The Impact of Membership Overlap on Growth: An Ecological Competition View of Online Groups

OrganizationScienceVol. 24, No. 2, March–April 2013, pp. 414–431ISSN 1047-7039 (print) � ISSN 1526-5455 (online) http://dx.doi.org/10.1287/orsc.1120.0756

© 2013 INFORMS

The Impact of Membership Overlap on Growth:An Ecological Competition View of Online Groups

Xiaoqing WangDepartment of Decision and Information Technologies, Robert H. Smith School of Business, University of Maryland,

College Park, Maryland 20742, [email protected]

Brian S. ButlerCollege of Information Studies, University of Maryland, College Park, Maryland 20742,

[email protected]

Yuqing RenCarlson School of Management, University of Minnesota, Minneapolis, Minnesota 55455,

[email protected]

The dominant narrative of the Internet has been one of unconstrained growth, abundance, and plenitude. It is in thiscontext that new forms of organizing, such as online groups, have emerged. However, the same factors that underlie the

utopian narrative of Internet life also give rise to numerous online groups, many of which fail to attract participants or toprovide significant value. This suggests that despite the potential transformative nature of modern information technology,issues of scarcity, competition, and context may remain critical to the performance and functioning of online groups. Inthis paper, we draw from organizational ecology theories to develop an ecological view of online groups to explain howoverlapping membership among online groups causes intergroup competition for member attention and affects a group’sability to grow. Hypotheses regarding the effects of group size, age, and membership overlap on growth are proposed andtested with data from a 64-month, longitudinal sample of 240 online discussion groups. The analysis shows that sharingmembers with other groups reduced future growth rates, suggesting that membership overlap puts competitive pressure ononline groups. Our results also suggest that, compared with smaller and younger groups, larger and older groups experiencegreater difficulty in growing their membership. In addition, larger groups were more vulnerable to competitive pressure thansmaller groups: larger groups experienced greater difficulty in growing their membership than smaller groups as competitionintensified. Overall, our findings show how an abundance of opportunities afforded by technologies can create scarcity inuser time and effort, which increases competitive pressure on online groups. Our ecological view extends organizationalecology theory to new organizational forms online and highlights the importance of studying the competitive environmentof online groups.

Key words : online groups; organizational ecology; competition; membership overlap; online communitiesHistory : Published online in Articles in Advance June 15, 2012.

IntroductionInformation and communication technologies haveenabled novel forms of collective action and the self-organization of large and distributed collaborative groups(Sproull and Kiesler 1991). Online groups, for instance,provide virtual spaces where globally distributed peo-ple can interact around a shared purpose (Rheingold2000, Sproull and Arriaga 2007). Individuals join onlinegroups to exchange information, to interact with like-minded others, and to organize and participate in col-laborative work or collective action (Sproull and Arriaga2007). Businesses use online groups to build brandloyalty (Porter and Donthu 2008), facilitate peer-to-peer customer support (Jeppesen and Frederiksen 2006),and foster knowledge sharing and collaboration amongtheir employees (Constant et al. 1996). Developmentof Internet technologies has significantly reduced thecosts of creating online groups on various platforms,

ranging from traditional discussion forums and mail-ing lists to social networking systems and wikis. Withthese technologies, online groups of software develop-ers, Wikipedia editors, and problem solvers have suc-cessfully produced widely used software (Lee and Cole2003), millions of online encyclopedia articles (Kane2011), and innovative solutions to problems faced bybusinesses and nonprofit organizations (Jeppesen andLakhani 2010). Online groups have the potential to cre-ate highly beneficial outcomes by bringing together peo-ple in ways that were previously difficult or impossible.

The proliferation of online groups, although benefit-ing society in many ways, has also created tension forInternet users, who must allocate their time and atten-tion among the many groups that appeal to their interestsand passions. According to a Pew Internet report, thegrowth of online users has slowed down over the last fewyears (Pew Research Center 2008), whereas the number

414

Wang, Butler, and Ren: The Impact of Membership Overlap on GrowthOrganization Science 24(2), pp. 414–431, © 2013 INFORMS 415

of Internet offerings—sites to visit and groups to join—has grown exponentially. Internet users, therefore, needto choose where to spend their limited time and effort. Areport by Nielsen also showed that the growth in onlinesocial networking sites and blogs has taken time awayfrom other Internet outlets, such as search, general inter-est portals and communities, and email (Nielsen 2009),an indication that online users’ time and attention havebecome increasingly scarce resources. Online groupsmay face increasing pressure to compete for human timeand attention, and yet the existing literature has little tosay about how intergroup competition is likely to affectonline groups.

We propose that competition among online groupswill affect them in ways that are analogous to theimpact of competition among traditional organizationsfor resources, such as raw materials, clients, and labor(Hannan and Freeman 1977). Organizational ecologytheories suggest that when resources in the environ-ment are scarce and the number of organizations tap-ping into these resources increases, resources acquiredby one organization become unavailable to others, reduc-ing their performance and chances of survival. An indi-vidual’s time is a scarce and limited resource: time spentin one activity (e.g., child care) takes away time fromanother activity (e.g., work) (Becker 1965). When sev-eral online groups rely on the participation of the samemembers, i.e., they have membership overlap, the mem-bers’ time spent in one online group (e.g., a Facebookgroup for a political cause) takes time away from anothergroup (e.g., a Wikipedia project on U.S. politics), poten-tially reducing the performance of both groups.

At the same time, it is unclear whether insights fromorganizational ecology theories regarding resource com-petition apply directly to competition among onlinegroups. Compared to traditional, off-line organizations,online groups rely more on member contribution thanphysical or financial resources to succeed. Being onlineenables groups to recruit members from many geo-graphic locations, greatly increasing the pool of avail-able people. Online groups also have more permeableboundaries, more fluid membership, lower participa-tion barriers, and fewer switching costs, which providesopportunities for members to simultaneously participatein multiple groups (Dahlander and Frederiksen 2012).Although these differences may create a resource-richenvironment for online groups, they may also makeonline groups more vulnerable to intergroup competi-tion, because members can join and exit groups easily.Consequently, it is unclear whether organizational ecol-ogy theories of resource competition generalize to onlinegroups. For example, how does membership overlapwith other online groups affect group growth? How dogroup characteristics such as size and age affect growth?Do the effects of membership overlap vary in groupswith different characteristics?

In this paper, we draw from organizational ecologytheories of resource competition (Hannan and Freeman1977) to explain the effects of membership overlap,group size, and age on the growth of online groups.In doing so, our work makes three contributions. First,our study contributes new insights to understandingonline groups, which are a new form of organizing onthe Internet. Most studies on online groups have focusedon either individual motivations, perceptions, and behav-iors (Bagozzi and Dholakia 2006, Butler et al. 2007,Jeppesen and Frederiksen 2006, Moon and Sproull 2008,Nambisan and Baron 2010, Wasko and Faraj 2005) orthe structure, culture, and dynamics within these groups(Faraj et al. 2011, Faraj and Johnson 2011, Stewart andGosain 2006). Some research has identified the char-acteristics of individuals who perform better than oth-ers in online groups (Dahlander and Frederiksen 2012,Jeppesen and Lakhani 2010). Only a few have attendedto the environment within which an online group exists(Grewal et al. 2006, Gu et al. 2007). This paper adds tothe literature that considers the external environment ofonline groups by showing that, like traditional organiza-tions, online groups are subject to competitive pressurein their ecological environment. Organizations interestedin creating and hosting these online groups must con-sider both internal design and the impact of the largersystem of groups. For individuals and businesses inter-ested in hosting online groups, our results provide prac-tical guidance on how to position groups to minimizethe impact of competition.

Second, we extend and adapt organizational ecologytheories to new organizational forms online. Studiesof traditional organizations, such as hotels and creditunions, have shown that size, age, and overlap den-sity with other organizations affect organizational perfor-mance, failure, and growth (Barron et al. 1994, Carrolland Hannan 2000, Ingram and Inman 1996). Our studyconsiders whether the same variables have effects ononline groups that are similar to those experienced bytraditional organizations competing for resources. Byexamining these questions, our study seeks to addresswhether online groups are truly novel organizationalforms or whether, despite the potentially emancipatingaspects of modern information technology, the changeof context—moving from off-line to online—does notchange the fundamental principles that govern the inter-dependencies among social and organizational entities.

Third, this work studies a mature population ofonline groups. Whereas much is made of the nov-elty of the Internet and associated technologies, manyof the most prominent examples of online platformsare well-developed, mature social systems. Wikipediawas founded in 2001. SourceForge.net, a platform foropen source projects, was started in 1999. InnoCentive,a commonly referenced example of crowdsourcing,began in 1998. All online platforms eventually mature.

Wang, Butler, and Ren: The Impact of Membership Overlap on Growth416 Organization Science 24(2), pp. 414–431, © 2013 INFORMS

Although mature populations are not always the focus ofonline group research, understanding their dynamics andinterdependencies is of theoretical and practical signif-icance for those interested in online groups. Moreover,organizational ecology theories suggest that competitionis most important when there are already many organi-zations in a population. Focusing on a mature populationwith many online groups—Usenet between 1999 and2005—allows us to characterize the effects of intergroupcompetition when they are most likely to matter.

The rest of this paper is organized as follows. In thenext section, we review the organizational ecology lit-erature; discuss how the literature may apply to onlinegroups; and propose hypotheses about how memberoverlap density, group size, and group age jointlyaffect online group growth. In the Methods section, wedescribe the data set gathered from 240 Usenet groupsand the measures constructed to test the hypotheses.We then present our analysis, results, and a discussion ofthe theoretical and practical implications of the findings.

Theory and HypothesesOrganizational Ecology LiteratureOrganizations need resources to survive and grow. Orga-nizational ecology theories suggest that organizationsdepend on their environment for these resources. Thisdependence requires organizations to simultaneouslycoexist and compete with other organizations. Whena new population of organizations emerges, the exis-tence of similar organizations provides legitimacy andopportunities for organizations to learn from each other(Aldrich and Ruef 2006). The coexistence of similarorganizations increases their survival rate and decreasestheir failure rate. At the same time, dependence oncommon, limited resources in the same population orenvironment (land, physical materials, labor, customers,etc.) can lead to competition among these organizations(Carroll and Hannan 1989, Hannan and Freeman 1977).When resources are scarce, there is an upper bound tothe number of organizations an environment can sup-port, known as its carrying capacity (Popielarz and Neal2007). Competing with many organizations for limitedresources decreases an organization’s likelihood of suc-cess or survival. Under these conditions, the growth ofone organization limits the growth of others.

The extent to which two organizations rely on com-mon resources depends on the degree to which theirniches overlap. A niche in organizational ecology refersto the location of an organization or a population oforganizations in a multidimensional space, defined bythe resources that the organization or population needs tosurvive (Hannan and Freeman 1977, McPherson 1983,McPherson et al. 2001). For instance, book publish-ers, newspaper publishers, and cardboard container firmsall compete for the same paper supplies; hence, they

share a niche. In contrast, regional restaurant chainsmay occupy different niches if they rely on facilitiesin distinct, nonoverlapping geographic areas. Most earlystudies in organizational ecology were conducted at thepopulation level, assuming that organizations in the samepopulation occupied the same niche and thus requiredthe same resources. These studies considered populationdensity—the number of organizations in a population—as a proxy for the number of organizations competing forcommon resources, and they examined its effect on thefounding and failure rate within the population. Otherwork suggested that organizations in the same popula-tion could occupy different niches. For instance, day-care centers that provide services for infants and thosethat provide services for toddlers require different staffand facility resources and do not directly compete withone another (Baum and Singh 1994). These studies wereconducted at the organization or niche level, and theyfocused on overlap density, which is the number of orga-nizations with overlapping niches as the focal organiza-tion weighted by the extent of niche overlap betweenthem. The more organizations in the same niche or occu-pying overlapping niches, the higher the demand forcommon resources.

The impact of either population density or overlapdensity on organizational performance depends on therelative strength of the beneficial and competitive effectsof coexistence. Prior work suggests that the benefits ofcoexistence, such as learning and legitimacy, decrease,and the effects of competition increase as more organi-zations enter a niche or as population or overlap densityincreases (Aldrich and Ruef 2006). Therefore, popula-tion or overlap density has a curvilinear effect on orga-nizational performance: when a population is emergingand density is low, the benefits of coexistence will out-weigh the negative effects of competition, so a marginalincrease in density leads to higher performance. When apopulation is mature and density is high, the effects ofcompetition dominate those of coexistence, and increas-ing density leads to lower performance (Kuilman andLi 2006).

When resources are limited, organizations in the sameenvironment differ in their ability to survive, succeed,and grow. Most existing research has found that largeorganizations have an advantage over small ones. Largeorganizations tend to survive longer not only becausethey have more resources and external ties (Bercovitzand Mitchell 2007) but also because they are better ableto acquire additional resources as needed. Consistentwith this view, most studies have shown a positive effectof size on organizational survival and a negative effect ofsize on failure, which implies a “liability of smallness”(Freeman et al. 1983, p. 692).

Existing literature also suggests several possibleeffects of an organization’s age, after size is controlledfor (Hannan 1998, Kitts 2009). Some studies show a


“liability of newness” effect (Freeman et al. 1983): neworganizations are more likely to fail because they donot have as many resources and capabilities to competewith others. Other studies suggest a “liability of adoles-cence” effect: the risk of failing increases initially andthen decreases later as organizations mature (Bruderland Schussler 1990, Fichman and Levinthal 1991). Fur-ther studies suggest a “liability of aging” effect: oldorganizations’ resources and capabilities decline overtime, and they are more subject to inertia, making themmore vulnerable to changes in the environment (Kitts2009, Sorensen and Stuart 2000). One way to recon-cile these seemingly contradictory findings is to considerthe coexistence of these effects at different stages of anorganization’s life: the liability of newness effect maydominate in early stages of an organization’s develop-ment, whereas the liability of aging effect may dominatein mature stages (Kitts 2009). Overall, organizationalecology theories focus on an organization’s need toacquire scarce resources from its environment to explainhow characteristics of the organization (such as size andage) and those of its environment (such as density ofcompeting organizations) combine to affect its perfor-mance and survival.

Online Groups, Membership Resources, andOrganizational EcologyOn one level, online groups are comparable to orga-nizations in terms of being “goal-directed, boundary-maintaining, and socially constructed systems of humanactivity” (Aldrich and Ruef 2006, p. 4). They areoften organized around a common purpose (e.g., Googlegroups about pets or Wikipedia projects) (Sproull andArriaga 2007). Online group infrastructures provideboundary-maintaining mechanisms to distinguish mem-bers from nonmembers and sometimes limit groupcontent to its members (Butler and Wang 2012). Liketraditional organizations, participants’ activities shapeand are shaped by the purpose, norms, and values ofonline groups (Stewart and Gosain 2006). Hence, it isreasonable to expect that organizational theories, includ-ing organizational ecology theories, might be applicablefor explaining the nature and consequences of interde-pendencies among online groups.

At the same time, online groups differ from tradi-tional organizations in several regards, which may be atodds with the fundamental assumption of resource com-petition in organizational ecology. To support operationsand survive, traditional organizations must acquire anduse a variety of resources, such as physical assets, natu-ral resources, labor, and financial capital (Bercovitz andMitchell 2007). Online groups lack the physical facil-ities, legal standing, and formal relationships of tradi-tional organizations, and thus their resource needs arelower. With the development of new technologies and

platforms, an online group can be created and man-aged with little or no investment in computer hardwareor software and access to scarce capital, materials, andlabor, and it is less likely to be affected by competi-tion for those resources. However, online groups remaindependent on their environment for resources that arecritical to their success and survival. Their most valuableresources are members’ time and effort (Butler 2001).Volunteer leaders promote groups, manage infrastruc-ture, and facilitate interaction among participants (Butleret al. 2007). Members provide information, answers,and social support for one another (Kollock and Smith1999), and they bring passion to their groups that drivesparticipation and contribution (Faraj et al. 2011). With-out members’ time and effort, online groups would notcontinue to exist and function. Thus, although onlinegroups may have less need for traditional organizationalresources, they remain dependent on their environmentfor the critical resource of members.

A second way that online groups may differ fromtraditional organizations is with respect to resourcescarcity. Traditional organizations operate under the con-straint of scarcity. Many of the critical resources atraditional organization may depend on are limited,including financial capital, physical facilities, raw mate-rials, and skilled employees (Hannan and Freeman1977). As a result, organizations are forced to competewith others in need of those same resources. In contrast,the environments in which online groups operate pro-vide access to a seemingly abundant population of users.Global reach, relatively low costs, and ease of use allcontribute to a dramatic increase in the number of peo-ple available to participate in online groups. Moreover,member resources are not exclusive in online environ-ments. Typically, resources acquired by one organizationare no longer available to others. A location used byone hotel cannot be used by another hotel at the sametime. A technical expert employed by one firm is notavailable to advise a competing organization. Resourceexclusivity creates competitive pressure that affects theperformance and survival of organizations. However, inonline groups, membership is rarely exclusive. Mem-bership in online groups is often open and fluid (Farajet al. 2011). It is common for people to participatein multiple groups to meet different needs (Dahlanderand Frederiksen 2012). A cancer patient, for example,may join one group to access information on his or hercondition and another group to socialize (Shaw et al.2000). A musician may visit multiple online groups totry out new ideas and receive feedback from a largeand diverse audience (Dahlander and Frederiksen 2012).Together, the reduction in geographic and economic con-straints and the lack of resource exclusivity online createa resource-rich environment in which online groups canpotentially exist without concern for limits on the avail-ability of potential participants.


Low traditional resource needs, potentially globalreach, and a lack of exclusive membership all have thepotential to reduce the competitive pressures faced byonline groups. As a result, it may be argued that onlinegroups will not be affected by competition for scarceresources. However, each of these factors also has thepotential to contribute to competition. Although com-munication technologies can significantly increase thenumber of accessible people, the population of onlineusers will not grow forever. The number of new userscoming onto the Internet has been declining in recentyears (Pew Research Center 2008). At the same time,the number of new sites, services, and activities availableonline has drastically increased, causing individuals todrop out of some services (Nielsen 2009). These trendssuggest that the number of individuals available to joinany particular system or type of activity is not likely tobe unbounded. As platforms like Facebook and Twitteremerge, older platforms like MySpace have seen stag-nant growth or even decline (MacMillan 2009). Thus,whether at the level of the entire online user popula-tion or within particular platforms, online groups are stillsubject to constraints on the number of available partic-ipants. Even if the global population of Internet userscontinues to grow, when individuals can join multiplegroups, resource scarcity and competition remain presentat a finer level. Online groups require time and energyfrom engaged participants to remain viable and success-ful (Butler 2001), and individuals’ time, energy, andattention is limited (Becker 1965). For each individual,time spent in one group takes away time that could bespent contributing to another group. Low setup costs anda lack of physical constraints mean that the population ofonline group “shells” can easily increase. Global reachmeans that individuals can choose from groups createdanywhere in the world. As the available groups increaseand the demand from multiple groups exceeds the finitetime and effort that individuals have to spend on onlinegroups, users may terminate their affiliations with somegroups with little or no direct costs. The same factorsthat create the impression of abundance and unfetteredgrowth in online environments have the potential to cre-ate scarcity and competitive pressures analogous to thoseexperienced by traditional organizations. Therefore, it isnecessary to carefully consider if, and how, competitionfor members’ time and effort factors into the ability ofan online group to survive and grow.

An Ecological View of Online GroupsApplying organizational ecology theory to the onlinecontext, we propose an ecological view of onlinegroups to examine the presence and implications ofintergroup competition for member resources. Althoughmany online group studies have considered how individ-ual motivations and other characteristics affect individ-ual willingness to participate in online groups (Bagozzi

Table 1 Mapping Organizational Ecology Constructs in theContext of Online Groups

Constructs Manifestations

Environment Other online groups, relationships among thegroups, and people with different interests

Population Set of online groups with a similar technologyplatform (e.g., Usenet or Facebook groups)

Resources Members and their time and effortNiche overlap Membership overlap among online groups

within a populationOverlap density Number of other online groups in a population

that share common members with a focalgroup

Performance Membership growth

and Dholakia 2006, Wasko and Faraj 2005), little or noattention has been given to an online group’s environ-ment and how it affects the performance of the group.Online groups, like organizations, do not exist in a vac-uum. They must acquire members from the pool ofpotential participants in the larger environment to sus-tain their day-to-day activities. Organizational ecologyprovides a potentially useful basis for explaining rela-tionships between an online group’s environment and itsperformance. Table 1 summarizes how we map core con-cepts of organizational ecology in the context of onlinegroups.

Studies in organizational ecology have examined manyoutcomes of resource competition including bankruptcy(Barron et al. 1994), performance and growth (Carrolland Hannan 2000), alliance strategy (Nam et al. 2010),and, most commonly, founding and failure rates (Barronet al. 1994, Dowell and Swaminathan 2000). Becauseof their status as legal entities, organizations often haveclearly defined creation and termination events. Further-more, because of their heavy reliance on physical andfinancial resources, traditional organizations typicallyexperience a clear break in operations when they run outof these resources. Unlike traditional organizations, fail-ure of online groups is rarely clearly defined. Becausethe expense of maintaining a virtual space is minimal, itis common for an online group to continue to exist asa technical structure long after it has ceased function-ing as a meaningful social entity. Although membershipgrowth (or decline) may not be equated with success inall online contexts, continued decrease in membershipsignificantly affects the activities of any group regard-less of its intended goals and purposes. Thus, although itmay be possible in some contexts to identify terminationevents, measures of growth and decline in membershipare more generalizable indicators of the health of onlinegroups than measures of online group “deaths.”

Characterizing the effects of an online group’s com-petitive context requires identification of competinggroups. One common approach in the organizationalecology literature is to characterize a competitive envi-ronment in terms of the firms that share a common


organizational form, such as hotels (Ingram and Inman1996), day-care centers (Baum and Singh 1994), orcredit unions (Barron 1999), within a geographic region.Online groups are created on particular technologicalplatforms (e.g., Usenet, Facebook, Wikipedia). Theseplatforms affect the form of the groups and the intercon-nections among them (Butler and Wang 2012). As such,groups created on the same platform, such as Usenetnewsgroups or Wikipedia projects, can be seen as con-stituting identifiable populations of online groups.

However, population-level analysis is of limited util-ity for explaining outcomes at the group level. Particularorganizations are rarely subject to competitive pressurefrom every firm in a given population. The num-ber of organizations in a population requiring mutu-ally exclusive resources may have no effect or evena positive effect on a particular organization (Baumand Singh 1994). Therefore, organizational ecologyresearchers have used overlap density to characterizean organization’s competitive environment (Barnett andMcKendrick 2004, Baum and Singh 1994, Podolny et al.1996). Overlap density refers to the number of otherorganizations in a population that share a resource nichewith the focal firm (Baum and Singh 1994). Over-lap density is weighted by the extent of niche overlapbetween two organizations: the number of organizationsthat rely on 100% of the same resources as the focalorganization is directly counted, and the number of orga-nizations that rely on only some of the same resourcesis weighted by the extent to which their niches overlapwith the focal firm. For example, an organization thatshares 100% of its resources with firm A and 50% withfirm B has an overlap density of 1 + 005 = 105. Definedin this way, overlap density reflects the level of compe-tition for the resources that a focal firm needs.

Consistent with the organizational ecology literature,we define member overlap density as the number ofonline groups in a population that share members witha focal group, weighted by the degree of membershipoverlap between each pair of groups. Membership over-lap may be a result of multiple groups having simi-lar content or purposes (e.g., three groups focusing oncar purchasing) or a result of people’s diverse interests(e.g., the same person participating in both a parentinggroup and a technical support group). As noted above,the most critical resource for online groups are mem-bers’ time and effort. Online groups with membershipoverlap depend on the same resource: the shared mem-bers’ time and effort. The extent to which two groupsshare members reflects the extent of their niche overlap.Hence, member overlap density provides a measure ofthe intensity of competition for member attention that afocal group faces.

Organizational ecology theories suggest that resourcecompetition will be most prevalent when a populationis mature and density is high (Aldrich and Ruef 2006).

Although online groups may face competition for mem-bers at all times, the impact of intergroup competitionwill likely be most significant in large, mature popu-lations where there are already many well-establishedonline groups. Therefore, in the next section we developa series of hypotheses regarding resource competition inthe context of a mature population of online groups.

Research Hypotheses

Member Overlap Density and Membership Growth.Organizational ecology theories predict a negative effectof density on organizational performance and survivalin mature populations. In a well-established population,density, or the number of organizations requiring com-mon resources, is already high, and resources are scarce.The competition effect of density dominates its positiveeffect, and therefore, increasing density hinders survivaland growth (Carroll and Hannan 2000). There is abun-dant evidence in the organizational literature that sup-ports this competitive effect. For example, Podolny et al.(1996) found that the number of semiconductor com-panies relying on the same patents as a focal companyreduced the growth of the company. Similarly, Baum andSingh (1994) found competitive effects among day-carecenters that accept children of similar ages. Competitionfor scarce resources is often believed to be the reasonbehind industrial declines (Ruef 2004).

Member overlap density is expected to have similareffects on online groups’ growth. In a mature popula-tion of online groups, the population is crowded andresources are scarce. Online groups with overlappingniches would need to compete for members’ limited timeand attention, as members affiliated with multiple groupsstruggle to allocate their efforts among them. At thistime, increasing overlap density makes it more difficultfor any particular online group to retain existing mem-bers and to recruit new participants (McPherson andRotolo 1996). Thus, we expect a negative effect of mem-ber overlap density on the growth of online groups inmature populations.

Hypothesis 1 (H1). Member overlap density is neg-atively associated with subsequent membership growthin a mature population of online groups.

Organizational ecology suggests that the effect ofdensity is nonlinear as a population develops. Legiti-macy increases and competition intensifies when den-sity increases (Hannan and Freeman 1977). Together,these effects account for the curvilinear effect of den-sity on organizational performance found in studies oftraditional organizations (see Nickel and Fuentes 2004for a review). For instance, Barron et al. (1994) foundan inverse U-shaped relationship between the local den-sity of New York City credit unions and their growthrate. Dowell and Swaminathan (2000) also found that


the density of American bicycle producers first reduced,and then increased, firm mortality rates. For mature pop-ulations, it is the second portion of the inverted U shape,in which density decreases organizational performanceat an increasing rate, which is relevant. In mature popu-lations of online groups, we expect the negative effectsof overlap density on growth to increase as more groupsshare members with the focal group. As member overlapdensity increases, competition between a focal group andthe other member-sharing groups may become fiercerand have larger negative effects on growth. The initialset of groups that compete for members with the focalgroup would have a small, negative impact on groupgrowth. Additional groups sharing members with a focalgroup are expected to reduce growth more than the ini-tial ones.

Hypothesis 2 (H2). Member overlap density reducessubsequent membership growth at an increasing rate ina mature population of online groups.

Group Size and Membership Growth. Organizationalresearchers have often argued that large organizationshave an advantage over small ones when they com-pete in the same environment because they have moreresources, more efficient routines, better external rela-tionships (Bercovitz and Mitchell 2007), and better abil-ity to attract additional resources such as skilled laborand capital (Audia and Greve 2006). In their study of thepopulation of hard disk drive manufacturers, Barnett andMcKendrick (2004) found that large manufacturers wereless likely to fail than small ones. Similarly, Ranger-Moore (1997) observed that large life insurance compa-nies were less likely to fail than small ones. However,others have pointed out that large organizations mayhave more difficulty growing than small ones becausethey have less room and motivation to grow (Greve2008). Large organizations may have reached an opti-mum size that minimizes costs (Barron et al. 1994),or they may find it difficult to adapt to their environ-ments and grow (Barron 1999). Reflecting on these con-flicting arguments, organizational studies to date haveshown mixed results on the relationship between sizeand growth rates. Some have found a negative relation-ship (Barron et al. 1994, Greve 2008), some have founda positive relationship (Banaszak-Holl 1991), and oth-ers have found no significant relationship between them(Bottazzi et al. 2002).

Both the advantages and disadvantages of being largemay manifest in the context of online groups. Largeonline groups have more member resources and as aresult can offer more content, making them potentiallymore attractive to current and potential participants thansmall groups (Butler 2001). Larger online groups alsoare more likely to be able to find volunteer leaders toperform administrative work, such as mentoring new

members, reinforcing norms, and maintaining infrastruc-ture (Butler et al. 2007), all of which would facili-tate the functioning of a group. However, large onlinegroups also suffer from social loafing problems (Kraut2003), which reduce participants’ motivation to con-tribute. For example, Jones et al. (2004) found that largeUsenet groups have lower response-to-post ratios thansmall ones. Online groups may also have an optimumsize beyond which growth would slow. Above a cer-tain point, continued growth of online groups createsinformation overload that can outweigh the marginalresource benefit of adding additional participants (Rabanet al. 2010). As an online group grows, members canbecome overwhelmed by both the amount of informa-tion and the number of members and thus be more likelyto leave. These effects are reflected in efforts made bysome online groups to limit the joining of new mem-bers (Shirky 2003) and the practice of creating sub-groups when an online group grows too large (Kim2000). Therefore, although size provides some benefitsfor the functioning of online groups, we expect its effecton membership growth to be negative. We posit thefollowing.

Hypothesis 3 (H3). Group size is negatively asso-ciated with subsequent membership growth in onlinegroups.

Few studies in organizational ecology have consideredthe interaction between organizational demographics andoverlap density. However, competitive pressure does notaffect all organizations equally (Barron 1999) and mayhave a differential effect on online groups as well. AsBarron (1999) argued, when the number of organizationsin a population increases and competition intensifies, itshould be more difficult for all organizations to obtainnew resources and grow. Large organizations may facea greater challenge than smaller ones if they find it dif-ficult to adapt. His analysis showed that density reducesthe growth rate of large organizations more so than thegrowth rate of small organizations. Following Barron(1999), we expect that large online groups would suffermore in growth than small ones when they share mem-bers with many other groups, partially because they arealready close to or beyond their optimal levels and par-tially because of their loose social structure and connec-tions among members. According to McPherson et al.(1992), membership turnover in voluntary groups is neg-atively associated with the number and strength of net-work connections within the group. Large groups, bothonline and off-line, typically have more diverse com-position and less cohesive networks than small groups.Members who are loosely connected to a group aremore likely to leave when they are distracted by con-tent from or connections with other groups. As a result,large online groups are more likely to lose members toother groups when they share members with many othergroups. Therefore, we posit the following.


Hypothesis 4 (H4). Member overlap density hasgreater negative effects on subsequent growth of largeronline groups than on subsequent growth of smalleronline groups.

Group Age and Membership Growth. The existing lit-erature suggests that in a mature population, age wouldreduce the performance of an organization at an increas-ing rate. A mature population has many well-establishedorganizations that have accumulated resources and capa-bilities by establishing connections with the environmentand developing a stable internal structure (Stinchcombe1965). Those very capabilities, routines, and structuresmay become outdated over time, and it is often moredifficult for old organizations to adjust and adapt to envi-ronmental changes than it is for young ones (Thornhilland Amit 2003). When that happens, old organizationsare more likely to decline than young ones (Freemanet al. 1983), and their vulnerabilities increase over time.Even among the newer organizations in the popula-tion, the difficulty of establishing internal structures andbuilding relationships is reduced because the organiza-tional form in a mature population is already estab-lished and legitimate. Therefore, the negative effect ofaging should dominate its positive effect in a maturepopulation.

Age may have similar effects on online groups as onoff-line organizations. In a mature population of onlinegroups, more groups are old and well established and aretherefore subject to the negative consequences of aging(Sorensen and Stuart 2000). The fast member turnover inonline groups may make it difficult for groups to retainthe capabilities that they have developed like traditionaloff-line organizations do (Faraj et al. 2011). Even whenthe norms, values, and policies of an online group areretained, they may become obsolete as the group ages,making older groups less able to maintain growth thannewer ones. The older an online group is, the more theymay suffer from their inflexibility. Thus, we expect thatthe effect of online group age on membership growthwithin a mature population will be negative, and thatgroup age reduces membership growth at an increas-ing rate.

Figure 1 Research Model and Hypotheses

H4: –H1: –H2: –

Member overlap density

Size

Membership growth

Age

H6: –

H3: –

H5: –

Hypothesis 5 (H5). Group age is negatively associ-ated with membership growth in a mature population ofonline groups, and it reduces membership growth at anincreasing rate.

As with the moderating effect of size, we hypoth-esize that group age moderates the effects of mem-ber overlap density on membership growth. In theory,the moderating effects could either exacerbate or alle-viate the negative effects of member overlap density.Old online groups are likely to have more committedmembers and more established networks among mem-bers, which buffers the groups from competitive pres-sure. At the same time, old online groups are likely tohave members who are burned out from past participa-tion (Cress et al. 1997) or who no longer need benefitsfrom the group, thus exacerbating the effects of competi-tion. Here, we follow Barron’s (1999) argument that anydisadvantage older groups have, compared to youngergroups, will be exaggerated as competition intensifies.Because we expect age to have a negative effect onmembership growth in mature online groups, we positthe following.

Hypothesis 6 (H6). Member overlap density hasgreater negative effects on the growth of older onlinegroups than on the growth of younger online groups ina mature population.

Figure 1 illustrates our complete research model. Fora mature population of online groups, we expect mem-ber overlap density to reduce membership growth atan increasing rate. We expect large and old onlinegroups to have slower growth than smaller and youngergroups, and we expect that group age reduces member-ship growth at an increasing rate. We also expect largeand old groups to be more vulnerable to competitionfrom membership overlap than small and young groups.Together, these arguments present an ecological theoryof online groups in terms of how intergroup competitionfor members interacts with online group characteristicsto affect growth.


MethodsDataWe tested the proposed hypotheses in the context ofUsenet newsgroups. First developed in 1979, Usenet is atext-based, peer-to-peer Internet communication infras-tructure that hosts online discussion groups called news-groups (Kollock and Smith 1999). In the early days,one needed a newsreader application to access Usenetmessages, just as one needed an email client to accessemails and an Internet relay chat (IRC) client to par-ticipate in IRC chat. In 2001, Google acquired thelargest Usenet archive service, Deja.com, and enabledbrowsing and posting to Usenet newsgroups through theWeb on Google Groups. Usenet was one of the old-est and most important communication channels in the1990s, together with email, discussion lists, and IRCchat (Kollock and Smith 1999). It hosted Tim Berners-Lee’s announcement of the first World Wide Web projectand the birth of the Linux open source software project.Our data show that as of 2005, there were approximately189,000 Usenet newsgroups and over 9 million partici-pants who had contributed to at least one Usenet news-group. Because of its unique technology and early pop-ularity, Usenet constitutes a distinct, mature populationof online discussion groups.

Figure 2 shows the number of newsgroups and Usenetmembers during our studied period. The number ofnewsgroups in Usenet has increased, especially duringthe second half of the studied period (Figure 2, left),suggesting an increasing demand for active participants.However, the number of Usenet members (Figure 2,right) has been declining since 2002, probably as a resultof competition from other interactive platforms (Herring2004) such as Web-based discussion forums and socialnetworking sites. These trends are consistent with ourcharacterization of Usenet as a mature population, withinwhich a growing number of online groups compete forincreasingly scarce member resources. As such, Usenet

Figure 2 Number of Usenet Newsgroups (Left) and Number of Members (Right) from October 1999 to January 2005

2040

6080

100

120

Num

ber

of n

ewsg

roup

s (in

thou

sand

s)

Month

Number of Usenet newsgroups

01,

000

2,00

03,

000

4,00

0

Use

net m

embe

rs (

in th

ousa

nds)

Jul 1999 Jan 2001 Jul 2002 Jan 2004 Jul 2005Jul 1999 Jan 2001 Jul 2002 Jan 2004 Jul 2005

Month

Number of Usenet members

is an appropriate setting to study the impact of compe-tition on online groups within a mature population.

We collected data on Usenet newsgroups from theMicrosoft Netscan database (Smith 2005). The Netscansystem collected information about Usenet newsgroups,their members, and message activities starting inSeptember 1999. In Usenet, there is no formal require-ment to sign up to become members of a group—thusno strict sense of “membership.” Individuals become vis-ible only when they post messages. Lurkers who readmessages without posting remain invisible. We definemembers of a Usenet newsgroup as participants whopost messages to the group, and we use this definition tomeasure our key constructs, such as group size, memberoverlap density, and growth rate.

We used several criteria to select the sample fromthe Netscan database. Adapting the classification inRidings and Gefen (2004), we first drew a stratifiedrandom sample of 400 newsgroups covering four broadtopics: hobby, technology, issue discussion, and healthsupport. Examples of the sampled newsgroups includealt.pets.ferrets (hobby), microsoft.public.security (tech-nology), talk.politics (issue), and alt.support.diabetes(health). Then, based on our observation of typicalactivities in Usenet, we removed groups with an aver-age of fewer than 20 posts per month to excludegroups that never attracted enough member participa-tion to be viable. After reviewing group descriptionsand message content in the Google Groups archive(http://groups.google.com), we further excluded groupsthat were mostly spam, announcements, or not Englishbased. Our final sample includes 240 newsgroups with107 hobby groups, 64 technology groups, 38 issuediscussion groups, and 31 health support groups. Weobtained monthly activity data for these groups and theirmembers from the Netscan database and retrieved theirstarting months from the Google Groups archive. Totest the hypotheses, we constructed a panel data set forthe 240 newsgroups over 64 months from October 1999


to January 2005. A few groups started after October1999, and they have fewer than 64 observations in thedata set. The final data set includes 14,929 group-monthobservations.

MeasuresGrowth Rate. Following the method in Podolny et al.

(1996), we measured each newsgroup’s monthly mem-bership growth by dividing the total number of activemembers in the current month (month t) by the totalnumber of active members in the previous month (montht−1). On the rare occasions when there was no activity,and thus no active members in a month t, the growth ratein month t+1 was set to be equal to the number of activemembers in month t + 1. The calculated growth rateranged from 0 to 316. Because this variable was highlyskewed, we took the natural log to improve its normal-ity. To ensure that results were not unduly influencedby highly transient participants, we also constructed analternative measure of growth rate by counting the num-ber of core members (instead of all members), defined asmembers who had posted a total of at least 11 messagesin the group. We used 11 messages as the thresholdbecause only 20% of newsgroup participants posted 11messages or more during their lifetime in their groups.Analysis of the growth of core members yielded simi-lar results to the analysis of the growth of all members.Thus, we report the results with the growth rate of allmembers in the paper.

Member Overlap Density. Prior work suggested thatboth the number of organizations in a niche and thedegree of niche overlap affect the intensity of compe-tition (Barnett and McKendrick 2004, Baum and Singh1994, Podolny et al. 1996). Following prior studies inorganizational ecology, we constructed a weighted mea-sure of member overlap density to account for both thenumber of newsgroups that shared members with a focalgroup and the degree of membership overlap betweenthe focal group and the other groups. Online group mem-bership can change quickly as members move in andout of groups. Therefore, we measured member overlapdensity at a monthly interval.

We considered two newsgroups as sharing a mem-ber if the member posted at least one unique messagein both of the newsgroups in a given month. Individu-als who only cross-posted messages (i.e., did not posta unique messages in multiple groups) were not con-sidered to be shared members. Posters who participatedin more than 150 groups in a month were excludedbecause they were most likely software systems ratherthan individuals. For each of the newsgroups in our sam-ple, we identified all other newsgroups in the Netscandatabase that shared at least one member with it in amonth, including the newsgroups that were not in thesample. We then counted the number of members that

the focal group shared with each of them. The degreeof membership overlap between the focal group i andanother group j in month t was calculated as the numberof shared members between the two groups divided bythe total number of posters in the focal group i in thatmonth. Member overlap density was then calculated asthe sum of the degree of membership overlap betweenthe focal group and each of the other groups with whichit shares members. For example, if a newsgroup shares10% of its members with one group and 15% of itsmembers with another, its member overlap density iscalculated as 001 + 0015 = 0025. The higher the numberis, the higher the competitive pressure the focal groupfaces from other groups:

Member overlap densityit =J∑

j=1

#SharedMembersijt#Membersit

0

Group Size. We measured a newsgroup’s size as thetotal number of active members who posted at least onemessage to the group in a given month. Sizes of orga-nizations have been measured in many ways, includ-ing sales, scope, assets, capacity (e.g., storage capacityof wineries or room counts of hotels), scale of opera-tions (e.g., number of subscribers of telephone compa-nies or assets of life insurance companies), and numberof employees (Barron et al. 1994, Bercovitz and Mitchell2007). Because online groups acquire resources throughtheir active member base, we measured group size asthe total number of its members (Butler 2001). As wementioned earlier, people who read messages withoutposting were not considered a member and were thusnot counted in this measure of group size.

Group Age. We measured group age as the numberof years that have elapsed since a group’s creation. Fora group created in September 1999, its age was codedas 1 from September 2000, 2 from September 2001, andso on. The Netscan database began collecting data inSeptember 1999 and did not have starting-month data fornewsgroups created before then. For groups created priorto September 1999, we used the month when they firstappeared in the Google Groups archive as the startingmonth.

Control Variables. We controlled for the average levelof participation and message complexity in a newsgroupin a given month. Average level of participation was cal-culated as the total number of messages divided by thetotal number of members posting to the newsgroup in amonth. Message complexity was calculated as the aver-age number of lines in a message in a newsgroup in agiven month. We also included three dummies to controlfor the topic of a group: technology, health support, andissue discussion, with hobby groups as the base case.Finally, we controlled for the overall member scarcityin Usenet with the variable Usenet membership size, thetotal number of Usenet posters in a given month.


Table 2 Descriptive Statistics of Main Variables 4N = 1419295

Variables (untransformed) Mean Median Std. dev. Min Max

Growth rate (percentage) 16020 −0040 3055 −100 311500000Member overlap density 7056 5058 6076 0 64010Group size (members) 209021 120000 287020 0 41546000Group age (years) 7001 7000 3028 0 18000Level of participation 4034 2085 4040 0 70098Message complexity (lines) 47049 33000 91007 0 41979000Usenet membership size 1,433,137 1,322,355 3951937050 7021089 3,273,088

Analysis and ResultsTable 2 displays the descriptive statistics. Most of thevariables were highly skewed except for group age andgroup type dummies. Therefore, we listed both the meanand median of the variables. During the studied period,Usenet had 702,089 to over 3 million users a month,with a median of 1.3 million users a month. Medianlevel of participation in the sampled newsgroups was2.85 messages per member per month, and the medianmessage length was 33 lines. Group size ranged from 0(when a newsgroup had no activity for a month) to over4,000 members with a median of 120. Group age rangedfrom 0 to 18 years, with an average age of 7 years. It wascommon for newsgroups to share members. The medianof member overlap density was 5.58, equivalent to shar-ing all members with 5.58 other Usenet groups. Figure 3shows the average of member overlap density and groupsize across the 240 newsgroups in our sample. Duringthe studied period, member overlap density increasedwhile group size decreased. Newsgroups in our sam-ple had a median membership decline of about 0.4%each month, equivalent to an overall decline of approxi-mately 21% over the five-year period. Changes in mem-bership occurred gradually, with few extreme changesfrom month to month. We found 87 observations witha monthly growth rate larger than 5. Sensitivity analy-sis suggested that excluding these observations does not

Figure 3 Average Member Overlap Density (Left) and Average Group Size (Right) for the 240 Newsgroups from October 1999 toJanuary 2005

Month

Jul 1999 Jan 2001 Jul 2002 Jan 2004 Jul 2005

Month

Jul 1999 Jan 2001 Jul 2002 Jan 2004 Jul 2005

67

89

10

Ove

rlap

dens

ity

Membership overlap density

150

200

250

300

Gro

up s

ize

Group size

change our results, and we reported the analysis with allobservations included.

We applied a natural-log transformation to all vari-ables except group age and group type to improve theirnormality. Before transforming, a small value, 0.1, wasadded to the values of the variables. All independentvariables except for group age and group type werelagged for one month to examine their effects on thegrowth rate in the subsequent month. The independentvariables were mean centered before creating the inter-action terms. Table 3 shows the correlations among thetransformed variables. Multicollinearity analysis showedthat the variance inflation factors (VIFs) for the indepen-dent variables were below 3, well within the generallyacceptable ranges (Hair et al. 1998). Therefore, multi-collinearity was not a concern for this analysis.

To test the proposed hypotheses, we estimated alongitudinal and multilevel model using the xtmixedprocedure in Stata (Rabe-Hesketh and Skrondal 2005),predicting growth rate as a function of member over-lap density and group age and size and controlling foroverall Usenet membership size, group type, and thegroup’s levels of participation and message complexity.The mixed effect model takes into account the multi-level structure of our panel data, with multiple observa-tions over time nested within each newsgroup. It alsoallows both fixed and random effects and provides moreflexibility in specifying a model. Newsgroup IDs were


Table 3 Correlations

Variables 1 2 3 4 5 6 7

1 Growth rate 12 Member overlap density −00073∗ 13 Group size −00197∗ 00044∗ 14 Group age −00028∗ 00174∗ 00235∗ 15 Level of participation −00076∗ 00088∗ 00562∗ 00141∗ 16 Message complexity −00099∗ 00322∗ 00096∗ 00191∗ 00116∗ 17 Usenet membership size 00016 00028∗ 00051∗ −0005∗ 00012 −00067∗ 1

Note. N = 141929.∗Correlation significant at p < 0005.

included in the model as random effects to controlfor unobserved newsgroup heterogeneity. Examining thedistribution of the residuals suggested that there mightbe outliers in the sample. We ran sensitivity analysiswithout the outliers, and the results remained unchanged.We therefore report the results with all observationsincluded. Because our data set was clustered withinnewsgroups, the homogeneity of variance assumptionwas likely violated. We specified the mixed effect modelto estimate a unique variance component for each news-group. Therefore, heteroskedacity is not likely a concern.We also ran the most conservative fixed effect modelswith standard errors robust to newsgroup clustering. Theresults remained largely the same: that is, the main effectof group age became insignificant, but all other resultsremained the same. Our results from the mixed effectmodel are thus robust to heteroskedacity, and we reportthem in Table 4.

Table 4 Explaining Membership Growth

Model 1: Control Model 2: Main effect Model 3: Interaction

Intercept −00284 (0.223) −10368∗∗∗ (0.214) −10502∗∗∗ (0.215)Support group 00036∗∗∗ (0.012) −00108∗ (0.051) −00118∗ (0.06)Technology group −00013 (0.010) 00028 (0.04) 00041 (0.046)Issue group 00038∗∗∗ (0.012) 00012 (0.048) 00043 (0.057)Level of participation −00047∗∗∗ (0.005) 00096∗∗∗ (0.01) 00097∗∗∗ (0.01)Message complexity −00067∗∗∗ (0.006) 00013 (0.007) −00005 (0.007)Usenet membership size 00020 (0.016) 00097∗∗∗ (0.015) 00106∗∗∗ (0.015)Member overlap density −00046∗∗∗ (0.007) −00085∗∗∗ (0.012)Member overlap density2 — — 00022∗∗∗ (0.004)Group size −00246∗∗∗ (0.006) −00283∗∗∗ (0.007)Group age −00015∗∗∗ (0.002) −00015∗∗∗ (0.002)Group age2 — — −00000 (0.000)Member overlap density ∗Size −00035∗∗∗ (0.003)Member overlap density ∗Age 00003 (0.002)Member overlap density ∗Support group −00070∗∗∗ (0.019)Member overlap density ∗Technology group 00042∗∗ (0.016)Member overlap density ∗ Issue group −00047∗ (0.018)Number of groups/Clusters 240 240 240% of variance explained (estimated) (%) 1.6 12.6 13.9Deviance (−2 ∗ Log likelihood) 18,524.1 17,481.8 17,384.3ãDeviance 1104203∗∗∗ 9705∗∗∗

Notes. N = 141929. Standard errors are in parentheses.∗∗∗p < 00001; ∗∗p < 0001; ∗p < 0005.

We ran three hierarchical models to test the hypothe-ses. The first model included all control variables; thesecond model included the controls plus group size,age, and member overlap density; and the third modelincluded the squared term of overlap density and groupage and the interactions between overlap density andgroup size and age. Mixed effect models do not produceR-squared, and thus we calculated estimated R-squaredmanually following Xu (2003). Model 1 had poor modelfit and explained only about 1.6% of the variance inmembership growth. Model 2 explained 12.6% of thevariance, and Model 3 explained 13.9% of the variance.A chi-square test examining the difference between thedeviance statistics (−2 log-likelihood) across the threemodels suggested that adding group size, age, memberoverlap density, and their squared terms and interac-tions significantly improved model fit at the 0.001 level(ãDeviance = 1104203 between Models 1 and 2 and 97.5between Models 2 and 3, respectively).


Effects of Member Overlap DensityHypothesis 1 posits that member overlap density is neg-atively associated with growth, and it was supported.As shown in Model 2 in Table 4, the coefficient ofmember overlap density was negative and statisticallysignificant (� = −000461 p < 00001). The coefficientsuggests that a one-unit increase in member overlap den-sity, or about a 13% increase from the mean (7056 ×

13% = 1), was associated with a 0.6% (00046 × 13%)decline in membership in the subsequent month. Thisdecline is equivalent to 1.5 times the median monthlymembership decline rate (0.4%) in the sampled news-groups. The one-unit increase in member overlap densitycould occur, for example, when a focal group shared allmembers with one additional group or shared 10% ofits members with 10 other groups. This result showedthat an online group’s ability to grow can be hindered bysharing more members with more groups. Hypothesis 2posits that member overlap density reduces growth at anincreasing rate. It was not supported. In Model 3, thesquare term of member overlap density was positive andstatistically significant (� = 000221 p < 00001), whichmeans that member overlap density reduced growth rateat a decreasing rate. As shown in Figure 4, the overalleffect of member overlap density remained negative inour sample, yet the negative effects became smaller asmember overlap density increased.

Effects of Group Size, Age, andTheir Interactions with Overlap DensityHypothesis 3 posits that larger groups have slowergrowth rates, and Hypothesis 4 posits that memberoverlap density has a stronger negative effect on thegrowth rate of larger groups than of smaller ones. Ourresults supported both hypotheses. Model 2 showed thatgroup size was negatively associated with growth rate(�= −00246, p < 00001). A 10% increase in group sizewas associated with a 2.46% decrease in growth rates.Model 3 showed a negative interaction between memberoverlap density and group size (�= −000351 p < 00001).This suggests that group size moderated the effect ofmember overlap density on growth rate: the decrease ofgrowth rate from member overlap was greater in largegroups than in small groups, suggesting that, compared

Figure 4 Predicted Relationship Between Member OverlapDensity and Membership Growth

–0.5

0

0.5

1.0

–6 –4 –2 0

Overlap density (logged)

2 4

Predicted growth rate (logged)

–0.5

0

0.5

1.0

–6 –4 –2 0

Overlap density (logged)

2 4

Predicted growth rate (logged)

with smaller groups, larger groups were more vulnerableand likely to lose members to competition.

Hypothesis 5 posits that older groups have slowergrowth rates and that group age reduces growth at anincreasing rate. Hypothesis 6 posits that member over-lap density has a stronger negative effect on growthrate for older groups than for younger ones. The resultsprovided partial support for Hypothesis 5 but no sup-port for Hypothesis 6. Model 2 showed that group agewas negatively associated with growth rate (�= −00015,p < 00001). A one-year increase in a group’s age wasassociated with a 1.5% decrease in growth rate. Model 3showed that neither the square term of group age nor theinteraction between member overlap density and groupage was significant. This result suggests that the rela-tionship between group age and membership growth waslinear and that competition from sharing members withother groups had similar effects on older and youngergroups.

Examination of the coefficients of the control vari-ables yielded some interesting results. As shown inModel 3 of Table 4, there were significant interactionsbetween group types and member overlap density. Usinghobby groups as the base, sharing members with othergroups caused greater decrease in the growth ratesof health support groups (� = −000701 p < 00001)and issue discussion groups (� = −000471 p < 00001),yet a smaller decrease in the growth rate of tech-nology groups (� = 000421 p < 0001). Compared withhobby groups, health support groups and issue discus-sion groups seemed to be more vulnerable to competitivepressure from sharing members with other groups, andtechnology groups were the least vulnerable among thefour categories.

DiscussionWe set out to study the effects of intergroup competi-tion on the growth of online groups. Analysis of lon-gitudinal data from 240 Usenet groups suggested thatonline groups, a new organizational form, were sub-ject to the same competitive pressure in their environ-ment as traditional organizations. Sharing members withother groups creates the need to compete for members’time and efforts, which decreases a group’s ability togrow. Although the month-to-month change in mem-bership appears to be small, when considered over theentire studied period of a group’s lifetime, the changescan be substantial. Usenet groups in our sample experi-enced a median decline of group membership by 0.4%every month. Over a five-year period, that decrease isequivalent to about 21% of membership decline. Hold-ing all else constant, every additional group with whichthe focal group shared all members further reducedits membership by 0.6%, a 150% increase from themedian membership decline level. Over the course of a


five-year period, an additional monthly decline of 0.6%amounts to an additional membership decline of about31%, which is a substantial effect of which online groupmanagers should take notice. In addition, we found thatlarger and older groups experienced greater difficultyin growing their membership. We also found the neg-ative effects of competition were exacerbated by groupsize; larger groups were more likely to lose membersthan smaller groups in the face of intense competition.Overall, our study shows that competitive pressures in amature online environment can have a significant impacton online groups’ ability to grow.

Our study provides several new insights and raisesimportant questions about online groups. Compared toprevious studies that focused on the internal designand dynamics of online groups, our work advocates anecological view and calls attention to the external envi-ronments of online groups and their effects on groupperformance. The ecological framework and the mea-sure of member overlap density can be readily appliedto online groups hosted on different technological plat-forms. As such, they provide a basis for assessing thelevel of competition felt by particular groups and thelikely consequences of the competition on the group’sability to survive and succeed over the long term.

We also contribute to the organization science liter-ature by testing and extending organizational ecologytheories to a new context. The dominant narrative ofonline environments is one of abundance and unboundedaccess. More information is available to more peopleworldwide than ever before, and more people can par-ticipate in more online groups. In this context, onlinegroups have emerged as a form of virtual organizationwith more fluid membership and permeable boundariesthan traditional organizations. In spite of the emancipa-tory expectations of online environments, our findingsdemonstrate that ecological theories continue to apply inthis new context. Online groups are subject to compet-itive pressure for critical resources just like traditionalorganizations, with a few caveats.

We found that in a mature population of onlinegroups, member overlap density reduces growth at adecreasing rate. This finding differs from typical organi-zational ecology studies (Barron 1999) that predict thatoverlap density reduces growth at an increasing rate.Although it is clear that Usenet, with 20 years of historyprior to the study, can be thought of as a mature popula-tion, our results raise questions about how long it takesa population of online groups to mature. Compared totraditional organizations, online groups primarily dependon members’ time and effort to succeed. Competitivepressure may have greater and more immediate effectson online groups because of the low barriers involvedin joining and exiting groups. As a result, a popula-tion of online groups may mature quickly. The age ofthe Usenet groups that we studied means that many

members have already been facing conflicting demandsfrom multiple groups, and the intensity of competitionfaced by Usenet groups is already high. As the pop-ulation continues to grow, the intensity of competitionmay peak, and further membership overlap would have adecreasing impact on reducing membership growth. Asa result, competition from a handful of groups signif-icantly reduces membership growth, whereas competi-tion from additional groups has a smaller impact. Howthese effects work in new populations and how quicklythe negative impact of competition becomes dominantin other online settings remain interesting questions forfuture research.

Our findings also suggest that large and old onlinegroups have slower growth than small and young onesin a mature population. Although organizational ecol-ogy literature has seen mixed effects of organizationalsize and age on growth, we found that being large andbeing old were liabilities instead of advantages for theUsenet groups in our sample. Large online groups notonly experienced slower growth than small groups over-all but were also more vulnerable to competitive pres-sures from other groups. Online groups’ dependence onthe time and effort of members, coupled with their fluidmembership and low switching costs, may have changedthe implications of being big in this new context. Forinstance, members of large groups may find it difficult toforge strong connections with others yet easy to be over-whelmed with high traffic volume. Although we focusedon growth as our dependent variable, the results wereconsistent with prior work that found a negative relation-ship between the size of an online group and its turnoverrate (Butler 2001) as well as its responsiveness (Joneset al. 2004).

Meanwhile, we found that old online groups grew ata slower rate than young groups in a mature popula-tion. Members of old online groups may find the discus-sions getting repetitive and stale over time. As a result,they were more likely to leave, which made it hard forthese groups to maintain and grow their memberships.However, we did not find a curvilinear effect of groupage on growth rate. Group age also did not moderatethe impact of member overlap density, suggesting thatage did not make a group more or less vulnerable whenthere were many competitors in the environment. Thesefindings may reflect the maturity of Usenet during thestudy time period, when most newsgroups in our sam-ple were already well established and competition formember resources was high. Aging may play a biggerrole in an online group’s earlier stages of development.Whether and how the effect of age differs during thelifetime of online groups warrants future research.

Although we did not hypothesize the effects of grouptypes, we found that, compared to hobby groups, healthsupport groups and issue discussion groups were moresensitive to competitive pressure, whereas technology


groups were less sensitive. Two differences between thegroup types might have caused these effects. First, tech-nology groups are more information oriented, and mem-bers join to exchange and access information related totechnologies. Health support groups, in comparison, aremore socially oriented, and members of these groupsjoin to make friends and seek social support from oth-ers (Ridings and Gefen 2004). Second, discussions intechnology groups are shorter and often involve simplerexchanges such as questions and answers around a spe-cific topic. In comparison, discussions in health supportand issue groups tend to be longer and more interac-tive, with members engaging in emotional conversationsand sometimes heated debates. Therefore, regular par-ticipation in health support and issue discussion groupsrequires greater investment in time and effort thanother types of groups, making them more vulnerableto competition from external groups. This explanationis consistent with findings from research on volun-tary associations, which show that organizations withhigher demand for members’ time and participation haveshorter membership durations (Cress et al. 1997).

Practical ImplicationsAs organizations increasingly leverage online groups forboth internal operation and external relations, the onlinespace will grow more crowded, and Internet users willhave more options to choose from on where and how tospend their time (Yoo 2010). Understanding the impli-cation of such crowdedness is critical to the design andmaintenance of viable online groups. Given the num-ber of groups that already exist on the Internet, it willbecome more difficult to build vibrant groups or commu-nities from scratch. Businesses and individuals who areinterested in launching online groups cannot only spendtime and money crafting the internal design of the group.They must also consider the environment in which theywill launch the groups. Our study suggests that carefullypositioning a group in a niche market with less competi-tion from other groups may increase the group’s abilityto grow and succeed. When competition is unavoidable,a clear statement of the group’s purpose might serve touniquely differentiate a focal group from other groups.Existing online groups, especially large groups and typesof groups that are more vulnerable to competition, canmonitor the extent of membership overlap with othergroups in their environments as an index of competition.

The open nature of the Internet and the many plat-forms available to host online groups make prevent-ing members from switching groups challenging. Onlinegroup designers, however, have the capability to managethe level of intergroup competition within a platform.Often, organizations create a system of online groups,rather than a single group, to achieve their purpose. Forexample, the Mozilla foundation’s MozillaZine consistsof 28 discussion boards dedicated to different Mozilla

products and topics. Likewise, Inspired, a company thatpromotes online health discussion, currently hosts 193online health support groups. Although open member-ship among groups within a system allows members tofreely join and leave groups, such freedom may lead toa high level of membership overlap that is detrimentalto the operation of individual groups. Within a systemof online groups, designers may consider implement-ing policies or technical features that limit the numberof groups a member can participate in simultaneously.They may also consider tools that would recommend asmall set of groups that closely match a member’s inter-ests yet minimize intergroup competition. Meanwhile,online group designers may consider enacting policiesthat increase the costs of leaving or switching to othergroups.

Limitations and Future ResearchThis study is subject to several limitations. First, westudied Usenet groups hosted on one of many technolog-ical platforms and examined the effects of membershipoverlap within the Usenet population. Future studiesshould attempt to replicate our study in other online con-texts, such as Weblogs, electronic mailing lists, socialnetworking sites, or open source and open content col-laboration projects. We believe that our findings, suchas the negative impact of membership overlap and theliability of being big and old, can be generalized tomature populations of online groups hosted on otherplatforms, although the magnitudes of the influence mayvary. Our findings on the differential effects of member-ship overlap on groups of different types provide somehints about how intergroup competition may operate inother online contexts. For example, open source softwaregroups that mainly involve technical exchanges may beless susceptible to the competitive pressure of sharingmembers with other groups than cancer support or socialnetworking groups. These are interesting areas for futureresearch.

Second, we studied Usenet groups in multiple topicdomains without focusing on a particular topic. As aresult, we treated two online groups as occupying thesame niche if they shared a member, even if theirpredefined topics or purposes were different. Whereasorganizational ecology research often focuses on orga-nizations within a single industry, our approach is con-sistent with the principle of ecology theory in that itrecognizes that resource niches may (or may not) coin-cide with the socially constructed identities of the orga-nizations involved. In this regard, our work presents asmall step towards understanding cross-population com-petition (Ruef 2000). Building on this, future researchshould examine the effects of membership overlap bothwithin and across platforms of online groups.

This study focused on a mature population of onlinegroups that was itself facing competition for resources.


Our studied period was at a time when many new plat-forms were becoming popular and began competingwith Usenet groups for members. Because of this, weincluded the Usenet membership size variable to con-trol for the effect of Usenet resource scarcity resultingfrom competition from other platforms. Future researchshould gather data on the complete life history of awhole population of online groups to fully understandthe effects of overlap density, size, and age withinpopulations.

Fourth, we examined growth rate as the online groupperformance outcome. We explored several alternativemeasures, such as member attraction, member retention,and the growth of core members, and found the resultsto be similar to the results for growth. Future researchshould examine a broader set of measures of group per-formance, such as members’ satisfaction with a group,quality of contribution in a group, and the degree towhich a group achieves its own purposes (e.g., to col-laboratively develop software code or Wikipedia arti-cles). Finally, whereas this study focuses on group-leveldynamics, future work can investigate how group ecol-ogy may affect individual outcomes. Are all individualsaffected by membership overlap equally, or do individ-ual motivations and roles interact with group ecologyto affect individual behaviors? Answers to these ques-tions will help us better understand the ecology of onlinegroups and learn how to reduce the potential negativeeffects of member competition.

ConclusionOnline groups play an increasingly important role inorganizations and our society by improving individualand social life, generating innovative solutions for busi-nesses, creating artifacts of lasting value for society, andenabling new forms of collective action. However, ittakes more than a technical infrastructure to make anonline group successful. How a group operates matters,group demographics matter, and in this paper we showthat a group’s external environment matters as well. Ourwork shows the theoretical and practical importance ofstudying the ecological environment and its impact ononline groups. Only by recognizing the impact of poten-tially competing forces and by understanding how tocope with competition can businesses and individualsachieve their goals of building viable online groups.

AcknowledgmentsThe authors thank Samer Faraj and three anonymous review-ers for their invaluable comments, Robert Kraut for feed-back on early versions of the manuscript, and Marc A. Smithand Microsoft Research for providing data. This researchwas supported by the National Science Foundation [GrantsIIS-0325049 and OCI-0951630].

ReferencesAldrich H, Ruef M (2006) Organizations Evolving (Sage, London).

Audia PG, Greve HR (2006) Less likely to fail: Low performance,firm size, and factory expansion in the shipbuilding industry.Management Sci. 52(1):83–94.

Bagozzi RP, Dholakia UM (2006) Open source software user com-munities: A study of participation in Linux user groups. Man-agement Sci. 52(7):1099–1115.

Banaszak-Holl J (1991) Incorporating organizational growthinto models of organizational dynamics: Manhattan banks,1791–1980. Unpublished dissertation, Cornell University,Ithaca, NY.

Barnett WP, McKendrick DG (2004) Why are some organizationsmore competitive than others? Evidence from a changing globalmarket. Admin. Sci. Quart. 49(4):535–571.

Barron DN (1999) The structuring of organizational populations.Amer. Sociol. Rev. 64(3):421–445.

Barron DN, West E, Hannan MT (1994) A time to grow and a timeto die: Growth and mortality of credit unions in New York City,1914–1990. Amer. J. Sociol. 100(2):381–421.

Baum JAC, Singh JV (1994) Organizational niches and the dynamicsof organizational mortality. Amer. J. Sociol. 100(2):346–380.

Becker GS (1965) A theory of the allocation of time. Econom. J.75(299):493–517.

Bercovitz J, Mitchell W (2007) When is more better? The impactof business scale and scope on long-term business survival,while controlling for profitability. Strategic Management J.28(1):61–79.

Bottazzi G, Cefis E, Dosi G (2002) Corporate growth and industrialstructures: Some evidence from the Italian manufacturing indus-try. Indust. Corporate Change 11(4):705–723.

Bruderl J, Schussler R (1990) Organizational mortality—The lia-bilities of newness and adolescence. Admin. Sci. Quart.35(3):530–547.

Butler BS (2001) Membership size, communication activity, and sus-tainability: A resource-based model of online social structures.Inform. Systems Res. 12(4):346–362.

Butler BS, Wang X (2012) The cross-purposes of cross-posting:Boundary reshaping behavior in online discussion communities.Inform. Systems Res. 23(3, Part 2):993–1010.

Butler B, Sproull L, Kiesler S, Kraut R (2007) Community effortin online groups: Who does the work and why? Weisband S,ed. Leadership at a Distance (Lawrence Erlbuam Associates,Mahwah, NJ), 171–194.

Carroll GR, Hannan MT (1989) Density delay in the evolution oforganizational populations: A model and five empirical tests.Admin. Sci. Quart. 34(3):411–430.

Carroll GR, Hannan MT (2000) The Demography of Corporationsand Industries (Princeton University Press, Princeton, NJ).

Constant D, Sproull L, Kiesler S (1996) The kindness of strangers:The usefulness of electronic weak ties for technical advice.Organ. Sci. 7(2):119–135.

Cress DM, McPherson JM, Rotolo T (1997) Competition and com-mitment in voluntary memberships: The paradox of persistenceand participation. Sociol. Perspect. 40(1):61–79.

Dahlander L, Frederiksen L (2012) The core and cosmopolitans:A relational view of innovation in user communities. Organ. Sci.23(4):988–1007.


Dowell G, Swaminathan A (2000) Racing and back-pedaling into thefuture: New product introduction and organizational mortalityin the U.S. bicycle industry, 1880–1918. Organ. Stud. 21(2):405–431.

Faraj S, Johnson SL (2011) Network exchange patterns in online com-munities. Organ. Sci. 22(6):1464–1480.

Faraj S, Jarvenpaa SL, Majchrzak A (2011) Knowledge collaborationin online communities. Organ. Sci. 22(5):1224–1239.

Fichman M, Levinthal DA (1991) Honeymoons and the liability ofadolescence—A new perspective on duration dependence insocial and organizational relationships. Acad. Management Rev.16(2):442–468.

Freeman J, Carroll GR, Hannan MT (1983) The liability of newness:Age dependence in organizational death rates. Amer. Sociol. Rev.48(5):692–710.

Greve HR (2008) A behavioral theory of firm growth: Sequentialattention to size and performance goals. Acad. Management J.51(3):476–494.

Grewal R, Lilien GL, Mallapragada G (2006) Location, location, loca-tion: How network embeddedness affects project success in opensource systems. Management Sci. 52(7):1043–1056.

Gu B, Konana P, Rajagopalan B, Chen HWM (2007) Compe-tition among virtual communities and user valuation: Thecase of investing-related communities. Inform. Systems Res.18(1):68–85.

Hair JF Jr, Anderson RE, Tatham RL, Black WC (1998) MultivariateData Analysis, 5th ed. (Prentice-Hall, Upper Saddle River, NJ).

Hannan MT (1998) Rethinking age dependence in organizational mor-tality: Logical formalizations. Amer. J. Sociol. 104(1):126–164.

Hannan MT, Freeman J (1977) The population ecology of organiza-tions. Amer. J. Sociol. 82(5):929–964.

Herring SC (2004) Slouching toward the ordinary: Currenttrends in computer-mediated communication. New Media Soc.6(1):26–36.

Ingram P, Inman C (1996) Institutions, intergroup competition, andthe evolution of hotel populations around Niagara Falls. Admin.Sci. Quart. 41(4):629–658.

Jeppesen LB, Frederiksen L (2006) Why do users contribute tofirm-hosted user communities? The case of computer-controlledmusic instruments. Organ. Sci. 17(1):45–63.

Jeppesen LB, Lakhani KR (2010) Marginality and problem-solvingeffectiveness in broadcast search. Organ. Sci. 21(5):1016–1033.

Jones Q, Rafaeli S, Ravid G (2004) Information overload and the mes-sage dynamics of online interaction spaces: A theoretical modeland empirical exploration. Inform. Systems Res. 15(2):194–210.

Kane GC (2011) A multimethod study of information quality inwiki collaboration. ACM Trans. Management Inform. Systems2(1):Article 4.

Kim AJ (2000) Community Building on the Web: Secret Strategies forSuccessful Online Communities (Peachpit Press, Berkeley, CA).

Kitts JA (2009) Paradise lost: Age-dependent mortality of Americancommunes, 1609–1965. Soc. Forces 87(3):1193–1222.

Kollock P, Smith MA (1999) Communities in Cyberspace (Routledge,New York).

Kraut R (2003) Applying social psychological theory to the prob-lems of group work. Carroll JM, ed. HCI Models, Theories,and Frameworks: Toward a Multidisciplinary Science (MorganKaufmann, New York), 325–356.

Kuilman J, Li J (2006) The organizers’ ecology: An empirical studyof foreign banks in Shanghai. Organ. Sci. 17(3):385–401.

Lee GK, Cole RE (2003) From a firm-based to a community-basedmodel of knowledge creation: The case of the Linux kerneldevelopment. Organ. Sci. 14(6):633–649.

MacMillan D (2009) Facebook lures advertisers at Myspace’sexpense. BusinessWeek (July 9) http://www.businessweek.com/technology/content/jul2009/tc2009079_233343.htm.

McPherson M (1983) An ecology of affiliation. Amer. Sociol. Rev.48(4):519–532.

McPherson JM, Rotolo T (1996) Testing a dynamic model of socialcomposition: Diversity and change in voluntary groups. Amer.Sociol. Rev. 61(2):179–202.

McPherson JM, Popielarz PA, Drobnic S (1992) Social networks andorganizational dynamics. Amer. Sociol. Rev. 57(2):153–170.

McPherson M, Smith-Lovin L, Cook JM (2001) Birds of a feather:Homophily in social networks. Annual Rev. Sociol. 27(1):415–444.

Moon JY, Sproull LS (2008) The role of feedback in managingthe Internet-based volunteer workforce. Inform. Systems Res.19(4):494–515.

Nam I, Gruca TS, Tracy R (2010) The effects of competition onreferral alliances of professional service firms. Organ. Sci.21(1):216–231.

Nambisan S, Baron RA (2010) Different roles, different strokes: Orga-nizing virtual customer environments to promote two types ofcustomer contributions. Organ. Sci. 21(2):554–572.

Nickel MN, Fuentes JM (2004) Relationship between legitimation,competition, and organizational death: Current state of the art.Internat. J. Management Rev. 5(1):43–62.

Nielsen (2009) Global faces and networked places—A Nielsen reporton social networking’s new global footprint. Report, The NielsenCompany, New York.

Pew Research Center (2008) Pew Internet and American Life ProjectSurvey, March 2000-April 2006. Report, Pew Research Center,Washington, DC.

Podolny JM, Stuart TE, Hannan MT (1996) Networks, knowledge,and niches: Competition in the worldwide semiconductor indus-try, 1984–1991. Amer. J. Sociol. 102(3):659–689.

Popielarz PA, Neal ZP (2007) The niche as a theoretical tool. AnnualRev. Sociol. 33(1):659–689.

Porter CE, Donthu N (2008) Cultivating trust and harvesting value invirtual communities. Management Sci. 54(1):113–128.

Raban DR, Moldovan M, Jones Q (2010) An empirical study ofcritical mass and online community survival. Proc. 2010 ACMConf. Comput. Supported Cooperative Work (ACM, New York),71–80.

Rabe-Hesketh S, Skrondal A (2005) Multilevel and LongitudinalModeling Using Stata (Stata Press, College Station, TX).

Ranger-Moore J (1997) Bigger may be better, but is older wiser?Organizational age and size in the New York life insuranceindustry. Amer. Sociol. Rev. 62(6):903–920.

Rheingold H (2000) The Virtual Community: Homesteading on theElectronic Frontier (MIT Press, Cambridge, MA).

Ridings C, Gefen D (2004) Virtual community attraction: Why peoplehang out online. J. Comput.-Mediated Comm. 10(1):Article 4.

Ruef M (2000) The emergence of organizational forms: A communityecology approach. Amer. J. Sociol. 106(3):658–714.


Ruef M (2004) For whom the bell tolls: Ecological perspectiveson industrial decline and resurgence. Indust. Corporate Change13(1):61–89.

Shaw BR, McTavish F, Hawkins R, Gustafson DH, Pingree S (2000)Experiences of women with breast cancer: Exchanging socialsupport over the chess computer network. J. Health Comm.5(2):135–159.

Shirky C (2003) A group is its own worst enemy. Clay Shirky’swritings about the Internet. Accessed June 1, 2010, http://shirky.com/writings/group_enemy.html.

Smith MA (2005) Netscan: A social accounting search engine (http://netscan.research.microsoft.com), Accessed October 15, 2007,http://netscan.research.microsoft.com.

Sorensen JB, Stuart TE (2000) Aging, obsolescence, and organiza-tional innovation. Admin. Sci. Quart. 45(1):81–112.

Sproull L, Arriaga M (2007) Online communities. Bidgoli H, ed.Handbook of Computer Networks (John Wiley & Sons, NewYork), 248–279.

Sproull L, Kiesler S (1991) Connections: New Ways of Working inthe Networked Organization (MIT Press, Cambridge, MA).

Stewart KJ, Gosain S (2006) The impact of ideology on effective-ness in open source software development teams. MIS Quart.30(2):291–314.

Stinchcombe AL (1965) Social structure and organizations. March JG,ed. Handbook of Organizations (Rand McNally & Company,Chicago), 142–193.

Thornhill S, Amit R (2003) Learning about failure: Bankruptcy, firmage, and the resource-based view. Organ. Sci. 14(5):497–509.

Wasko MM, Faraj S (2005) Why should I share? Examining socialcapital and knowledge contribution in electronic networks ofpractice. MIS Quart. 29(1):35–57.

Xu R (2003) Measuring explained variation in linear mixed effectsmodels. Statist. Medicine 22(22):3527–3541.

Yoo CS (2010) Is the Internet a maturing market? Comm. ACM53(8):24–26.

Xiaoqing Wang is an assistant professor in the Depart-ment of Decisions, Operations, and Information Technologiesat the Robert H. Smith School of Business at the Universityof Maryland. She received her Ph.D. from the University ofPittsburgh. Her research interests include online communities,computer-supported collaborative work, and technology adop-tion and diffusion. Her work has appeared in the top informa-tion systems journals.

Brian S. Butler is an associate professor in the Collegeof Information Studies and associate professor of InformationSystems at the Robert H. Smith School of Business at theUniversity of Maryland. His current work with online commu-nities and social computing focuses on modeling the dynamicinterplay between technology and organizing processes. Hiswork has appeared in the leading information systems andmanagement journals.

Yuqing Ren is an assistant professor of information anddecision sciences at the Carlson School of Management at theUniversity of Minnesota. She holds a Ph.D. from CarnegieMellon University. Her areas of interest include online com-munity design, distributed collaboration, knowledge manage-ment, and computational modeling of social and organizationalsystems. Her work has been published in the top organizationand management science journals.

Documents

The Impact of Membership Overlap on Growth: An Ecological Competition View of Online Groups