesion perspective emphasizes (Leenders 2002). This argument is consistent with the emphasis in neo-institutional research on mimetic isomorphic processes within industries (DiMaggio and Powell 1983). Prior research shows the extent to which organizations produce similar, competitive products or employ the same technologies are more likely to imitate each other's adoptions of innovations (Bothner 2003, Davis and Greve 1997, Flingstein 1985).

Based on prior research (Guler et al. 2002, Winship and Mandel 1983), we define role equivalence as the extent to which OSS projects engage developers and users in the same technologies. As such, OSS projects are role equivalent to the extent they employ the same technology platform in their development. Indeed, the primary data source we use in this study, Source-Forge, organizes all OSS projects it hosts into common domains or "foundries" based on their common usage of a technology platform, such as projects that use the Perl programming language. Projects in the same foundry typically target and compete for similar users and for developers with similar skills (Hahn et al. 2008). The extent to which OSS projects compete to maintain relationships with users and developers increases the extent to which the projects are social substitutes (from the perspective of users and developers). This increasing competition should increase the incentives OSS projects have to monitor and imitate each other's adoption of practices that can influence the performance of their respective projects, such as the choice of license. Because license choice affects OSS project performance and such choice involves considerable uncertainty, a licensor of a new project should be particularly sensitive to the license(s) previously adopted by equivalent projects. Compatibility issues associated with an OSS project's choice of license may increase the tendency to mimic the previous licensing decisions of equivalent projects. Complementarities often exist among software products and the exploitation of these complementarities influences a product's success (Gallaugher and Wang 2002). The license that governs the development and distribution of a software product influences its ability to exploit complementarities with other software products because such licenses influence the extent to which products are legally compatible and can be mingled and mixed by developers and users (Lerner and Tirole 2002). Licensors may fear the opportunity costs of choosing an inappropriate license--one that reduces its compatibility with other projects--and thus look for guidance on the appropriate choice from equivalent projects. In sum, a licensor will be more likely to monitor and imitate the license choices of projects in the same foundry.

HYPOTHESIS 2 (H2). A licensor is more likely to choose a license type that role equivalent projects have adopted more widely.

4.3. Infectiousness of Prior Adopters

The infectiousness of a prior adopter of an innovation describes how influential the information about its actions is for the adoption decision of a potential adopter. Because infectiousness is a property of information generated by prior adopters, its influence depends on whether or not it can reach a potential adopter and the distance it must travel to do so (Strang and Tuma 1993). In other words, the influence of infectiousness depends on the social proximity between a prior and potential adopter. Prior research suggests prior adopters can be more or less influential as social referents for a subsequent potential adopter based on characteristics such as their size, performance, or status (Greve 2005). We focus on prior adopter performance as the primary driver of its infectiousness. (4) License adoptions by higher performing OSS projects will be more influential on the choice of license type by new projects for several reasons. First, successful projects attract more attention, which leads to more information being available to potential adopters about them (Greve 2005). Second, potential adopters often attribute the level of success an actor subsequently achieves to their choice of innovation (Greve 2005), which suggests potential OSS license adopters may attribute a prior adopter's success, in part, to its choice of license, making information about such adopters more influential on the adoption decisions of others. However, such attributions may be unnecessary because people often copy the behavior of successful or prestigious referents, even when the focal behavior has nothing to do with the referent's success or prestige (Henrich and Gil-White 2001). Third, whereas the adoption of an OSS license is characterized by substantial uncertainty, innovations adopted by high status organizations are viewed as less uncertain and therefore more likely to be imitated by others (DiMaggio and Powell 1983). In this way, an actor's status provides an uncertainty-reducing signal about its underling quality and competence, increasing the extent to which others attend to and are influenced by its actions (Podolny 2001). Such status signals are particularly influential when recipients find it difficult to search for other useful information sources to inform their adoption decision (Simcoe and Waguespack 2011). In the OSS community, the success of a project reflects its social status in the community (Stewart 2005). Thus, the adoption of a particular license type by successful projects is likely to be imitated by new projects. Because the social proximity of established projects determines the extent to which they will be used as social referents by a focal project, we expect the license types chosen by socially proximate projects will have a stronger effect on a focal project's license choice when these projects are more successful.

HYPOTHESIS 3 (H3). A licensor is more likely to choose a license type that has been adopted by successful projects that are socially proximate.

4.4. Susceptibility of Licensor

Susceptibility refers to how much a potential adopter is affected by information about the practices adopted by others (Greve 2005). Holding constant the information about prior adopters that reaches potential adopters, a potential adopter is more susceptible when his or her adoption decision is more sensitive to and influenced by this information (Greve 2005). Little research has examined the characteristics of potential organizational adopters that affect their sensitivity to the choices of prior adopters (Wejnert 2002). In particular, research does not consider when or how the experiences of organizational members who have worked with particular innovations in other organizations affect their current organizations' adoption of such innovations. Potential adopters differ in their susceptibility to the social influence of prior adopters based on differences in a potential adopter's motivations to search for new practices and learn from the actions of others (Greve 1998). In addressing an important limitation of extant innovation adoption research, we argue that these motivational differences will be affected by the nature of the experience key organizational members (i.e., the project manager) have had with the innovations in their work on other projects that had previously adopted.

The success and failure of other OSS projects on which the licensor of a new project has worked will influence her susceptibility. Individuals and organizations are motivated to search for new practices and learn from others' efforts when they are dissatisfied with their own performance (Cyert and March 1963). Thus, the search for new behaviors or practices is triggered by the need to solve the problem of poor performance (Cyert and March 1963). Actors simplify the evaluation of their performance by assessing it relative to their aspiration level and by dichotomizing actual performance as "success" or "failure" depending on whether it was above or below their aspiration level (March and Simon 1958). An aspiration level is "the smallest outcome that would be deemed satisfactory by a decision maker" (Schneider 1992, p. 1053). Aspiration levels are determined by a process of social comparison (Cyert and March 1963, Festinger 1954, Greve 1998). Individuals compare themselves and their organizations with referent others who are similar to the focal actor and who therefore serve as a reference group (Greve 2005). A measure of the aggregate performance of the reference group, such as the average, constitutes the social aspiration level for an actor, who compares its own performance to this level to determine success or failure (Cyert and March 1963, Greve 2005). Poor performance triggers the search for new behaviors and practices, whereas success decreases the likelihood of such search. In the context of innovation adoption, performance below an actor's social aspiration level increases its willingness to consider adopting innovations adopted by its referent group, whereas success reduces incentives to change existing behavior and practices (Greve 2005). Poor performing organizations have been found to be more susceptible to social influence when faced with decisions to adopt novel practices (Davis and Greve 1997, Kraatz 1998).

In the OSS context, we expect the extent to which a licensor of a new project has previously worked on successful or unsuccessful projects will influence her susceptibility to the social influence of other socially proximal projects. As discussed previously, socially proximal projects are likely to serve as the referent group for the licensor of a new project. Relatively poor performance in prior projects should increase a licensor's propensity to search for appropriate practices, which will increase his or her susceptibility to outside influence. Experience working on successful OSS projects, in contrast, will tend to reduce a licensor's willingness to attend to the adoption decisions of other projects. Prior successful experience also sends a positive signal about the capabilities of the licensor to the community (Podolny 2001, Stewart 2005), which increases the project's attractiveness to potential contributors, irrespective of the license adopted. Such a licensor is less likely to seek help from others to determine the appropriate license type for his or her new project. Therefore, we expect that a licensor's susceptibility to influence from other projects decreases with the success of his or her prior OSS projects.

HYPOTHESIS 4 (H4). The effects of social proximity (social cohesion and role equivalence) on the likelihood a licensor will adopt a particular license type will decrease with the relative success of the other OSS projects on which the licensor worked.

To answer this question, I talked via e-mail with a search engine optimization professional, Alan Webb, who couldn't have a better last name for what he does. Webb is CEO of Abakus Internet Marketing <www.abakus-internet-marketing.de/en/index.shtml>. Exemplifying the world wide nature of the World Wide Web, Abakus is based in Germany.

About 85 percent of Web sites are discovered through a search engine, Webb says. The search engine market leader for some time now has been Google. Because Google has been used as the search technology by other search sites, such as Yahoo, AOL, Netscape and iWon, Google's technology controls 76 percent of the search market, according to WebDex <www.webdex.biz>, another search engine optimization company, based in Dallas. (Yahoo recently bought the Inktomisearch technology, which MSN also uses, and may switch to it.)

For new sites, the figure for Google is even higher, approaching 95 percent, Webb says. Because Google owns the searchengine space to get people to your site, you need to show them the way through Google.

The reason for Google's success is the relative relevance of its search results. According to research Webb has uncovered, about 70 percent of searchers therefore don't look past the top 10 sites presented, or first page. Fully 90 percent don't go past the first three pages.

Clearly then it's imperative for a Web site that wants to be found to be on Google's first page. But don't bother to pay for a sponsored link on that page. The average click-through rate for them is only about 6 percent, says Webb.

How your site shows up on Google depends on some things you can't easily control, such as how many other sites link to yours, and on some things you can easily control, such as your site's title and description.

Every Web page should have a title tag, which shows up at the top of viewers' browsers. You create a title tag in the <HEAD> section of a Web page using a short piece of HTML code, which directs how pages are seen by Web browsers. The title tag is the piece of HTML code that receives the highest weighting by Google.

When creating a title tag, even experienced Web designers often get it wrong, says Webb, by choosing the wrong words. Don't use a title tag such as this: <TITLE>Welcome to MyWebSite.com!</TITLE>

This will do nothing for your search engine results.

Instead, use two or three keyword phrases, consisting of one to three words, separated by a hyphen, that clearly spell out what your site is all about. Choose those keyword phrases that Google searchers will most likely type in. Use the same keyword phrases in the text of the page itself, ideally two or three times.

In choosing optimal keyword phrases, the pay site Wordtracker <www.wordtracker.com> or free site Overture <inventory.overture.com> can help.

But don't be guilty of "keyword flooding," Webb says. Some Web designers overeagerly load their title tags with a dozen or more keyword phrases.

Search engines treat this as "spam" and penalize a site for this in their rankings.

You should also use a meta description tag. It should be placed after the title tag and looks like this: <META NAME="Description" CONTENT-"The [specific] industry's hardest working company in providing [specific] solutions to [specific] customers">.

The meta description is what searchers will see after the title ha the list of Google results and other search engines, If you don't use a meta description tag for the description of your site that appears on the search results page, searchers instead will see the text around the first occurrences of the searched-for term, which may not provide enough information for searchers to want to click through to your site.

Particularly for business sites, sometimes it makes sense to hire a professional to improve your search engine rankings. Google factors in more than 100 different HTML, design and off-page factors in ranking sites. Testing different combinations is often needed.

One good place to look for professional help is the Web site Search Engine Optimization Consultants <www.seoconsultants.com>. Good sources of additional information about scarch engine optimization are Spider Food <www.spiderfood.net>, SearchEngineWatch.com <www.searchenginewatch.com> and SEO Chat <www.seochat.com>.

Consumers using a search engine face the option of clicking organic or sponsored links. The organic links are ranked according to their relevance to the search query, whereas the sponsored links are allocated to advertisers through a competitive auction. Because consumers tend to trust organic links more, advertisers often try to increase their visibility in the organic list by gaming the search engine's ranking algorithm using techniques collectively known as search engineoptimization (SEO). (1)

A notable example of the dramatic impact that an SEO campaign can have is that of JCPenney, an American retailer. This retailer's organic links skyrocketed during the 2010 holiday shopping season and suddenly climbed to the top of the searchresults for many general keywords such as "dresses," "bedding," and "furniture" (Segal 2011). JCPenney eventually fired their SEO contractor after finding out that they used black hat techniques that eventually led to a punitive response from Google. Search engine optimization is widespread in the world of online advertising; a 2010 survey of 1,500 advertisers and agencies revealed that 90% of them engaged in SEO, compared with 81% who purchased sponsored links (Econsultancy 2010). In the past few years, search engine optimization has grown to become a multibillion dollar business (VanBoskirk 2009).

This paper explores the economics of the SEO process and its effects on consumers, advertisers, and search engines. Using a game-theoretical model, we fully characterize the incentives and trade-offs of all players in the ecosystem. Our model consists of (i) advertisers with exogenous qualities and potentially correlated valuations for clicks, competing for the attention of consumers; (ii) a search engine that offers both organic and sponsored links and can set minimum bids; and (iii) consumers who engage in costly search to find the highest quality site. To capture the effect of SEO, we model the imperfections in the algorithms used by search engines, assuming that there is a measurement error that prevents thesearch engine from perfectly ordering links according to quality. Advertisers can, in turn, manipulate the potentially erroneous quality observations to their advantage through SEO and improve their ranking. A key parameter of our model is the effectiveness of SEO, determining the extent to which SEO efforts by advertisers affect the organic results.

We first ask how SEO changes the organic results and whether these changes are always detrimental to consumers and high-quality advertisers. The interest in this question stems from the strong stance that search engines typically take against SEO by emphasizing the potential downside on organic link quality. To justify their position, search enginestypically claim that their manipulation results hurt consumer satisfaction and decrease the welfare of "honest" sites. In contrast, search engines also convey the message that the auction mechanism for sponsored links ensures that the best advertisers will obtain the links of highest quality, resulting in higher social and consumer welfare. This reasoning suggests that consumers should trust sponsored links more than organic links in equilibrium and would prefer to start searching on the sponsored side. A substantial contribution of using a sophisticated model for consumers is that we are able to derive their optimal search behavior. Contrary to claims by search engines, we find that search engines fight SEO because of the trade-off advertisers face between investing in sponsored links and investing in influencing organic rankings. As a consequence, search engines may lose revenue if sites spend significant amounts on SEO activities instead of on paid links and content creation.

To approach the issue of diminished welfare from SEO, we first focus on the case where sponsored links are not available to advertisers and consumers. This base model serves as a benchmark and gives us a deeper understanding of the nature of the competition for organic links when using SEO activities. Our first result reveals that SEO can be advantageous by improving the organic ranking. In the absence of sponsored links, this only happens when advertiser quality and valuation are positively correlated. That is, if sites' valuations for consumers are correlated with their qualities, then consumers are better off with than without some positive level of SEO. By contrast, if there are sites that extract high value from visitors yet provide them with low quality, then SEO is generally detrimental to consumer welfare. The SEO process essentially allows sites with a high value for consumers to correct the search engine's imperfect ranking through a contest.

The second question we ask focuses on the full interaction between organic and sponsored links when SEO is possible. The institutional differences between the organic and sponsored lists are critical to the understanding of our model. First, advertisers usually pay for SEO services up front, and the effects can take months to materialize. On the other hand, bids for sponsored links can be frequently adjusted depending on the ordering of the organic list. Second, SEO typically involves a lump-sum payment for initial results, and the variable portion of the cost tends to be convex, whereas payment for sponsored links is on a per-click basis, with very little or no initial investment. Finally, there is substantial uncertainty in the outcome of the SEO process depending on the search engine algorithms, whereas sponsored links are allocated through a deterministic auction.

It is noteworthy that the presence of sponsored links accentuates the results of the base model and that SEO favors the high-quality advertiser regardless of the correlation between quality and valuation. The intuition is that sponsored links act as a backup for high-quality advertisers in case they do not possess the top organic link. When consumers have lowsearch costs, they will eventually find the high-quality advertiser, reducing the value of the organic position for a low-quality player. In equilibrium, consumers will start searching on the organic side, and high-quality sites will have an increased chance of acquiring the organic link as SEO becomes more effective.

Although SEO clearly favors high-quality advertisers, we find that there is a strong tension between the interests of consumers and the search engine. As advertisers spend more on SEO and consumers are more likely to find what they are looking for on the organic side, they are less likely to click on revenue-generating sponsored links. This tension may explain why search engines take such a strong stance against SEO, even though they favor a similar mechanism on the sponsored side. Furthermore, we obtain an important normative result that could help search engines mitigate the revenue loss stemming from SEO: we find that there is an optimal minimum bid the search engine can set that is decreasing in the intensity of SEO. Setting the minimum bid too high, however, could drive more advertiser dollars away from the sponsored side toward SEO.

As common as the practice of SEO may be, research on the topic is scant. Many papers have focused on sponsored links and some on the interaction between the two lists. In all of these cases, however, the ranking of a website in the organic list is assumed exogenous, and the possibility of investing in SEO is ignored. On the topic of sponsored search, works such as those by Rutz and Bucklin (2007) and Ghose and Yang (2009) focus on consumer response to search advertising and the different characteristics that affect advertising efficiency. Other recent examples, such as those by Chen and He (2011), Athey and Ellison (2011), and Xu et al. (2011), analyze models that include both consumers and advertisers as active players.

A number of recent papers study the interplay between organic and sponsored lists. Katona and Sarvary (2010) show that the top organic sites may not have an incentive to bid for sponsored links. In an empirical piece, Yang and Ghose (2010) show that organic links have a positive effect on the click-through rates of paid links, potentially increasing profits. Taylor (2013), White (2013), and Xu et al. (2012) study how the incentives of the search engine to provide high-quality organic results are affected by potential losses on sponsored links. The general notion is that search engines have an incentive to provide lower-quality results in order to maximize revenues.

The work of Xing and Lin (2006) is the closest antecedent to our paper. It defines "algorithm quality" and "algorithm robustness" to describe the search engine's ability to accurately identify relevant websites. Their paper shows that when advertisers' valuations for organic links are high enough, SEO is sustainable, and SEO service providers can then free ride on the search engine as a result of their "parasitic nature." The relationship between advertiser qualities and valuations and the strategic nature of consumer search is not taken into account. An earlier work by Sen (2005) develops a theoretical model that examines the optimal strategy of mixing between investing in SEO and buying ad placements. It is noteworthy that the model shows that SEO should not exist as part of an equilibrium strategy.