Libre Open Source Software development in developing and developed countries: A conceptual framework with an exploratory study

Libre Open Source Software development in developing and developed countries: A conceptual framework with an exploratory study

Decision Support Systems 46 (2008) 173–186 Contents lists available at ScienceDirect Decision Support Systems j o u r n a l h o m e p a g e : w w w...

432KB Sizes 0 Downloads 18 Views

Decision Support Systems 46 (2008) 173–186

Contents lists available at ScienceDirect

Decision Support Systems j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d s s

Free/Libre Open Source Software development in developing and developed countries: A conceptual framework with an exploratory study Ramanath Subramanyam 1, Mu Xia ⁎,1 Department of Business Administration, College of Business, University of Illinois at Urbana–Champaign, Champaign, IL 61820, United States

a r t i c l e

i n f o

Article history: Received 11 September 2007 Received in revised form 3 June 2008 Accepted 19 June 2008 Available online 2 July 2008 Keywords: Open source software Software development Developer motivations Project-level preferences

a b s t r a c t We propose a conceptual framework to understand drivers of motivation for developers in Free/Libre Open Source Software (FL/OSS) development across geographic boundaries. We identify generic motivations (such as sharing and learning, financial and career concerns, and satisfying functional needs), and project-level preferences (such as preferences for large teams, modular and universal projects), as well as the links between the two types of preferences of FL/ OSS developers in three regions — North America, China, and India. Our findings from an exploratory study indicate the presence of intrinsic motives in all three regions, with North American developers exhibiting stronger intrinsic motives such as sharing and learning. Project-level preferences differ considerably across the three samples. Finally, we observe that generic motivational factors are related to project-level preferences, although differently in the three regions. For instance, Chinese developers, who are driven by intrinsic motives, are drawn toward projects that are larger in scale, more modular, and universal in nature. In contrast, Indian developers with similar project preferences are mostly motivated by extrinsic motivations. © 2008 Elsevier B.V. All rights reserved.

1. Introduction FL/OSS (Free/Libre Open Source Software) refers to software programs that are distributed with its source code, often under a license that sets conditions for modification, reuse and re-distribution. Each FL/OSS program (project) is based on voluntary contributions, over the Internet, from a community comprising geographically distributed developers. FL/ OSS often offers the more reliable alternative, performs better [52], encourages creativity, and can find and fix defects more

⁎ Corresponding author. E-mail addresses: [email protected] (R. Subramanyam), [email protected] (M. Xia). 1 Both authors contributed equally. We thank UIUC CIBER for their support for the project, all the seminar participants at University of Illinois and Georgia Institute of Technology for their comments. We particularly thank Cynthia Beath for her detailed and insightful comments on an earlier version of this paper. We are also grateful to Kexin Zhao, Ya Tang, and Woo-je Cho for their excellent research assistance. 0167-9236/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.dss.2008.06.006

rapidly [39], compared to proprietary software. FL/OSS has seen dramatic growth globally both in programmer participation and market share. It is reported that worldwide there are about one million developers [34] working on more than 100,000 FL/OSS products [17]. Corporate adoption has increased significantly as well [7]. A large and growing body of prior IS development and management literature has examined contribution motivation in the context of FL/OSS. Motivations have been found to range from problem solving [11,45], to reputation benefits and career opportunities [28,4,42]. While most recognized FL/OSS practitioners reside in developed countries, support from developing countries has surged recently [23,29–31]. Yet, most of the extant research in FL/OSS has been performed in the western cultural context. Their findings cannot be readily generalized to developers from developing countries, which have different cultural and economic contexts. This is a crucial issue for the viability of FL/OSS as developing countries have an increasing and rapidly growing supply of developers; relatively recent trends indicate that

174

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

Table 1 Taxonomy of FL/OSS incentive research Generic motivation factors

Study

Findings

Sharing and learning (intrinsic)

Lakhani and Wolf [27] Shah [45]

• Improving skills is an important motivator for participation • Community norms of reciprocity and need-driven learning often motivate participants • Reputation effects matter to developers • Higher status in a merit-based ranking within the Apache Project does lead to significantly higher wages • Most developers do not foresee a link between participation and better job prospects • Developers start open source to “scratch[ing] a developer's personal itch”, i.e. to satisfy a need. • User benefits are key to a number of open source projects • User needs explain contribution to not only source code, but also less glamorous tasks such as documentation and support, as users benefit from the opportunity to learn through public posting of answers and questions • User needs are heterogeneous and users modifying their own software are more satisfied than non-innovating users • A majority (58%) of developers cite user needs as a motivator of contribution. These include personal needs (34% of all respondents) and professional needs (30% of all respondents). • Many developers participate because of their own needs.

Financial benefits and reputation (extrinsic)

Lerner and Tirole [28] Hann et al. [15] Shah [45]

User needs (internalized extrinsic)

Raymond [40] Lerner and Tirole [28] Lakhani and von Hippel [26]

Franke and von Hippel [12] Lakhani and Wolf [27]

Shah [45]

programmer supply in some of these developing countries has already outnumbered those from the western world2. If FL/OSS cannot engage the emergent and vast number of developers from developing countries, the needs of their users may remain largely unaddressed and community interest in FL/OSS could wane. Consequently, FL/OSS initiatives may stagnate and participation could eventually decline. Thus, there is a need to examine motivational differences among developers across regional boundaries, while taking into account the underlying differences in developers' valuations of merit from participation. In a discussion centered on FL/OSS developer motivation, one cannot ignore the fact that the actual participation of developers occurs in the confines of a project–; the primary unit of FL/OSS from an organizational perspective. Understanding project preferences, in the long run, should help to predict the kinds of projects that are likely to be successful FL/ OSS initiatives. Research on motivations, thus, cannot ignore the role of project characteristics on a participant's intention to contribute to FL/OSS. Yet, there seems to be a limited understanding of developers' project preferences and predilections. This paper develops theoretical arguments for the comparison of developers' motivations to participate and their project preferences in two settings: developed and developing countries. Specifically, we chose North America (NA) to represent developed countries and China and India for developing countries. We use NA in a loose sense to represent the western world, where the idea of FL/OSS originated and still where most of the activities have been taking place. We

2 For example, in India, the software industry employs more than 250,000 employees, sustaining annual growth rates of 30 to 40% in revenues and employment over more than ten years. Also, the number of IT graduates in India increased from 42,800 in 1997 to 71,000 in 2001. By comparison, the number of IT graduates in the United States increased from 37,000 in 1998 to 52,900 in 2000 [1].

selected India based on its leading position in the global software and business process outsourcing market and the vast number of software developers [1,46]. China, with its recent enormous economic growth and a burgeoning domestic IT and software market (e.g., [33]), offers another natural candidate for the study. The two developing countries, due to their sheer size, have relatively large and independent software developer communities, which provide a contrast to the western community. These three regions also differ significantly in several cultural and economic dimensions that are relevant to FL/OSS [18]. Our primary goal is to propose theoretical arguments for the discussion of the underlying rationale for motivational and project-level differences across developers of different cultural and economic backgrounds. We first investigate how three generic kinds of motivations noted in literature – intrinsic (sharing and learning), extrinsic (financial incentives) and internalized extrinsic (user needs) – affect developers' participation. Project-level preferences of developers and their relationships to the generic motivations are also analyzed by noting the cultural and economic backgrounds across the three regions. Based on our proposed conceptual framework, we designed an interview and a survey, and collected and examined responses from developers in the three regions. We followed a multiple case study approach suggested by Yin [54] and augmented it with survey data of developers from the three regions. Our findings from the small-scale study indicate the presence of intrinsic motives in all three regions. They also suggest that project-level preferences differed across the three samples. Finally, we observe that generic motivational factors were related to project-level preferences, although differently in three regions. Due to our small sample size, we call for largerscale studies that can complement our conceptual arguments and preliminary findings, which should bring more insights into this important area of understanding contribution behavior to FL/OSS.

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

We believe that understanding the differences between developed and developing countries is important for making policy decisions related to FL/OSS, especially in developing economies such as India and China. For example, we find evidence of sharing and learning being important in these regions. If this result holds in a larger-scale study, promoting FL/OSS education will enable novices and inexperienced developers to obtain design and programming skills through this medium. Moreover, the governments' investments in FL/ OSS are justified, not only from cost and security perspectives, but also from the software industry growth standpoint. This is because FL/OSS can be used as a means to provide exposure to systematic, collaborative software development — an experience that is hard to obtain except for those working for capable and innovative multinational software companies. Prior research has established that to foster virtual communities such as open source communities, it is vital that participants willingly sharing knowledge with other members (e.g., [6]). The rest of the paper is organized as follows. In Section 2, we introduce the background information and review prior literature. Section 3 presents details of the data collection process and the variable definitions. We discuss our empirical findings in Section 4. Section 5 provides some implications of our results and concludes our paper. 2. Background and literature review 2.1. Generic motivational factors: a taxonomy Participant motivation, as mentioned before, is the primary focal point of a large body of FL/OSS research. Researchers ask: Why do developers contribute to FL/OSS projects, when there is no direct monetary reward [41,16,28,27,43,45]? In general, the motivations can be categorized as either extrinsic, i.e., those that offer direct or delayed monetary rewards in mediated forms, or intrinsic, i.e. those that are pursued for their own sake. In addition to extrinsic and intrinsic motivations, Roberts et al. [42] introduce the notion of “internalized extrinsic motivations”. They argue that solving a problem of personal interest (satisfying user needs) or contributing for reputation concerns belong to this category as they are extrinsic but have to be internalized due to their selfregulating nature. In studying the relationship between these motivations and project preferences across different countries, we selected one central motivation from each of the three categories based on its importance noted by prior research. First is sharing and learning, which is the most widely-cited intrinsic motivation [27,28,45]. Second we consider financial and career benefits that developers derive from participating in FL/OSS development, as it is the most recognized extrinsic motivation [15,45]. We also study user needs, considered by prior researchers (e.g. [42]) as an important internalized extrinsic motivational factor. A summary of findings from prior research on these constructs is shown in Table 1. We will further elaborate on the motivations in the following subsections. We note that research on this wide-spread global phenomenon [14] could benefit from a cross-cultural perspective. Despite the proliferation of FL/OSS research, most research

175

projects have been conducted in the context of FL/OSS communities in western cultures, such as SourceForge.net or projects such as Apache, where developers often share western values (e.g. [16,27,45]). It is not known whether and how the findings from these papers apply to newer “geographic domains”, such as developing countries like China and India, where motivations for participation may differ due to different cultural as well as economic backgrounds. 2.1.1. Intrinsic motivation: sharing and learning Among all the incentives for participation, intrinsic incentives are probably the most documented, since these are unique to FL/OSS compared to proprietary software development. For example, the entire FL/OSS development process provides ample opportunities for experienced developers to share their knowledge and their inexperienced counterparts to learn [3]. Indeed, sharing and learning has been identified as an important motivation in FL/OSS [28,53,42]. Researchers do find that developers of FL/OSS, including project initiators as well as later participants, place a high premium on the sharing and learning effect of participating in FL/OSS [53,45]. Complementary to such learning effects, Shah [45] also finds that community norms of reciprocity and need-driven learning often motivate developers of some FL/OSS communities. For instance, she reports that some developers may contribute in order to elicit solution improvements, initiate and sustain discussions of their solutions, and to bring their solutions to the attention of others in the community. The cultural and economic sphere to which the developers belong also has a critical impact on the extent to which

Table 2 Hofstede measures for North America, China and India Measure

North Americaa

China

India

Individualism Uncertainty Avoidance Power Distance Masculinity Long-Term Orientation

86 47 40 57 26

20 40 80 50 118

48 40 77 56 61

Notes from Hofstede [18] regarding the cultural measures: Individualism is the degree to which individuals are integrated into groups. On the individualist side we find societies in which the ties between individuals are loose: everyone is expected to look after him/herself and his/ her immediate family. On the collectivist side, we find societies in which people from birth onwards are integrated into strong, cohesive in-groups, which continue protecting them in exchange for unquestioning loyalty. Uncertainty Avoidance index deals with a society's tolerance for uncertainty and ambiguity. It indicates to what extent a culture programs its members to feel either uncomfortable or comfortable in unstructured situations — novel, unknown, surprising, different from usual. Power Distance index is the extent to which the less powerful members of organizations and institutions accept and expect that power is distributed unequally. Masculinity refers to the distribution of roles between the genders The women in feminine countries have the same modest, caring values as the men; in the masculine countries they are somewhat assertive and competitive, but not as much as the men, so that these countries show a gap between men's values and women's values. Long-Term Orientation — values associated with Long-Term Orientation are thrift and perseverance; values associated with Short Term Orientation are respect for tradition, fulfilling social obligations, and protecting one's ‘face’. a We averaged the scores of USA and Canada for this measure; they were similar in these five dimensions.

176

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

developers contribute voluntarily. Accordingly, sharing and learning, a strong intrinsic driver, may be affected by the extent of individualism in a culture. From Hofstede's widelycited research on country-level indicators (Table 2), it is evident that the level of individualism is very high in North America, in contrast to China and India [18,19]. Prior research in the context of information systems development and group support systems has maintained that the role of culturespecific factors, such as individualism and power distance [18], is important when examining team member motivations and contributions (e.g., [21,50]). We expect that societies with high extent of individualism (e.g., North America) are likely to involve loose ties between individuals, who are expected to look after themselves and the immediate family [18]. In such societies, open source development provides means for likeminded individuals across the population to share and learn from each other. In collectivistic societies, we expect that familial norms would overpower individuals' motivations to share and learn from developers across the population. In addition, economic factors might also affect their sharing and learning motivation. In developing countries, because of the income potential for possession of FL/OSS skills, developers may be more motivated to learn, and FL/OSS projects provide valuable opportunities. The combined effect of the two opposing forces is not clear. 2.1.2. Extrinsic motivation: career concern and financial benefits Even though direct financial reward is unlikely to be provided to developers working on an FL/OSS project, indirect financial incentives cannot be ignored (e.g., [42]). Among them, career concerns and reputation effects are the most recognized. Since FL/OSS communities provide the ability for developers to signal their capability through their work, employers can observe such participation and reward skilled ones with higher-paying jobs [28]. Investigating the potential prevalence of such a phenomenon, Roberts et al. [42], in their study of the Apache HTTP Server project, find that while greater FL/OSS participation per se, as measured in contributions made, might not lead to wage increases, a higher status in a merit-based ranking within the Apache Project was associated with significantly higher wages. This suggests that participation could be used as a signaling mechanism to potential employers. Extrinsic motivation is influenced by both cultural idiosyncrasies as well as the economic environment. We observe high scores of the power distance measure (Table 2) in China and India in comparison to North America. Specifically, the power distance measures across India and China are also closer in the scale. These scores reflect the greater inequality in power and wealth in these societies [18], which might motivate developers to weigh financial and career concerns to a greater extent. Consequently, we expect that financiallydriven motives for participation in FL/OSS development are likely to be stronger among developers in China and India compared to those in North America. Factors related to the economic environment complement these culture-specific effects. As a case in point, we observe that, in developing countries, despite the relatively higher income of software developers compared to the general population, a typical programmer's income still does not afford her/him the freedom to work on tasks with little

monetary compensation. On a related note, we observe that there is a noted absence of a “geek culture” in developing countries in comparison to certain developed ones. Unlike their counterparts in western countries, a lot of whom grew up writing programs, a majority of developers in some of these countries are drawn to the field for its high income potential. Complementing these economic realities are the regulatory environments across countries which may also affect developers' motivations. In many countries, FL/OSS is promoted by the government as an alternative to proprietary software for factors such as low cost, openness, better security, and limited explicit switching costs [14,30]. For instance, governments of China [29,30], India [31], and Brazil [23], have set up programs and regulations to encourage the growth of FL/OSS through distribution of FL/OSS software, education, and government procurement preferences and tax benefits for FL/OSS software firms. Such regulatory preferences translate into economic benefits for FL/OSS developers and entrepreneurs. The resulting demand for FL/OSS developers presents an additional benefit of gaining experience in this field. Since all the reasoning predicts the same direction, extrinsic motives such as potential financial benefits are likely to be more important for developers of China and India in comparison to developers in North America. 2.1.3. Internalized extrinsic motivation: user needs It is well accepted that one of the major advantages of FL/ OSS is the integral involvement of the user in both software creation and production (e.g. [28,45]), as they participate in the production of software they will use in the future. Lerner and Tirole [28], in their paper on developers' incentives to contribute to FL/OSS projects, draw observations based on interviews of developers of four major FL/OSS projects. They identify two sources of benefits for developers as users. First, many FL/OSS projects have been started because the developers wanted to “scratch an itch”, i.e. write the software to accommodate their own needs that are not satisfied by existing programs, as reported in Ghosh et al.'s [13] survey of FL/OSS developers. Such “lead users” have very strong motivations to innovate and are an important source of product innovations for firms [51,36]. Franke and von Hippel suggest that such developers are motivated when they generate innovations [12]. Second, developers receive direct benefits as users when they fix a bug in the program or customize it to meet their own needs. Neither of these benefits can be easily achieved in proprietary software development settings. Other researchers also report empirical evidence of user needs as a source of motivation. For example, Lakhani and Wolf [27] survey developers of projects hosted by SourceForge.net. They find that a majority (58%) of developers cite user needs as a motivator of contribution. These include personal needs (34% of all respondents) and professional needs (30% of all respondents). Moreover, as observed by Lakhani and von Hippel [26], user needs explain contribution to not only source code, but also less glamorous tasks such as documentation and support, as users benefit from the opportunity to learn through public posting of answers and questions. While researchers have considered heterogeneity in user valuations and abilities [12,20,25] when studying the competition between FL/OSS and proprietary software, there is

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

limited research on how user needs in different countries affect the development of FL/OSS3. Users and developers across countries vary greatly in their perception, needs and valuations of FL/OSS, and their ability to contribute to FL/OSS development initiatives. Moreover, there is limited prior understanding of progress in the FL/OSS movement in emerging software markets and production bases, such as India and China. Culture-specific factors also underlie the extent to which user needs explains contributions of OSS developers [40,10]. A user's need in the context of open source software is one form of self-actualization need, as opposed to deficiency need which is a prerequisite for the former [32]. According to Maslow, people are expected to attain self-actualization needs once all the deficiency needs (lower level) have been fulfilled; this is evident when people seek knowledge and self fulfillment (among other higher order needs) and reach beyond esteem, safety and other basic needs. We expect that greater proportions of developers in developed economies are likely to be motivated by self-actualization needs as opposed to lower-level needs. We anticipate lower proportions of such developers in developing economies. Corroborating this argument, we believe that, in most developing countries, where computing advances are relatively nascent and imported software often dominates its local counterparts in its prevalence, goals of satisfying deficiency needs are likely to be more pronounced in comparison to developed countries, where numerous important FL/OSS initiatives have often originated and provide an outlet for individuals to satisfy their self-actualization goals. The extent to which user needs motivate developers is likely to vary significantly across the regions. 2.2. Project preferences Unlike in proprietary software development, where firmlevel, project-level, and individual-level factors are all important [8], FL/OSS development can only be analyzed at the project and individual levels, since FL/OSS typically relies little on direct industry forces and firm-level inputs. Hence, it is crucial to understand project preferences and individual preferences, as well as the relationship between these two factors. Further, unlike in proprietary software, participation in FL/OSS is purely voluntary and project selection is a personal choice of the programmer. In aggregate, developers' choices determine which FL/OSS projects get developed, whereas the need to understand programmer's preference is minimal in proprietary production as project choice is rarely dependent on the developers' personal decision. This choice, in the long run, should help predict the kinds of FL/ OSS projects that are likely to be successfully developed. Yet, limited FL/OSS literature exist that study developers' projectlevel preferences; studies have focused on the relationship between developer motivation and choices of the license mechanisms for such projects (e.g., [48,37]).

3 An exception is Lakhani and Wolf's study [27], on SourceForge projects, that involves developers in multiple countries. However, 83% of the study respondents come from either North America or Western Europe.

177

We posit that when potential developers evaluate which project to contribute to, they choose those that are most likely to be successful. According to Crowston et al. [9], measures of success of FL/OSS can be categorized into four phases: system creation and maintenance, system quality, system use, and system consequences. Out of the four categories, only measures in the first two, namely system creation and maintenance, and system quality, are observable ex ante by potential developers. Because of the high correlations between measures within each category, we select one key measure in each stage to represent success in that phase: the number of developers for the creation and maintenance stage, and modularity for system quality. Both measures are easy to evaluate from an outsider's perspective and have been documented to influence people's choice (e.g., [53,2]). These measures correspond to the initiation and ongoing phases of OSS development research as noted in [37]. In addition to the two project-level measures, we introduce content universality as a third dimension that is likely to influence a programmer's project choice. Even though FL/OSS currently already enables collaboration across geographical and cultural boundaries, it is not clear whether developers choose projects that address universal needs as opposed to those satisfying only local needs. This factor is important because future users of the software, as well as the skills required to build them can be vastly different. 2.2.1. Modularity Prior research has argued that the modularity of a product should, in theory, influence developers' motivations to contribute in FL/OSS projects (e.g.,[2,4]). For instance, Baldwin and Clark [2] note that if a code base is modular, working developers should do better in a collective environment than coding alone. This suggests that projects that are designed using a modular architecture might entice contributing developers since they have a greater incentive to participate in the community. While anecdotally some FL/OSS projects are touted for their modularity and are the target of significant developer interest (e.g. [28,42]), there is limited empirical evidence validating programmer preferences for projects designed in a modular manner. 2.2.2. Large teams Projects developed in large teams have some inherent advantages and disadvantages. Larger projects offer more opportunities for developers to collaborate with and learn from peers sharing similar interests, while smaller projects present an opportunity for faster appreciation of one's work and limit the coordination challenges that are common in large teams (e.g. [5]), as well as provide a more intimate atmosphere conducive to learning. In large projects, participants take many different roles ranging from active developers who contribute as project leaders and core members to peripheral participants such as bug reporters and readers [53]. These diverse roles enable novices to perform auxiliary tasks and augment their skills from contributing at the periphery of these projects. Coordination and integration challenges in large projects can provide invaluable opportunities for hands-on learning for beginners, while providing an opportunity for the experienced ones to showcase their design, development and management skills.

178

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

2.2.3. Universal vs. local projects Developers have the choice of working on projects of universal relevance or local relevance. On the one hand, developers may choose to contribute to universal projects such as operating systems initiatives and web servers. On the other hand, they can choose to work on local projects such as local language dictionaries, or localization efforts related to a universally popular program. Note that the term local content is not restricted in the language sense. It is defined as software written for a geographically-congregated audience. For example, a software program to calculate taxes in a particular region has to incorporate all the local tax codes could be considered as local in nature, in that it cannot be used elsewhere without significant modifications. Further, as noted by Mockus et al. [35], developers in open source projects are highly likely to be users of such initiatives as well. Hence their needs are likely to drive their project preferences. On a related note, Krishnan and Subramanyam [24] find that users in different countries tend to significantly differ in their preferences on functionalities for the same software product. Such differences in user preferences, in conjunction with localization needs, may lead to different development priorities, both for proprietary and FL/OSS software. It is important to investigate whether FL/OSS programmer priorities are affected by such context-specific user preferences. 2.3. Relationship between generic motivations and project preferences Understanding the relationship between generic motivations and project-level preferences for FL/OSS developers can help predict projects that are likely to receive more interest from the community. We expect that the greater the fit between developers' motivations and their project preferences, the more likely the community would benefit from the individual's participation. We posit that certain conceptual linkages do exist between these two factors. In other words, depending on the developers' motivation in participation, they may prefer projects with certain characteristics. Furthermore, these linkages are likely to vary with the geographic and cultural boundaries of developers as we theorize next. 2.3.1. Link between “sharing and learning” and project preferences Novice developers can learn from FL/OSS projects by reviewing source code, participating in peripheral activities such as bug reporting, maintenance, documentation, and further development of OSS projects [53]. Likewise, developers skilled in such activities can share their experience and find it rewarding to coordinate these tasks among peers. Naturally, a project with a large developing team is likely to have more code, increased development scope, and provides more such opportunities for developers. Likewise, modular projects also make code reading, maintenance, documentation, and related tasks easier to manage for developers. Therefore, we expect developers motivated by sharing and learning to gravitate towards larger and more modular projects. As discussed earlier, for “sharing and learning” as a motivation, the cultural and economic forces act in different directions between developed and developing countries. It is

not clear which one is dominant over the other. Since universal projects draw participation from experts around the world, they provide an ideal medium for learning opportunities and sharing design and programming expertise. On the other hand, the communication and management structures of projects across cultural and geographic boundaries might reflect local customs and nuances, which in turn could affect the link between their sharing and learning intentions and their project preferences. Developers in collectivist cultures may perceive large projects as means to conduct joint and collaborative learning. 2.3.2. Link between financial, career concerns and project preferences Developers with financial and career motivations could use FL/OSS as a signaling mechanism (e.g., [28,42,45]). Larger projects provide more opportunities for developers to showcase their design and programming skills. The larger the project, the stronger the signaling effect, both in terms of the audience that is familiar with the programmer's expertise, as well as the sophistication of the project. Meanwhile, a more modular project allows a programmer to focus on a subset of the entire code, thus can participate in more projects and/or produce more output with the same amount of effort, making her more valuable in the job marketplace. Baldwin and Clark [2] argue that the presence of modularity leads to matched incentives for developers to contribute to the development of FL/OSS software and permits efficient exploitation of talent by allowing different work tasks to be assigned to the most efficient producer. Finally, we expect that universal projects provide greater visibility and more opportunities for developers in seeking future financial benefits, in comparison to local projects. As noted earlier, extrinsic motives such as potential financial benefits are likely to be more important for developers in developing countries. Accordingly, we expect that developers from these regions may prefer larger projects to hone their design and development skills. Similarly, as discussed by Baldwin and Clark [2], extrinsically motivated developers are expected to prefer modular projects. However, given the extents of variance in project management and programming expertise in these regions, it is not clear whether the relationship expected between extrinsic motivation and modular project preference holds true across cultures. Likewise, differences in career choices and maturities of local markets make it unclear whether universal projects or local projects provide greater extrinsic benefits to developers in developing regions. 2.3.3. Link between user needs and project preferences One unique aspect of FL/OSS is that developers themselves are the first and often the major users of the programs they help to write [35]. If a programmer is interested in addressing user needs by working on a FL/OSS project, a smaller project might provide a greater opportunity to address a larger proportion of an individual's needs. A motivated contributor might be driven by her/his ability to contribute through innovative ideas [12], which is more likely to receive prompt attention and approval in a smaller scale project. In a larger project, only a fraction of one's needs is likely to be addressed during development. Conversely, a participant may regard the

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

Fig. 1. Conceptual model.

number of developers in the project as a proxy for user interests and need for a new software product. It is not clear which of the relationships is dominant over the other, and how this varies across regional boundaries. If a participant is drawn towards FL/OSS to satisfy an unmet need, it is likely that he/she prefers a modularly-designed product in comparison to alternatives. Modular projects are argued to provide compatible incentives for individual participation, and provide an environment which can help transform a product concept to a fulfilled solution (e.g., [2,38]). As argued earlier, the goals of satisfying deficiency needs (lower-level needs in Maslow's hierarchy) are likely to be more pronounced in developing countries while developers in developed regions might benefit from using FL/OSS as means to satisfy their self-actualization goals. Hence, in NA, where the individualism factor is high, we expect developers to be highly heterogeneous in the way they select projects. While some developers may prefer to share their skills in a small project, others welcome the opportunity to showcase their skills in (and learn from) medium-sized or large-scale projects. Likewise, the inclination to contribute to universal projects as opposed to local projects depends on the needs that the programmer (user) wants to address. The low power distance measure suggests that developers in NA are considered relatively equal in status to their peers and thus they might be more inclined to collaborate with the larger pool of audience possible in universal projects. In a developing region where English is not the spoken language, local (language) software projects might be preferred over universal projects if unrealized user needs are an important motivation. The conceptual model for our study is shown in Fig. 1. 3. Data collection For this study, we followed the multiple case study protocol suggested by Yin [54] and augmented it with survey data. We started the study with interviews and preliminary discussions with six expert subjects from three geographical regions — North America, China and India. These participants were contributors to FL/OSS software in three distinct collaboration forums. For three of the cases (participants), who were key informants known to the authors, interview data were coded using axial coding and categorization (open coding) of data [49]. This process helped us create connections between the coded categories.

179

Two factors (extrinsic and intrinsic motivation) in the conceptual model served as initializers for the categorization process. More coding categories, especially the project preference categories, emerged during the actual coding of data. Through this process, several important patterns emerged. First, we observed that the nature of the project plays a key role in garnering attention of potential developers, with some product categories inherently drawing more interest from the interviewees. Second, notions of extrinsic motivation and intrinsic motivation differed based on what culture the participant was accustomed to. Third, we found that certain aspects of intrinsic motivation such as learning were strongly tied to participant's views on what projects would provide the best platform for learning. We then reviewed prior research and theories on FL/OSS communities to gain greater understanding about how the patterns we observed were consistent with or in contrast to prior theories. We iterated between theory and data to develop an initial framework for understanding the motivations of participants to contribute to the community. This process helped us create a survey to gather data on a larger scale from the members of the FL/OSS communities. However, since a majority of the developers contributed their personal time to the community outside (and in addition to) their work-related commitments, our key informants requested us to shorten any questionnaire or instrument that would be employed to gather data from the developers. This led us to creating a short and non-time consuming questionnaire. Nevertheless, we provided open items in the instrument that helped respondents convey their opinions and views that do not relate to any of the other questions. During the survey stage, the same questionnaire was sent to developers from the three communities. Developers were given consistent explanations of the meanings of the survey questions. For the Chinese community, the questionnaire was first translated into the Chinese by an informed bilingual coder and then verified for consistency by a neutral bilingual researcher before being administered to the developers. Data were collected during the year 2004–2005. The participants from the sample were contributors from 50 most active projects in SourceForge. For China, we selected participants from HuiHoo.org, one of the most active FL/OSS communities in China. Finally, for India, we surveyed contributors to active projects on the Sarovar.org FL/OSS collaboration forum. These three forums were chosen since they capture a representative pool of participants from that geographic locale. We obtained responses from about 40% of surveyed participants in the Indian community (19 participants), about 20% from the Chinese forum (28 participants), and 5% of participants from the North American surveyed pool (16 participants). We suspect that the difference in response rates, especially the relative low response rate of North American participants, could be attributed to the following several reasons. First, FL/OSS software has received much more attention and support from the government in developing countries (e.g., [22]). Hence, developers in these communities are much more eager to know how FL/OSS development is done in the developed countries than vice versa. Consequently, they tended to be more cooperative than their western counterparts in volunteering inputs for our

180

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

study. In contrast, participants of SourceForge had received too many survey requests in the past few years, their interest in participation had worn off considerably. This was brought to our attention by several respondents, some of whom declined to participate in our study primarily due to this reason. Second, we suspect that North American developers employ stronger email filtering rules than their developing country counterparts do, which lead to fewer deliveries of the survey. We should note that for those North American developers that did respond, the level of detail in their responses to open-ended comments compare favorably with responses from the other two samples. The most recent projects on which the participants from our sample have worked are summarized by target application platform (Fig. 2a) and by domain of the software (Fig. 2b). These numbers exclude the participants who did not choose to share details of their current projects, which represent 10% of our sample. We observe that a majority of participants have been involved in projects that concern tools for software development. In terms of software type, networking and systems software are the preferred types by the participants of the survey. 3.1. Methodological issues Performing a cross-cultural analysis using survey-based data poses several challenges that have to be explicitly addressed to ensure the validity of the results. Two such issues deserve special consideration — translation equivalence between the different questionnaires and sample

equivalence between different cultures [47]. We ensured that the meaning of the survey questionnaire was consistent after the survey was translated into Chinese by using a bilingual translator well-versed in both languages. A bilingual research assistant's help was also sought out to ensure that coding of responses was done appropriately. Our confidence in the sample equivalence is based on the following. First, the primary means by which the participants of FL/OSS community contribute and collaborate with others towards FL/OSS solutions is through Internet communities such as SourceForge, Freshmeat, Sarovar, and Huihoo. Hence, the sample of participants is relatively homogeneous in the manner in which they contribute to the project. Second, in the three cultural contexts, the participant in the survey is an active contributor to an active project. We next provide definitions of the variables used in our analyses and discuss our results. 3.2. Variable definitions 3.2.1. Generic motivational factors Participation in FL/OSS due to sharing and learning measures the extent to which a participant is drawn towards FL/OSS projects due to her belief in sharing work and learning from other participants. This measure is captured on a sevenpoint Likert scale. The second variable, participation in FL/OSS to provide oneself financial benefits, is also assessed on a seven-point Likert scale. This variable represents the extent to which an FL/OSS participant is drawn towards projects that provide the participant with financial benefits. Using a seven-

Fig. 2. a: Current project preferences of participants in the sample (by target application platform). Note: The target platforms are not mutually exclusive. b: Participant preferences in the sample (by software type).

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

181

Table 3 Descriptive statistics and correlations Variable

NA

China

India

Pairwise Pearson correlations (⁎ =N significance of p ≤ 0.05)

Mean

S.D.

Mean

S.D.

Mean

S.D.

Financial benefits

User need

Modular projects

Generic motivational factors Sharing and learning Financial benefits User needs

6.38 3.56 4.25

1.54 1.41 1.34

5.79 4.86 4.90

1.86 1.77 1.82

6.21 4.47 4.74

1.13 1.58 1.05

0.18 1 0.54⁎

0.34⁎ 0.54⁎ 1

0.45⁎ 0.39⁎ 0.46⁎

0.14 0.18 0.16

0.52⁎ 0.17 0.21

Project preferences Modular projects Projects involving large teams Universal projects

5.44 4.06 4.81

1.26 1.69 0.98

5.25 4.34 5.17

1.84 1.57 1.79

5.84 3.53 5.11

0.96 1.07 1.45

0.39⁎ 0.18 0.17

0.46⁎ 0.16 0.21

1 0.18 0.25⁎

0.18 1 0.17

0.25⁎ 0.17 1

Control factor Prior experience in open source development (in years)

3.37

2.77

1.88

1.60

4.02

2.96

0.02

0.32⁎

−0.08

−0.01

Projects involving large teams

Universal projects

−0.03

Variables indicate likelihood of participation.

point Likert scale, we measure participation in FL/OSS to satisfy functional needs and aim to capture the extent to which a participant is drawn towards initiatives that address needs not already met by commercially available alternatives. We averaged two items (reliability score of α = 0.7) to create this score. 3.2.2. Project preferences The variable, participation in modular projects, is measured on a seven-point Likert scale. It intends to capture the extent to which an FL/OSS participant is motivated to participate in a project for its modularity. Participation in projects with large teams is assessed using a seven-point Likert scale measure to capture the extent to which a participant gravitates towards projects developed in large teams. We substantiated this measure by querying the participant the size, in terms of number of modules, of the most recent project that she/he participated. We found the correlation between the two items to be 0.80. The third project-based measure, participation in universal projects was also based on a seven-point Likert scale. We created this score

using two items on the survey (reliability α = 0.84). This variable intends to capture the extent to which an FL/OSS contributor is motivated to participate in projects aimed at a universal audience. When this measure is low, participants prefer projects targeted at local audience or locale-dependent functionality. The control variable, participant experience in FL/OSS development, was captured through a survey item indicating the years of experience of the participant in FL/OSS software development. The sample size for our analysis is 63. The 63 cases were selected from the set of 71 responses for our analysis. The other eight responses belonged to geographical regions not tested in this analysis. The summary statistics for the variables in our study, including correlations, are shown in Table 3. 4. Empirical results and discussion In this section, we present our results in a parallel fashion with our theoretical analysis in Section 2. In Sections 4.1 and 4.2, we compare generic and project preferences across different regions, respectively. In Section 4.3, we study the

Table 4 Population comparisons — NA, China, India Variable

Generic motivational factors Sharing and learning Financial benefits User needs Project preferences Modular projects Projects involving large teams Universal projects Control Factor Control factor Prior experience

North America vs. China

North America vs. India

China vs. India

Z (means)

F (variance)

Z (means)

F (variance)

Z (means)

F (variance)

+1.72 ⁎ −2.49 ⁎⁎ −1.37

0.55 1.33 3.49 ⁎

+1.17 −1.71 ⁎ −0.85

0.10 0.00 0.05

−0.56 +0.99 +0.66

1.41 1.11 6.27 ⁎⁎

−0.08 −0.44 −1.59

1.36 0.05 5.19 ⁎⁎

−0.85 +1.07 −0.69

5.37 ⁎⁎ 1.67 5.38 ⁎⁎

−0.68 +1.69 ⁎ +0.48

8.19 ⁎⁎⁎ 3.98 ⁎ 0.42

+2.09 ⁎⁎

0.24

−0.72

1.94

−2.43 ⁎⁎

4.71 ⁎⁎

Wilcoxon Rank Sum/Mann–Whitney test (null — means are equal): Z-statistics shown for mean comparisons; Levene's test for equality of variance test (null — variances are equal): F-statistics shown. Note: ⁎⁎⁎ 0.01 level of significance, ⁎⁎ 0.05 level of significance; ⁎ 0.10 level of sig.

182

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

Table 5 Alignment between generic factors and project preferences Project preferences Generic motivational factors

Sharing and learning Financial benefits User needs Control: prior experience

Modular projects

Projects involving large teams

Universal projects

NA (R2 = 0.45)

China (R2 = 0.65)

India (R2 = 0.48)

NA (R2 = 0.11)

China (R2 = 0.25)

India (R2 = 0.27)

NA (R2 = 0.19)

China (R2 = 0.51)

India (R2 = 0.20)

n.s. n.s. n.s. +0.61 ⁎⁎

+0.61⁎⁎⁎ +0.20 ⁎ n.s. n.s.

n.s. +0.32 ⁎ +0.33 ⁎ +0.34 ⁎

n.s. n.s. n.s. n.s.

+0.68 ⁎ n.s. n.s. −0.50 ⁎

n.s. +0.51 ⁎ n.s. n.s.

n.s. n.s. n.s. n.s.

+0.73 ⁎⁎ n.s. n.s. n.s.

n.s. +0.55 ⁎ −0.39 ⁎ n.s.

MANOVA correlations are shown. Note: ⁎⁎⁎ 0.01 level of significance; ⁎⁎ 0.05 level of significance, ⁎ 0.10 level of significance.

difference, among different regions, of the relationship between the generic motivations and project preferences. For both generic motivations and project preferences, we performed pair-wise comparison across the three samples using a non-parametric approach. Further, to test whether there were systematic differences between the early sample of participants and the later sample, we tested for differences in means between the two groups. None of the variables were significant at the p = 0.05 level (the lowest p-value of the test of significant difference between variable means was p = 0.253). Also, because of the small sample size, the results could be sensitive to distributional assumptions. Hence, we adopt a non-parametric approach which is more robust to outliers and influential observations. We performed population mean comparisons using the Mann–Whitney rank-sum test. We find that in the three pair-wise analyses, the null hypothesis of equal variance was not rejected. To verify whether the results are sensitive to distributional assumptions, we performed ANOVA contrast tests, which were consistent with the non-parametric Mann–Whitney test results presented in Table 44. Finally, to test the relationships between generic motivations and project preferences, we perform Multivariate Analysis of Variance (MANOVA) tests of relationships between the two sets of factors and report the results in Table 5. 4.1. Generic motivations across regions 4.1.1. Intrinsic motivation: sharing and learning We performed dependent sample pair-wise t-tests to identify the most important generic motivational factor. We find that sharing and learning is the most important one in all the three regions, followed by user needs and financial benefits respectively. This ranking was consistent across the three samples though the statistical significance of the difference between user needs and financial benefits was not found in the China sample. This suggests that the three communities are strongly driven by the desire to learn development and share their skills with other developers in the community (e.g., [42]). When comparing across regions (Table 4), we find that the sharing and learning factor is significantly stronger in North America in comparison to China. Between NA and India, the difference is positive but not statistically significant. The 4 We have not included the results of the ANOVA contrast tests due to space limitations.

difference between NA and China could be due to several reasons. First, the degree of maturity of the software market is the highest in NA. Hence, the level of sophistication and the extent of choices available among proprietary and FL/OSS software is high in this region compared to the other two. This implies that the participants bring their niche interests to each initiative and understand the merits of FL/OSS relative to alternative approaches. For instance, one NA participant, when asked for his reason to participate in FL/OSS, responded …“There is more of a community donating code/ideas for the software so it soon becomes a better product for the user and quicker than it would if developed using a proprietary approach.” Likewise, participants were also particular about the nature of FL/OSS software products that they intended to contribute in. For example, another participant in the NA community stated…“I participate in projects that I want to use afterwards. And I prefer to stay in the kind of projects for end-users, not targeted at administrators or such.” Another reason for the difference between the regions is the extent of specialization and highlyspecific learning interests among the NA participants. Such a targeted interest was evident in some responses; one participant expressed … “My personal interest is in the area of programming. I'm primarily oriented towards compiling, emulation, and systems tools. I would not participate in any sort of Office-related project.” Under such circumstances, sharing of specific knowledge can be achieved better through a medium such as FL/OSS. These results are in line with the individualism argument noted earlier. FL/OSS provides an opportunity for participants to develop expertise in niche areas and interact with like-minded (but uncommon) individuals who share similar interests. In contrast, participants in the Chinese community and the Indian community (to a lesser extent) were often drawn towards the philosophy of FL/OSS and rarely expressed niche interests in particular specialties. This aspect was evident from responses of some Chinese participants: “… You can understand the entire framework by reading documents and codes, and you can know how to apply some design methods and select them from many alternatives…”. Likewise, Indian participants expressed a general interest in the philosophy rather than a specialized technical interest: “Programming is my form of art. It is to be peer reviewed and appreciated as widely as possible. I don't care much for which company I work for as long as I get the opportunity for the above.” Similarly, when asked for the reason for interest in FL/OSS, another experienced software developer responded “… (a) there is a community built around the project. (b) it helps the project evolve at a much greater rate. (c) it is more fun working for an open source initiative. (d) there is much more ego satisfaction.”

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

From an economic standpoint, when participants are motivated by income potential, specific skills (interests) are less likely to be important in comparison to generic programming and design skills, which are pre-requisites to attain maturity. 4.1.2. Extrinsic motivation: career concern and financial benefits We found considerable differences in participants' financial motives in the three regions. Complementary to the finding of Hann et al. [15] that career concerns are among the least important motivational factors in FL/OSS contributions, our results add to the discussion of applicability of these findings to developing country participants. We find that participants from the Chinese as well as the Indian communities were strongly driven by career concerns and financial benefits in comparison to North American participants (Table 4). There could be several reasons for these differences, as we expected. For instance, there are higher economic inequalities in China and India than in NA. Participants, hence, are likely to be drawn toward initiatives that raise their current and future income potential. Another reason could be the indirect financial benefits of participation in FL/OSS. This was evident from views of a few Chinese participants on their motivation for FL/OSS participation. For instance, one respondent noted a trend that…“…Salaries for FL/OSS developers are higher than, say, Windows developers. While some developers find working in FL/OSS more satisfying than commercial software, most do not have any preference…”. 4.1.3. User needs We find that the variance of user needs measure is unequal across the NA–China and China–India pairs of samples (in pair-wise Levene's test shown in Table 4). Hence, we cannot compare the means of the samples directly. One reason for these differences could be that, in NA, typically there are several commercial alternative software products for a given FL/OSS product (often there is always at least one major commercialized/proprietary alternative for most FL/ OSS products). However, in countries such as China, this is not the case. We find that projects in these communities often do not compete with proprietary alternatives. Hence, while some participants are strongly driven by the need for a product (given the lack of alternatives in the market place), others are motivated by other factors. With respect to NA and India (both regions where software versions in English are dominant), the participants are relatively homogenous in functional needs. In summary, our findings show that there are strong differences in generic motives for participation. While all the three communities are strongly driven by sharing and learning, we find that North American participants were more driven by these motives than developers in China. In addition, financial concerns were more important to Chinese and Indian participants in comparison to North American FL/ OSS participants. Further, participants in the three samples varied considerably in their motivation to participate due to their functional needs. 4.2. Project preferences 4.2.1. Modular projects Variance differences are high across samples. We find strong sample variances across North America and India; as

183

well as between China and India (the statistical significance was not found in the North America–China comparison). Further, considering the standard deviations shown in Table 3, the India sample is less diverse compared to NA and China. In other words, Indian developers have a more homogeneous (and generally higher) outlook on modularity than their NA and China counterparts. This is not surprising since preferences for modular projects are closely related to experience in rigorous software design — Indian developers are much more mature in software engineering principles in comparison thanks to their experience working for the disproportionate numbers of globally-ranked CMM5 Level 4 and Level 5 firms in the country. 4.2.2. Large teams We find some country-level differences across the three samples. While we do not find differences at the mean level across the North American and Chinese samples, we find statistically significant differences in variances between China and India. Even though the variances are different across the two samples (at the p = 0.10 level), we find that participants from the China sample were drawn towards larger projects than the Indian counterparts. This could be attributed to several factors. In China, since the software industry is immature and nascent compared to India and NA, experienced developers are relatively scarce in the software industry. FL/OSS initiatives provide developers an otherwise difficult-to-find opportunity for learning. Further, only in a larger project can a new FL/OSS programmer “learn the ropes” as peripheral contributors suggesting design improvements [53]. 4.2.3. Universal projects We find significant pair-wise variance differences between North America and both China and India samples (we do not find differences in variance as well as the means in the China– India comparison). This suggests that the population is very heterogeneous in these samples ruling out the possibility of direct comparisons at the mean level. These variance differences in conjunction with the magnitude of standard errors in Table 3 together indicate that the NA community might be relatively homogeneous in comparison to China and India. Chinese and Indian participants tend to be diverse in their preference of universal projects (vs. local projects). In brief, project preferences differed considerably across the three samples of participants. Participants in NA were relatively homogeneous in their preference for universal projects, while Chinese and Indian participants varied considerably in their preferences. Participants in China were attracted to large FL/OSS initiatives. By contrast, Indian developers are relatively homogenous in their preference for modularly-designed projects.

5 The Capability Maturity Model I (CMM-I) specified by the Software Engineering Institute of Carnegie Mellon University has often been employed to rate the capabilities of software firms in the awarding of large government and private party software development contracts. As of 2006, close to 75% of the CMMI Level 5 software centers were reported to be based in India (SCAMPI Results: http://sas.sei.cmu.edu/pars/pars.aspx, last accessed on Mar 28th, 2008).

184

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

4.3. Relationship between generic motivations and project preferences 4.3.1. Sharing/learning and project preferences The sharing and learning preferences of participants are linked with all three project preferences, albeit in different forms in the three samples. Specifically, in the China sample, we find that participants driven by sharing and learning motives are more likely to participate in modular projects, in large teams, and in universal projects. This relationship is not evident in the NA and Indian samples. This finding is consistent with Baldwin and Clark [2]. The sharing and learning focus of participants in the China sample are evident in their project preferences. For instance, they expect modular projects to give them more opportunities for sharing their expertise as well as learning design and programming skills. Modular projects are relatively well managed and produce a usable product in a shorter time span, which should be of interest to participants driven by functional needs. Participants also viewed that projects developed by large teams provide more learning opportunities. These relationships were summed up by one participant who was strongly motivated by sharing and learning motives “… It is important for a programmer to have system design capabilities. And we can obtain such capabilities from (open source) practice … Open source software requires examination from a lot of developers (large projects) … the design is also changing in order to be perfect…” Further, universal projects (most likely aimed at global audiences) are perceived as providing greater opportunities for sharing and learning in comparison to local projects that are developed for the Chinese market. Given that we observe this relationship only in this collectivist culture [7], our results suggest that the role of modularity might be more pronounced in product development settings in cohesive groups. 4.3.2. Financial benefits and project preferences Indian participants driven by financial motives and career concerns are drawn towards modular projects, large team initiatives, as well as universal projects. This finding is in line with our expectation for developing countries. In India, with the burgeoning growth of the market for software skills and design expertise, developers benefit from participation in state-of-the-art global projects. For instance, competition for specialized skills is high in the growing software economy. Participation in FL/OSS may enable developers to differentiate themselves from others with similar skills and simultaneously signal their dexterity. This relationship is not as strong in the NA and China sample, even though the financial incentives-modular project preference link was evident in the China sample, confirming that in China as a developing country, at least one of the project preferences has a significant link with financial benefits as a generic motivation. 4.3.3. User needs and project preferences With respect to the relationship between motivation due to functional needs and project preferences, we find a weak relationship in the Indian sample but no significance in others. For instance, we find that participants who are involved in FL/OSS initiatives to satisfy a functional need were drawn towards modular projects. This result could also

stem from the strong software engineering background of participants found in the India sample. They are consistent with the view of Parnas [38] that modularity provides a means to hide irrelevant design aspects of each component from the user and it helps promote collaboration. Surprisingly, we find that Indian participants who were driven by functional needs did not prefer universal projects. This relationship, though, was statistically weak (as shown in Table 5). Such participants preferred to contribute to local projects. For example, one participant from this sample, who rated his motivation due to functional need as very high, noted that he had been and preferred working on local initiatives such as translations of software into Oriya, a language local to the eastern state of Orissa in India. The growth of the software industry in the Indian economy could be signaling opportunities in the local market, “gaps”, that might be filled by newly developed FL/OSS software. It could also be possible that the extensive promotion of FL/OSS software by the central and state governments (e.g. [46]) have increased the awareness of FL/OSS. The availability of talent pool, in conjunction with such governmental efforts, might be driving FL/OSS software efforts that address local needs. 4.3.4. Control factor (prior experience) We find evidence consistent with expectations. In the NA sample and the India sample, both of which can be considered relatively mature software development regions, we find that participants with significant prior experience tend to prefer participation in modular projects. In the China sample, we find that participants with limited prior experience tend to gravitate toward large team projects (likely due to potential learning opportunities). To sum up, we find strong relationships between the generic motivational factors and project preferences. We observe that both extrinsic as well as intrinsic motives might be associated with project preferences. Chinese participants who were driven by sharing and learning, as well as Indian participants driven by financial concerns, preferred modular, large and universal projects. 5. Concluding remarks In this paper, we propose a conceptual framework to study generic motivations and project-level preferences of FL/OSS developers in developed countries, represented by North America, and developing countries, represented by China and India. We provide theoretical arguments for the discussion of the underlying rationale for differences due to different cultural and economic backgrounds. We first investigate how three generic motivations noted in literature – sharing and learning, financial and career incentives, and user needs – affect developers' participation. Project-level preferences of developers and their relationships to the generic motivations are also analyzed. Based on our proposed conceptual framework, we designed an interview and a survey, and collected and examined responses from developers in the three regions. We believe that understanding the framework, as well as results from the preliminary empirical study, is important for making policy decisions related to FL/OSS, especially in developing economies such as India and China. For example, if sharing and learning is a strong motivational factor as we

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

have found, offering more FL/OSS-related educational opportunities will help improve developers' learning experience through participation in FL/OSS. In addition, we expect economic policies in developing countries favoring FL/OSSrelated businesses to have a positive impact on programmer participation. Our observation of financial motives from participation corroborates this viewpoint. Further, FL/OSS can be used as a means to provide exposure to systematic, collaborative software development, an experience that is hard to obtain except for those working for certain resourceful and leading edge multinational software companies. Developing country developers who are drawn to FL/OSS by both extrinsic and intrinsic motivations also prefer working in large, modular and universal projects, the same characteristics of projects developed in a selected few multinational software firms. This is quite different from developers in developed countries, where we observe no particular relationship between the two sets of motivational factors. Our results have important implications for managers and administrators of FL/OSS projects. First, to gain access to a larger pool of talented resources, organizations in developed countries that are looking to contract out software development could outsource projects to the FL/OSS community in developing countries. By providing and highlighting opportunities for learning, they could attract developers to work on such projects. Similarly, to draw more participants, administrators of FL/OSS communities need to create an environment conducive to learning. For instance, they can encourage senior developers to provide “hand-holding” help to their novice peers. Furthermore, since prospective participants are sensitive to how modular the design is and other project characteristics, administrators need to monitor progress and ensure that such favorable project characteristics are maintained until project fruition. Last, in developing countries like China and India, initiators and administrators of large and popular FL/OSS projects may need to better articulate benefits for participants' skill building to induce continued involvement of career-concerned developers. This study has certain limitations. First, given the small sample nature of our study, generalizability of results would be difficult. Hence, there is a need for a larger systematic study across a greater set of participants, as well as other developers who choose not to participate in FL/OSS. Second, we do not know the true contributions of these developers. Some contributors wrote new code, some were involved in planning, while others made contributions over prior-developed solutions. This information would allow us to study how motivations are related to the tasks they undertake. Third, gathering additional data from more countries would greatly enable us to make generalizable inferences. However, it is difficult to find the corresponding country-specific FL/OSS developer communities that would be necessary to provide comparable environments. Fourth, given the potential for non-response bias from participants in these communities, increasing the generalizability would require data collection efforts on a larger scale. One great challenge that we observed is that the prevalence of surveys of FL/OSS participants in the recent past has limited the willingness of many participants to share such information and opinions (often repeatedly) with researchers. Fifth, due to the need for a short length survey noted by many participants in our study, we used single-item measures for certain constructs, in addition to open-ended questions and items seeking data on their more recent projects to gather

185

confirmatory inputs from the participants. Multi-item constructs might have provided more robust and generalizable findings to complement our results. Finally, the representativeness of the sample is uncertain; it is possible that the current sample is not representative of the entire population of FL/OSS developers. The study gives preliminary insights into developer's motivation to participate in FL/OSS from different countries, which can be used to develop more in-depth studies. For example, now that we know programmer motivations differ in the three regions, it would be beneficial to learn what caused such differences: different cultural backgrounds of the developers, different wage structures and dispersions, or career goals or career development paths that exist for them? Also, it would be interesting to study if and to what extent the developers also participate in proprietary software development alongside their FL/OSS activities and compare the projects they work on with the FL/OSS projects to which they contribute given the potential competitive tensions across these two markets (e.g., [44]). Furthermore, complementary to this study, we could gauge the extent to which participants from one country ‘dominate’ a repository (i.e.., Sourceforge, HuiHoo, and Sarovar) in terms of their contribution. Likewise, it would be beneficial to learn about differences in project types and characteristics under such contribution asymmetries. References [1] A. Arora, A. Gambardella, The globalization of the software industry: perspectives and opportunities for developed and developing countries, NBER Innovation Policy & the Economy 5 (1) (2005) 1–32. [2] C.Y. Baldwin, K.B. Clark, The architecture of participation: does code architecture mitigate free riding in the open source development model? Management Science 52 (7) (2006) 1116–1127. [3] M. Blaug, The empirical status of Human Capital Theory: a slightly jaundiced survey, Journal of Economic Literature 14 (3) (1976) 827–855. [4] A. Bonaccorsi, C. Rossi, Why open source software can succeed? Research Policy, 32 (7) (2003) 1243. [5] F. Brooks, The Mythical Man-Month: Essays on Software Engineering, Addison-Wesley, Reading, MA, 1995. [6] C. Chiu, M. Hsu, E.T.G. Wang, Understanding knowledge sharing in virtual communities: an integration of social capital and social cognitive theories, Decision Support Systems 42 (3) (2006) 1872–1888. [7] CIO insight, Open Source software becomes everyday software, CIO Insight 15 (2005) 66–67. [8] T. Cottrell, B.R. Nault, Product variety and firm survival in the microcomputer software industry, Strategic Management Journal 25 (10) (2004) 1005–1025. [9] K. Crowston, J. Howison, H. Annabi, Information Systems Success in Free and Open Source Software Development: Theory and Measures, Software Process: Improvement and Practice, 2006, (Special Issue on Free/ Open Source Software Processes.). [10] J. Feller, B. Fitzgerald, A framework analysis of the open source software development paradigm, In Proceedings of the Twenty First International Conference on Information Systems (Brisbane, Queensland, Australia), International Conference on Information Systems. Association for Information Systems, Atlanta, GA, 2000, pp. 58–69. [11] J. Feller, B. Fitzgerald, Understanding Open Source Software Development, Addison-Wesley, Boston, 2002. [12] N. Franke, E. von Hippel, Satisfying heterogeneous user needs via innovation toolkits: the case of apache security software, Research Policy 32 (2003) 1199–1215. [13] R.A. Ghosh, R. Glott, B. Krieger, G. Robles, Free/Libre and Open Source Software: Survey and Study, FLOSS, Final Report, International Institute of Infonomics, vol. 30, 2002 http://floss.infonomics.nl/report/FLOSS_Final4.pdf (accessed on June 14, 2005). [14] J. Bessen, D.S. Evans, L. Lessig, B.L. Smith, Government policy toward open source software, in: R.W. Hahn (Ed.), AER Brookings Joint Center for Regulatory Studies Books and Monographs December, 2002. [15] I.-H. Hann, J. Roberts, S. Slaughter, Why developers participate in open source software projects: an empirical investigation, Proceedings Twenty-Fifth International Conference on Information Systems, 2004.

186

R. Subramanyam, M. Xia / Decision Support Systems 46 (2008) 173–186

[16] A. Hars, S. Ou, Why is Open Source software viable? A study of intrinsic motivation, personal needs, and future returns, Proceedings of the Sixth Americas Conference on Information Systems, 2000, pp. 486–490, Long Beach, California. [17] R. Hof, The power of us, Business Week (3938) (June 20 2005) http:// www.businessweek.com/magazine/content/05_25/b3938601.htm (last visited on August 14, 2008). [18] G. Hofstede, Culture's Consequences: International Differences in Work Related Value, Sage, Beverly Hills, 1980. [19] G. Hofstede, Culture's Consequences, Comparing Values, Behaviors, Institutions, and Organizations Across Nations, Sage, Thousand Oaks CA, 2001. [20] J.P. Johnson, Some economics of open source software. Unpublished working paper, 2000. [21] A. Kankanhalli, B.C.Y. Tan, K.K. Wei, M.C. Holmes, Cross-cultural differences and information systems developer values, Decision Support Systems 38 (2) (2004) 183–195. [22] S.M. Kerner, The Rapid Rise, Release of Asianux 2.0, August 25 2005 http://InternetNews.com, Available online at http://www.internetnews. com/entnews/article.php/3530091- (last visited March 4, 2006). [23] S. Kingstone, Brazil Adopts Open-source Software, BBC News, June 2 2005 Available at http://news.bbc.co.uk/2/hi/business/4602325.stm (last visited June 28, 2006). [24] M.S. Krishnan, R. Subramanyam, Quality dimensions in e-commerce software tools: an empirical analysis of North American and Japanese markets, Journal of Organizational Computing and Electronic Commerce 14 (4) (2004) 223–241. [25] J. Kuan, Open source software as consumer integration into production. Unpublished working paper (2000). [26] K. Lakhani, E. von Hippel, How open source software works: free user to user assistance, Research Policy 32 (6) (2003) 923–943. [27] K. Lakhani, R.G. Wolf, Why Hackers Do What They Do: Understanding Motivation and Effort in Free/Open Source Software Projects, Working Paper, MIT Sloan School of Management, September 2003. [28] J. Lerner, J. Tirole, Some simple economics of open source, Journal of Industrial Economics 5 (2002) 197–234. [29] L. Lessig, Open source baselines: compare to what? in: Robert Hahn (Ed.), Government Policy Toward Open Source Software (AEI-Brookings Joint Center for Regulatory Studies, 2002, pp. 50–68. [30] M. Li, Z. Lin, M. Xia, Leveraging the open source software movement for development of China's software industry, Information Technology and International Development 2 (2) (December 2004) 45–63. [31] I. Marson, Free CDs Spread Open Source in India, CNet News.com, May 25 2005 Available at http://news.com.com/Free+CDs+spread+open+ source+in+India/2100-7344_3-5720008.html?tag=nl (Last visited June 28, 2006). [32] A.H. Maslow, Motivation and Personality, 2nd ed.Harper and Row, New York, 1970. [33] Z. Meng, S.T. Lee, The value of IT to firms in a developing country in the catch-up process: an empirical comparison of China and the United States, Decision Support Systems 43 (3) (Apr. 2007) 737–745. [34] J. Messman, Adopting open source software, Chief Executive 204 (Dec. 2004) 32–34. [35] A. Mockus, R.T. Fielding, J.D. Herbsleb, Two case studies of open source software development: Apache and Mozilla, ACM Trans. Software Engg, Methodology 11 (3) (2002) 309–346. [36] P.D. Morrison, J.H. Roberts, E. von Hippel, Determinants of user innovation and innovation sharing in a local market, Management Science 46 (12) (2000) 1513–1527. [37] M. Nelson, R. Sen, C. Subramaniam, Understanding open source software development— a research classification framework, Communications of AIS 17 (12) (2006) 266–287. [38] D.L. Parnas, On the criteria to be used in decomposing systems into modules, Communications of the ACM 15 (12) (1972) 1053–1058. [39] J.W. Paulson, G. Succi, A. Eberlein, An empirical study of open-source and closed-source software products, IEEE Transactions on Software Engineering 30 (4) 246–256 (April 2004). [40] E.S. Raymond, Homesteading the Noosphere, 1998, available at http:// catb.org/~esr/writings/cathedral-bazaar/homesteading/ (last accessed on Sept 4, 2007). [41] E.S. Raymond, The Cathedral and the Bazaar: Musings on Linux and Open Source by an Accidental Revolutionary, O'Reilly and Associates, Sebastopol, California, 2000.

[42] J. Roberts, I.-H. Hann, S. Slaughter, Understanding the motivations, participation, and performance of open source software developers: a longitudinal study of the Apache projects, Management Science 52 (7) (2006) 984–999. [43] M.A. Rossi, Decoding the “Free/Open Source (F/OSS) Software Puzzle”: a survey of theoretical and empirical contributions, working paper (2004). [44] R. Sen, A strategic analysis of competition between open source and proprietary software applications, Journal of Management Information Systems 24 (1) (2007) 233–258. [45] S. Shah, Motivation, governance and the viability of hybrid forms in open source software, Management Science 52 (7) (2006) 1000–1014. [46] D.C. Sharma, Indian President Calls for Open Source in Defense, CNET news.com, July 7 2004 avaialbe at http://news.com.com/Indian+president+calls+for+open+source+in+defense/2100-7344_3-5259836.html (last visited July 20, 2006). [47] L.Y.M. Sin, G.W.H. Cheung, R. Lee, Methodology in cross-cultural consumer research: a review and critical assessment, Journal of International Consumer Marketing 14 (4) (2006) 126–144. [48] K.J. Stewart, A.P. Ammeter, L.M. Maruping, Impacts of license choice and organizational sponsorship on user interest and development activity in open source software projects, Information Systems Research 17 (2) (2006) 126–144. [49] A. Strauss, J. Corbin, Basics of Qualitative Research: Grounded Theory, Procedures, and Techniques, Sage Publications, Newbury Park, CA, 1990. [50] L.L. Tung, M.A. Quaddus, Cultural differences explaining the differences in results in GSS: implications for the next decade, Decision Support Systems 33 (2) (2002) 177–199. [51] E. von Hippel, The Sources of Innovation, Oxford University Press, New York, 1988. [52] D. Wheeler, Why Open Source/Free Software (OSS/FS)? Look at the numbers!, November 2004 Available at http://www.dswheeler.com/ oss_fs_why.html (Last visited August 26, 2007). [53] Y. Ye, K. Kishida, Toward an understanding of the motivation of open source software developers, Proceedings of 2003 International Conference on Software Engineering, May 3–10 2003, pp. 419–429, Portland, Oregon. [54] R.K. Yin, Case Study Research, Design and Methods, 2nd ed. Sage, Newbury Park, CA, 1994.

Ramanath Subramanyam is an assistant professor of Information Systems in the College of Business at the University of Illinois at Urbana– Champaign. He obtained his Ph.D. in 2004 from the University of Michigan. His research interests include Software Engineering and Project Management, Customer Integration in Software Development, IS/IT Outsourcing, and Business Value of Information Technology. His articles have been published (or are forthcoming) in Information Systems Research, IEEE Transactions on Software Engineering, Communications of the ACM, and Journal of Organizational Computing and Electronic Commerce.

Mu Xia is an assistant professor of Information Systems in the College of Business at the University of Illinois in Champaign, IL, USA. He received his Ph.D. from the University of Texas at Austin. He is interested in e-Business standards, online communities, and the formation and evolution of enterprise networks. His papers have been published in journals such as Journal of MIS, European Journal of Operational Research, International Journal of Electronic Commerce, and Electronic Markets. He is on the editorial board of Information Systems and e-Business Management. He is also the e-Business cluster co-chair in INFORMS Annual Meetings.