THE EMERGENCE OF NEW ORGANIZATIONAL POPULATIONS: MEDIA LEGITIMATION EFFECTS IN THE BROADBAND INTERNET ACCESS INDUSTRY PATRICK L.

Size: px
Start display at page:

Download "THE EMERGENCE OF NEW ORGANIZATIONAL POPULATIONS: MEDIA LEGITIMATION EFFECTS IN THE BROADBAND INTERNET ACCESS INDUSTRY PATRICK L."

Transcription

1 THE EMERGENCE OF NEW ORGANIZATIONAL POPULATIONS: MEDIA LEGITIMATION EFFECTS IN THE BROADBAND INTERNET ACCESS INDUSTRY PATRICK L. SCHULTZ Unversty of North Dakota Management Department Grand Forks, ND (701) KIMBERLY BOAL Texas Tech Unversty Rawls College of Busness Area of Management Texas Tech Unversty 15th and Flnt Avenues Lubbock, TX (806)

2 THE EMERGENCE OF NEW ORGANIZATIONAL POPULATIONS: MEDIA LEGITIMATION EFFECTS IN THE BROADBAND INTERNET ACCESS INDUSTRY ABSTRACT Buldng on the tradton of research n organzatonal ecology and nsttutonal theory, ths artcle extends our understandng of how legtmacy contrbutes to the creaton of new organzatonal populatons. Ecologsts have observed that both legtmacy and competton operate through densty-dependent processes to affect patterns of organzatonal foundngs. Drawng on Aldrch and Fol (1994), we argue that new populatons form from the nfluence of cogntve and socopoltcal legtmacy arsng around communcaton and nformaton exchange among ndustry partcpants. Usng data on the emergent stage of development of the broadband Internet access ndustry, we develop and test a meda coverage model of populaton entry rates, assessng how the legtmacy effects of varous forms of meda coverage affect entry rates n the growng populaton. Whle the study found sgnfcant support for both densty-dependent and meda-based models of legtmacy effects when treated ndvdually, when taken together densty-dependent effects became nsgnfcant, demonstratng the strong nfluence of the ongong dalogue occurrng among populaton partcpants as a factor n populaton emergence, and strongly suggestng a complex set of nterrelatonshps between dfferent aspects of legtmacy. 1

3 And t ought to be remembered that there s nothng more dffcult to take n hand, more perlous to conduct, or more uncertan n ts success, then to take the lead n the ntroducton of a new order of thngs. Machavell's The Prnce The hghly dynamc, technologcally-drven compettve envronment confrontng organzatons today requres greater nterest n and a research focus on the subject of ndustry creaton and entrepreneurshp. Aldrch and Fol (1994) observed that a fundamental problem facng nnovators and entrepreneurs n nascent ndustres s the lack of legtmacy of new technologes, frm strateges, and organzatonal forms whch could be used to obtan needed resources and support. The rse of the personal computer ndustry n the late 1970s and early 1980s, and the subsequent emergence from the 1990s onward of related technologes nvolvng the Internet fueled waves of new forms of busness centered on telecommuncatons and nformaton systems. Along wth ths, fnancal markets and governmental regulators have struggled n evaluatng these new ndustres. Lackng the nsttutonal supports of more establshed busness forms, emergng ndustres present a varety of challenges to partcpants that do not appear n more mature areas of actvty new forms of organzatonal actvty do not ft neatly nto exstng categores of acceptable busness practce, patterns of seemngly approprate behavor have yet to appear, and regulatory structures have not developed. In order to fully understand the creaton of new populatons of organzatons, a clearer understandng of how legtmacy arses durng the earlest moments of ndustry formaton must be acheved. The purpose of ths study s to examne how dfferent dmensons of legtmacy nfluence the dynamcs of actvty n new organzatonal populatons and to propose ways n whch these 2

4 concepts can be measured and the relatonshps between them tested, n a way whch advances establshed research n both organzatonal ecology and nsttutonal theory. Legtmacy has been of sgnfcant nterest n two major areas of organzatonal research: organzatonal ecology and nsttutonal theory. Organzatonal ecologsts, usng the denstydependence model of legtmacy (Hannan, 1991; Hannan and Carroll, 1992; Hannan and Freeman, 1987; Hannan and Freeman, 1988), have nspred a large stream of research on the pattern of foundng and falure rates n organzatonal populatons over tme, and produced a number of valuable nsghts and fndngs. Insttutonal theory research dentfes legtmacy by the degree to whch organzatons conform n a number of ways to regulatory or legal pressures, a shared sense of value or expectaton, or socally constructed categores of meanng (Scott, 1995: 45-47). What s evdent from both research tradtons s that each shares an nterest n how the socal envronment of organzatons affects the character or composton of organzatonal groups (.e., ndustres, populatons, or felds). Despte ths shared nterest n the legtmacy concept, both nsttutonal theory and organzatonal ecology dffer n ther vews of the legtmacy concept and how that concept appears n emprcal research. The result has been a schsm between the two theoretcal tradtons on a concept that each feld shares (Carroll and Hannan, 1989a, 1989b; Zucker, 1989; Hannan, Carroll, Dundon, and Torres, 1995; Baum and Powell, 1995; Hannan and Carroll, 1995). Much of the concern centers on accurately portrayng the complexty of the legtmacy concept, and the dffculty n measurng t and ts nfluence on populaton processes. From the ongong debate and commentary n the lterature, a number of observatons can be made about how they approach legtmacy. 3

5 Frst, the true complexty of the legtmacy concept has not been fully explored, partcularly ts multdmensonal character. Ths has been partcularly problematc n the organzatonal ecology lterature, where the densty of an organzatonal form (the number of nstances of a partcular organzatonal form n exstence at a certan tme) s used as an ndcator of the legtmacy of that form. In most studes, legtmacy refers smply to "taken-forgrantedness." Baum and Powell (1995) noted that legtmacy s a multdmensonal concept, and many factors beyond densty can contrbute to legtmacy. Delacrox and Rao (1994) clamed that densty tself s subject to a number of ndependent factors, each wth ther own dynamcs and effects on the organzatonal populaton. Second, each feld of research has observed that the other has mssed a crtcal step n analyzng the mpact of legtmacy: each has not effectvely operatonalzed the legtmacy concept despte ts mportance. For nstance, whle populaton ecology and nsttutonal theory make substantal contrbutons to the understandng of organzatons, the exchange between Carroll and Hannan (1989a, 1989b) and Zucker (1989) resulted n a realzaton that research n both felds typcally do not drectly measure legtmacy and nstead rely on nferences drawn from other more readly observable phenomena. For ecologsts, more drect measurement of the legtmacy concept s avoded because of a belef n ts fundamentally unobservable character (Carroll and Hannan, 1989b) and the assumpton that t s controlled by easly observed measures of organzatonal densty (Hannan, 1989, 1991). For nsttutonal theorsts, legtmacy of organzatonal practces are explaned by the effects of macro-level actors or events on the adopton of actvtes at lower levels of analyss. Fnally, the concept of legtmacy lacks a strong connecton to acton and change occurrng among ndustry actors. Ths occurs because of the hghly abstract nature of the 4

6 elements nvolved and the emphass on explanng stablty, conformty, and order (Gddens, 1979 and 1984; Hybels, 1995; Sewell, 1992). It s clear that many organzatons are not smply passve actors; they play a role n the mantenance of ther own legtmacy wthn an ndustry (Ashforth and Gbbs, 1990; Elsbach, 1994; Olver, 1991 and 1997). What s less clear s how legtmacy develops not just for ndvdual organzatons but for new organzatonal forms at the creaton of an ndustry, when densty s low t may not fully capture that developmental process. Ths study contrbutes to ths debate n a number of ways. After a dscusson of the organzatonal ecology perspectve, dfferent dmensons of legtmacy dentfed n the lterature wll be dscussed. Buldng on the framework ntally presented n Aldrch and Fol (1994) and further developed n Aldrch (1999), the legtmacy concept wll be dscussed n terms of ts cogntve and socopoltcal dmensons. Motvated by suggestons from Baum and Powell (1995), and recent work by Hoffman and Ocaso (2004) and Pollock and Rndova (2003), new ways of operatonalzng those attrbutes of legtmacy wll also be dentfed, by tappng nto ndustry dscourse through news meda sources rather than organzatonal densty. By separatng legtmacy nto component dmensons, each can be ndependently assessed for ts effect on organzatonal populatons. Fnally, these new measures of legtmacy wll be ncluded n a model of new market entry rates n the context of the emergence of the broadband Internet access ndustry from the years 1993 to The broadband access ndustry s deal for ths study for a number of reasons. Frst, the ndustry arose at the ntersecton of a number of dfferent actors and nterests, ncludng both broadly dversfed telecommuncatons ndustry players and broadband ndustry specalst frms, networkng and computng ndustry companes, regulatory enttes, and consumers. The legtmacy of the ndustry would ultmately be shaped by the nteracton of ths 5

7 dverse set of nfluences. Second, the legtmacy of the ndustry would be nfluenced by a competton over the utlty of broadband n a varety of dfferent applcatons (e.g., hgh-speed access to Internet, enhanced vdeo servces, new telephone functons, or other yet-to-be developed applcatons). THEORY AND HYPOTHESES Densty-Dependent Model of Legtmacy and Competton Organzatonal ecology has made major contrbutons to the study of legtmacy and ts mpact on the hstorcal development of ndustres and organzatonal populatons. The denstydependence model of legtmaton and competton, ntally developed by Hannan, Freeman, and Carroll (Hannan and Freeman, 1987 and 1988; Carroll and Hannan, 1989a; Hannan and Carroll, 1992), descrbes the relatonshp between the prevalence of a partcular organzatonal form, referred to as the densty of that organzaton type, to legtmacy and competton effects occurrng wthn that group of organzatons. Intal ncreases n the number of frms n an emergent populaton contrbute to the legtmacy of a new organzatonal form by sgnalng to potental entrepreneurs that opportuntes exst and that resources are avalable for the formaton of new frms; as densty ncreases the rate of foundng ntally ncreases along wth t. At hgher levels of organzatonal densty, rvalry ncreases n ntensty as exstng frms compete over ncreasngly scarce resources. Hgh densty levels then sgnal to potental entrants that opportuntes are beng competed away by the large number of frms already n the envronment. As the densty of frms ncreases, competton effects domnate the actvty occurrng wthn the populaton leadng to a decrease n the foundng rate. 6

8 The combnaton of legtmacy and competton effects produced by the densty level of an organzatonal form create and nverted U-shaped relatonshp between densty and foundng rates. Support for ths relatonshp between foundng and entry rates and densty has been reported n many organzatonal populatons, ncludng labor unons (Hannan and Freeman, 1987), newspapers (Carroll and Hannan, 1989a; Delacrox and Carroll, 1983), breweres (Carroll and Swamnathan, 1991; Carroll and Wade, 1991; Carroll, Presendoerfer, Swamnathan, and Wedenmayer, 1993), New York lfe nsurance companes (Budros, 1994), Toronto day care centers (Baum and Olver, 1992; Baum and Sngh, 1994), natonal automoble ndustres (Rao, 1994; Hannan et al., 1995), savngs and loan assocatons (Haveman, 1993), and trade assocatons (Aldrch, Zmmer, Staber, and Beggs, 1994), just to name a few. 1 Carroll and Hannan (2000) broadened the applcaton of the model of densty-dependent legtmacy and competton by hghlghtng the varety of ways the densty-dependence model of legtmacy-competton could be appled dependng on how populaton densty s operatonalzed and what s consdered the relevant foundng events. For nstance, foundng events typcally consst of true de novo organzatonal start-ups, but may also nclude spn-off frms splt from larger parent enttes, and de alo entres created from dversfcaton strateges or complete mgratons of frms from one populaton to another (Carroll and Hannan, 2000: 41-43). An example of densty-dependent processes operatng n the context of organzatons enterng from other organzatonal populatons s seen n Haveman s (1993) study of dversfcaton and market entry n a populaton of savngs and loan assocatons. She proposed that the dynamcs of legtmaton and competton descrbed by populaton ecology also apply to change n exstng organzatons, such as frm entry nto new product markets, and found support for the nverted 1 Two recent and comprehensve revews descrbng the varety of studes that have been conducted testng the denstydependence model nclude Baum (1996) and Carroll and Hannan (2000). 7

9 U-shaped relatonshp predcted by the densty-dependent legtmacy and competton thess. The mplcaton s that densty contrbutes to the dynamcs of legtmacy and competton regardless whether t s the product of entrepreneural foundngs or entres from neghborng ndustres. Carroll and Hannan (2000: 51-54) also note that there exst a wde varety of possble unts of analyss of nterest, begnnng at the establshment level where entry events and densty correspond to operatonal enttes wthn frms to the organzatonal level n whch entry events and densty correspond to the number of actve organzatons present. Ths allows consderable flexblty and choce n the applcaton of the densty-dependent model by organzaton theorsts. Study of the approprate unt of analyss should provde an accurate depcton of the processes n whch legtmacy s generated and compettve rvalry operates occurrng n the populaton. Therefore, n general the entry rate could nclude both foundngs by completely new organzatons and entres by exstng ones n other ndustres, wth entry rate and densty cast at the establshment or organzatonal unt of analyss, the specfc choce of each dependng on the context of a partcular study: Hypothess 1: The entry rate has an nverted U-shaped relatonshp wth the populaton densty. Hypothess 1a: The legtmacy of a populaton postvely affects the rate of entry. Hypothess 1b: Competton effects wthn a populaton negatvely affect the rate of entry. Cogntve and Socopoltcal Legtmacy Effects of Meda Coverage Cogntve and socopoltcal components of legtmacy. Despte the wdespread support for and usefulness of the fndngs of ecology research, there reman some dffcultes n usng the 8

10 densty-dependence model n understandng the works of legtmacy processes n populatons. One problem nvolves the use of one varable populaton densty as an ndcator for legtmacy. It s common n organzatonal research to dstngush between legtmacy stemmng from taken-for-grantedness versus legtmacy arsng from normatve approval, relatonal support, and regulatory or poltcal evaluaton (Scott, 1995: 35-47). For nstance, the mpact of legtmacy has been descrbed as comng from a basc recognton or acceptance of an organzatonal form or behavor (Berger and Luckmann, 1966; Zucker, 1977; Meyer and Rowan, 1983), the perceved value of the form to organzatonal members and the communty (Selznck, 1949; Selznck, 1957; Zald and Denton, 1963), and the approval and sancton of other socal enttes (Meyer and Scott, 1983; DMaggo and Powell, 1983). Recognzng the multtudnous ways n whch legtmacy has been conceptualzed, Suchman offered an nclusve defnton of legtmacy as a generalzed percepton or assumpton that the actons of an entty are desrable, proper, or approprate wthn some socally constructed system of norms, values, belefs, and defntons (1995: 574). These mages of legtmacy llustrate both cogntve and socopoltcal dmensons of the legtmaton process, each of whch act together to nfluence the creaton of new organzatonal forms and change n exstng populatons (Aldrch and Fol, 1994; Aldrch, 1999). The qualty of taken-for-grantedness s the key descrpton of cogntve legtmacy, and Aldrch noted that at ts hghest level a new product, process, or servce s accepted as part of the sococultural and organzatonal landscape (1999: 230). Socopoltcal legtmacy reflects the attanment of acceptance by crtcal envronmental enttes or conformty to socally constructed standards of behavor or legally establshed codes of conduct. The cogntve dmenson of legtmacy acheves ts authorty by ts unquestoned character n the mnds of ndvduals, whle the socopoltcal 9

11 dmenson depends largely on the nfluence of shared values and norms n creatng and sustanng stable patterns of organzatonal actvty. Whle organzatonal ecology studes also dstngush between cogntve and socopoltcal legtmacy processes, some have argued that by focusng on densty-dependence t tends to emphasze cogntve legtmacy n the development of organzatonal populatons. Socopoltcal factors are largely vewed as drven by densty-dependent processes (Hannan and Carroll, 1995). However, nsttutonal theorsts typcally vew each form of legtmacy as complementary and fundamentally nter-related. Baum and Powell (1995) noted that the emphass on only one part of the legtmaton process results n an ncomplete pcture; aspects of legtmacy other than taken-for-grantedness need to be assessed and combned wth densty for a more complete understandng of the process. Smlarly, Delacrox and Rao (1994) observed that a varety of dfferent processes affectng ndustry development are bundled together wthn populaton densty, reflectng ts reputaton, the nformaton and experence of ts partcpants, and the establshment of supportng structures such as technologcal standards and ndustry assocatons. Rather than focusng on one or the other, the nfluence of a number of dfferent cogntve and socopoltcal legtmacy processes may be masked by relyng only on densty. The result s a great deal of dffculty n determnng the true role of each element on the emergence and development of organzatonal populatons. Whle densty may serve as a parsmonous ndcator of populaton processes, t also masks a number of socal and economc factors that nfluence the dynamcs wthn an organzatonal populaton. Of course, ths crtcsm s not at all lmted to ecologcal research; research n the nsttutonal tradton also suffers from ths weakness, and an ongong challenge wthn the feld of organzaton studes s to devse measures of cognton and culture that more 10

12 effectvely dstngush between dfferent legtmaton processes and ther effects on organzatonal densty (Baum and Powell, 1995: 530; Schneberg and Clemens, forthcomng). Meda coverage model of legtmacy effects. Aldrch and Fol (1994: 648) noted that legtmacy could be assessed "by measurng the level of publc knowledge about a new actvty" or by assessng publc acceptance of an ndustry. Publc knowledge and understandng of new organzatonal actvtes reflects the underlyng shared schemas that form the bass of legtmate socal acton. Baum and Powell called attenton to the analyss of nformaton exchange and communcaton content as an establshed method whch acheves the need for a hstorcal scope and a means for measurng legtmacy separately from densty (1995: ). They note that detaled meda coverage exsts on a wde range of ndustres, and cte a number of references where content analyss has been used to study the context of socal acton. In Abrahamson s (1996) study of the dssemnaton of management fashons, he observed that the mass meda served as a pont of dffuson for manageral deas between both fashon setters and fashon users, and thus played a role n promotng actual manageral practces. For many ndvduals and organzatons, the mass meda serves as an mportant source of nformaton about events and ssues gong on about them n the envronment, a process whch then goes beyond beng nformatve to affectng long-term changes n belefs and atttudes (Alper and Ledy, 1969). Terreberry (1968) clamed that legtmacy s ntmately ted to nformatonal exchange regardng rewardng uses for resources and favorable organzatonal actvtes. The mass meda s n effect an arena through whch varous actors express ther nterests (Hlgartner and Bosk, 1988, McCombs and Shaw, 1972). Actors use the mass meda as a tool to advance ther own agendas, drect attenton to ssues, and present and defne stuatons n ways that sut them best. For nstance, varous groups, both wthn and outsde an ndustry, can use the 11

13 mass meda to generate awareness of and promote ther own vson of the best organzatonal forms and practces (Abrahamson, 1991); the meda serves as means by whch groups can stake out ther own clams to legtmacy (Coombs, 1992). In Pollock s and Rndova s study of the effect of mass meda-based nformaton exchange on IPO performance, they explaned that n performng ts functons of nformng, hghlghtng, and framng, the meda presents market partcpants wth nformaton that affects mpresson formaton and the legtmaton of frms (2003: 632). However, ths process runs n both drectons: ndvduals and organzatons both learn about ssues and judge ther mportance as well as attempt to assert ther own nterpretatons they both attempt to nfluence and are themselves nfluenced by nteractng through the condut of the mass meda. Among content analyss researchers, the mass meda and the busness press are routnely seen as means n whch organzatonal communtes create, share, and perpetuate shared conceptons of facts and meanngs regardng socal acton (Gerbner, 1969: 123), socal processes clearly ted to the generaton of cogntve and socopoltcal legtmacy. Examnng the nformaton exchange occurrng wthn the mass meda can be used to smply descrbe trends n communcaton content (Holst, 1968: 610), but can also be used to examne aspects of culture and socal change (Berelson, 1954: 503; Franzos, 1997; Holst, 1968: 634; Tarrow, 1989), to trace the development of scholarshp and research nterests n scentfc felds (Berelson, 1954: 491; Duncan, 1991; Whaples, 1991), and to assess the focus of attenton of varous socal groups (Hoffman and Ocaso, 2001; Klennjenhu, Rdder, and Retberg, 1997; Weber, 1990: 9). The analyss of publc meda coverage of an ndustry should also be an effectve way of measurng the legtmacy of organzatonal forms and strateges snce t s central n both dssemnatng knowledge about new organzatonal practces and n nfluencng the behavors of ndvdual 12

14 actors lke potental nvestors or entrepreneurs, and should therefore be crtcal n the establshment of taken-for-grantedness. General meda coverage. As the ntensty of general meda coverage of a partcular organzatonal actvty ncreases, nformaton s created and exchanged among nterested actors who become more aware of the proft potental or opportunty avalable, thus contrbutng to the development of an ndustry s cogntve legtmacy. As partcpants ncrease ther exposure to developments n the ndustry through an ncreasng volume of meda coverage, the ndustry becomes ncreasngly taken-for-granted and legtmate n ther mnds. Ths should motvate the foundng of new organzatons or the expanson of exstng ones to take advantage of those opportuntes; these actvtes contrbute further to the process as they are reported upon and dscussed n the news meda. The argument here s that as the overall level of publc dscourse of a partcular ndustry ncreases, through the arena of the mass meda, that ndustry wll gradually become seen as a natural part of the envronment and the strateges and organzatonal forms assocated wth that ndustry wll become ncreasngly taken-for-granted, and as a consequence the number of entrants nto the populaton should also ncrease: Hypothess 2: The volume of general meda coverage of an ndustry postvely affects the entry rate. Postve and negatve evaluatons n meda coverage. Beyond emergng from awareness or taken-for-grantedness among enttes, legtmacy also ntegrates the socal values of the broader socal envronment wth the actons of those enttes (Berger and Luckmann, 1966: 92-92; Parsons, 1960). In organzatons, respondng to external demands rooted n a socal value system results n a change process n whch achevng legtmaton means takng on value and becomng frmly embedded n the system tself (Selznck, 1949; Selznck, 1957; Zald and 13

15 Denton, 1963). Socopoltcal legtmacy depends on a postve evaluaton and approval of an organzaton or ts actvtes from elements n the envronment, and an organzaton gans legtmacy when t makes a contrbuton to stakeholders based on crtera rooted n the system of values an belefs operatng n the envronment (Meyer and Rowan, 1977; Rowan, 1982; DMaggo and Powell, 1983; Scott and Meyer, 1983). Those postve judgments can center on a number of organzatonal factors. Organzatons are evaluated on the socal and economc mert of ther doman of actvty, the consequences of ther actons, as well as the methods and procedures used n producng those outcomes (Dowlng and Pfeffer, 1975; Suchman, 1995). For nstance, organzatonal actvtes are legtmate to the extent that they produce results that are consstent wth socally constructed notons of worthness (e.g., hgh economc or ethcal value), and by the degree to whch the processes used are consdered useful (e.g., scentfc or ratonal). Acceptance and postve evaluaton can stem from the degree n whch practces conform to expectatons regardng organzatonal actvty, whch mght nclude the percepton that the organzaton has sound management practces and approprate busness strateges or that t operates n a way that s consstent wth ethcal standards or legal regulatons. It s also necessary n many nstances for ndustres to be seen as havng new and untapped opportuntes for growth from emergng markets or from the applcaton of a technologcal breakthrough n ts products, servces, or operatons. Delacrox and Rao (1994) noted that organzatonal densty was partly reflectve of the reputaton of the ndustry among nterested enttes. Densty demonstrates the reputaton of the ndustry to observers, ncludng both awareness and respectablty of organzatons and strateges n a populaton. In the case of emergent ndustres new entrants must establsh a postve evaluaton and approval of ther strateges and product offerngs to consumers and nvestors. The 14

16 reputaton generated n ths process bulds famlarty wth the ndustry among nterested actors and allows them to generate evaluatons of both ndvdual organzatons and the ndustry as a whole. Crucal to ths part of the socopoltcal legtmaton process s the role of fnancal news coverage n the exchange of knowledge and nformaton regardng organzatonal actvtes. In cases of populatons of proft-seekng busnesses, fnancal organzatons may be key stakeholders n the evaluaton of organzatons and the reputaton of the ndustry, and nformaton exchanged wthn the arena of the busness press should play a role n the development of socopoltcal legtmacy through ts nfluence on flows of resources between ndustry partcpants. News coverage regardng nvestment n the market, partcularly n entrepreneural actvty and new product development, should tend to sgnal to organzatons and potental entrepreneurs whether there are opportuntes avalable n the market. Pollock and Rndova noted that the nformaton the meda provde about a frm may affect the processes of mpresson formaton and legtmaton not only through the volume of the nformaton, but also through ts tenor, or ts framng as postve or negatve (2003: 634). Indcatons of publc nterest n nvestng n the area, corporate nterest n developng new products or n buldng relatonshps wth other frms that are already n the market, and the avalablty of venture captal for nvestment n the market should have a postve mpact on foundng and entry rates nto the populaton. On the other hand, meda coverage of the possblty or actualty of bankruptcy or other modes of falure n the market, or of a shakeout or overall downturn n the prospects of the ndustry, should negatvely effect foundng and entry rates. Hypothess 3a. The volume of postve ndustry evaluatons n meda coverage of an ndustry postvely affects the populaton entry rate. 15

17 Hypothess 3b. The volume of negatve ndustry evaluatons n meda coverage of an ndustry negatvely affects the populaton entry rate. Coverage of governmental and regulatory nterests. Governmental or other regulatory nterests nfluence organzatons by constranng or releasng resource flows to them, montorng and lcensng ther actvtes, and alterng compettve relatonshps n organzatonal populatons (Baum, 1996: 95-96). Because of ther powerful nfluence on organzatonal forms and practces, these factors are frequently ncluded n research studes. For nstance, Hannan and Freeman's (1987, 1988) work on labor unons examned the effect certan laws had durng dfferent hstorcal perods on foundng and falure rates. Edelman (1990, 1992) focused on the nfluence the legal envronment of the tme had on the dffuson of due process practces and equal employment opportunty structures n organzatons. Sngh et al. (1991) looked at the nfluence dfferent laws had on the foundng and falure of nonproft chartable organzatons n the cty of Toronto. Beyond the effects of laws durng partcular tme perods, researchers have sought other measures to assess the connecton between regulatory factors and legtmacy. In the study conducted by Sngh et al. (1991), the further effects of organzatonal certfcaton by regulatory enttes were also analyzed. Baum and Olver (1992) found that the relatons between lcensed day care facltes and government and communty agences nfluenced foundng rates. The resources avalable to those agences through budget allocatons also mpacted foundng rates. Although the typcal strategy n ecologcal studes s to look at tme perod or relatonal effects stemmng from regulatory enttes, the nterest here s n how knowledge of governmental nvolvement and potental or actual regulaton nfluences legtmacy, and subsequently the entry rate nto the market. Meda coverage of such events as congressonal hearngs, executve speeches and statements, and departmental or agency actvtes or regulatons of commercal 16

18 actvtes all contrbute to the socopoltcal legtmacy of the ndustry. However, government and regulatory agences can both support an organzatonal populaton as well as lmt or restran t. For nstance, the research cted earler on Toronto nonproft organzatons reported that durng the early 1970s regulatory actvtes fostered the emergence of new organzatons, but regulatory actons later that decade reversed those effects (Sngh et al., 1991; Tucker, Sngh, Menhard, and House, 1988). In stuatons where a new ndustry or organzatonal form s emergng or seekng to become establshed, governmental attenton can postvely nfluence legtmacy and densty, snce such actons gve approval and stature to the emergent ndustry as well as sgnal the government's wllngness to advance or support t (Aldrch, 1999: 230). For nstance, Budros (1992) found that the passage of a crucal pece of legslaton provded legtmacy to the fledglng New York state lfe nsurance ndustry and postvely affected organzatonal entres. In the feld of Internet commerce, governmental agences and legslatve actvty n the Unted States have predomnately served to facltate onlne commercal actvty. Laws such as the Natonal Communcatons Competton and Informaton Infrastructure Act and the Internet Tax Freedom Act serve to promote busness on and access to the Internet, and the Natonal Informaton Infrastructure ntatve brought attenton to the potental of nformaton systems and wdespread Internet access (Moschovts et al, 1999). Thus, the expectaton s that: Hypothess 3c. The volume of meda coverage of governmental nterest n a partcular ndustry postvely affects the populaton entry rate. Combnng densty-dependent and meda-based legtmacy effects. The dscusson of hypotheses 2 and 3 sought to expand the understandng of the concept of legtmacy by dentfyng ts underlyng cogntve and socopoltcal components and the effect of each on market entry rates. By combnng both aspects of legtmacy wth densty-based ndcators of 17

19 legtmacy, the populaton ecology model of densty-dependence can be extended wth a more complete understandng of the mpact of both cogntve and socopoltcal legtmacy on the process. As stated earler, the legtmacy-competton model of densty dependence clams that densty works n two processes affectng the vtal rates n a populaton. The process of legtmaton s based on the vew that early ncreases n densty reflect the establshment of the vablty and approprateness of organzatonal actvtes; thus, densty s proposed as postvely affectng foundng or entry rates. On the other hand, competton s based on the observaton that further ncreases n densty contrbute to a scarcty of resources allocated among a large number of frms; n ths case, densty s vewed as negatvely affectng foundng rates. The need to accommodate densty dependence as both a negatve and postve functon has been accomplshed tradtonally usng curvlnear functons of densty; ths perspectve forms the bass of hypotheses 1a and 1b dscussed prevously. However, f densty does ndeed reflect an underlyng multdmensonal legtmaton process n the ndustry, and f cogntve and socopoltcal legtmacy can be effectvely measured by lookng at the dscourse and content of meda coverage of the ndustry, then these measures should contnue to be sgnfcant even when assessed smultaneously wth denstybased measures. Measures of cogntve and socopoltcal legtmacy should be related to entry rates due to ther value n transmttng nformaton about emergent forms of organzatons or strateges, whle densty should reman related to entry rates both as an ndcator of actual actvtes n the populaton and as a measure of the competton among market partcpants. Hypothess 4: The legtmacy and competton effects of populaton densty, as well as cogntve legtmacy effects of meda coverage, are related to the entry rate. 18

20 Hypothess 5: The legtmacy and competton effects of populaton densty, as well as socopoltcal legtmacy effects of meda coverage, are related to the entry rate. Table 1 summarzes the hypotheses proposed n ths study Insert Table 1 About Here DATA AND RESEARCH METHODS The rse of e-commerce and Internet busness provdes a hstorcally convenent and nterestng case n whch to study the process of ndustry emergence and legtmaton. Whle only a small part of the wde range of actvtes collectvely known as electronc commerce, the Internet access busness s a crucal element n that phenomenon snce t provdes essental nfrastructure requred for electronc transactons to take place (Barua, Pnnell, Shutter, and Whnston, 1999; Coleccha, 1999; Organzaton for Economc Co-operaton and Development, 1997). Whle the Internet access ndustry tself conssts of many dfferent elements, some of whch are clearly stll emergng n the marketplace, the focus of ths study s the ntroducton and legtmacy of organzatons provdng servces assocated wth hgh speed, broadband Internet access between 1993 and the frst quarter of Cable and DSL Broadband Technologes A prelmnary examnaton of accounts of the ndustry n the busness and trade press, a revew of publc utlty commsson publcatons on the broadband ndustry, and nformal dscussons wth trade assocaton and publc utlty commsson staff members revealed that two technologes are the domnant competng forms of broadband access to consumers: cable and 19

21 dgtal subscrber lne (DSL) access. Both cable access and DSL are frequently consdered together n dscussons of broadband, both have smlar technologcal roots, both appeal to the same customer groups by provdng essentally dentcal servces, and broadband access provders often see each other as compettors regardless of any technologcal dfferences. More recent wreless and satellte technologes stll have a long way to go to have any mpact on the ndustry, n terms of technologcal development, market share growth, or ndustry attenton, and so are not consdered here n ths study. Broadband cable access s a technology that produces a very hgh-speed Internet connecton over the coaxal and fber optc cable nfrastructure used to delver cable televson. Internet access s provded through the exstng cable televson wrng present n most resdences. The technology conssts of both the connecton servce as well as specalzed equpment, such as a cable modem and networkng hardware, used to connect the user's personal computer to the cable company's network. Cable access avods the use of copper telephone lnes used n dal-up servces n order to provde data transfer speeds many tmes faster than normally avalable. The prmary competton for cable access n the broadband market dgtal subscrber lne, or DSL s a hgh-speed access method that extends the technology n tradtonal telephone lnes and hardware. Dgtal sgnals formng the user's connecton to the Internet are transferred over standard copper telephone lnes avalable n nearly every home and offce. Unlke dal-up modem connectons, DSL allows the smultaneous use of both Internet access and normal telephone servces whle addtonally provdng up to ffty tmes faster data transfer rates. The technology combnes both the network connecton servce as well as requred hardware, such as 20

22 a specal DSL modem and networkng nterface cards, used to make the lnkage between the user's personal computer and the wall telephone jack. Ths research settng s approprate for a study on the emergence of legtmacy and the creaton of new organzatonal populatons for a number of reasons. The rse of the broadband access ndustry has been characterzed by sgnfcant support from large, well-funded corporatons wth sgnfcant nterests n buldng telecommuncatons nfrastructure and establshng ther domnance n the ndustry. Snce cable access s provded over some of the same nfrastructure as cable televson, servce s typcally offered by cable system companes, lke AOL Tme Warner, AT&T Broadband, Cox Communcatons, etc. DSL technology s based on extensons of exstng telephone communcaton technology and s thus typcally offered by the regonal Bell operatng companes, such as SBC Communcatons and Verzon. For companes operatng n other ndustres, movng nto broadband sgnfcantly redefnes ther tradtonal character, potentally allowng expanson nto totally new domans of actvty but exposng them to new competton n areas where they have lttle experence. In addton to preexstng companes, a varety of new forms of organzaton have emerged typfed by examples such as Covad Communcatons and Rhythms NetConnectons dedcated solely to developng and takng advantage of new broadband technologes. In addton, the ndustry receves a tremendous amount of attenton from regulatory enttes nterested n controllng the deployment of new broadband servces, and who feel empowered by ther authorty over older telecommuncatons forms. Added to ths stuaton s the relatve newness of broadband and lack of awareness among busnesses and consumers durng ths perod; the overwhelmng proporton of Internet users at the tme stll connected through tradtonal dal-up modem servces 2 2 Boardwatch Magazne (2000) reported that over nnety percent of Internet users connected through dal-up servces, whle cable and DSL users amounted to just under fve percent. 21

23 (Boardwatch Magazne, 2000: 13). The ndustry represents a convergence of a number of competng nterests nterested n defnng the approprate forms of organzaton and busness, types of servces, and codes of conduct among partcpants. Snce the legtmacy of the broadband ndustry was stll beng negotated among busness and consumer users, servce provders, technologsts and ndustry experts, fnancal analysts, regulators, and other nterested socal actors, the broadband Internet access ndustry provdes an deal settng to study the effects of legtmacy on the emergence and development of new forms of organzatonal actvty. Sample Selecton The provson of broadband servces s based on the development of an organzaton and nfrastructure at the local level; even for frms that are natonal or global n scope, broadband actvtes are operated at the cty, or even the neghborhood, level (Dempsey, Frsan, Mehrotra, Rao, and Whte, 1998). Broadband nfrastructure s centered on the lnkage between ndvdual customers and nearby local network facltes that aggregate connectons to the Internet; t ncludes nvestment n cable connectons to customers physcal facltes, network upgrades and mantenance, and computer servers that route data between customers n and the Internet. Lkewse, broadband competton s also localzed; because of the need for close proxmty between broadband operatons and customers, servce provders n one metropoltan area do not compete wth servce provders n another. Because of these factors, the broadband ndustry can be characterzed by a group of subpopulatons whose boundares are defned at the cty level, and that nclude entres by both new frms as well as exstng frms n each cty subpopulaton. These factors suggest an establshment level of analyss, as descrbed by Carroll and Hannan (2000: 51); densty-dependent compettve effects operate locally, n each of the geographc locatons that companes operate n. 22

24 Furthermore, the localzed nature of broadband access creates a stuaton where economes of scale accrue to frms that can access the most number of customers wthn the geographc reach of ther network facltes. In other words, broadband provders ntally start up n urban areas wth hgh resdental populaton densty and appear later n less dense metropoltan areas and rural settngs. Any research sample chosen must account for effects on foundng and entry rates produced by the dffuson of broadband nto dfferent resdental populaton levels rather than by legtmacy and competton. Therefore, ths study covered the populaton of broadband provders n the sx largest ctes n Texas Austn, El Paso, Dallas, Fort Worth, Houston, and San Antono allowng for the observaton of ndustry actvtes n a typcal set of large metropoltan areas, and avodng extreme dfferences n urban populaton densty as an mportant factor. Restrctng the choce of ctes to one state also has the beneft of controllng for potental dfferences between state-level regulatory regmes. Wth respect to the tme frame for the study, the assumpton used was that the begnnng of the resdental-level Internet access ndustry began n 1993 when the graphcal Internet browser Mosac was developed by the Natonal Center for Supercomputng Applcatons (Moschovts et al, 1999). Ths applcaton resulted n popular nterest n the Internet and ts nformaton nterface known as the World Wde Web, and motvated frms to provde access to those nformaton servces. The resultng sample for ths study covered all the hgh-speed broadband provder foundngs and entres, measured n monthly tme ntervals between January 1993 and Aprl 2000, n the sx largest ctes n the state of Texas, yeldng a total of 528 observatons (88 monthly perods 6 ctes). To mantan consstency wth organzatonal ecology research, lagged ndependent varables were used resultng n the loss of one tme perod of data. 23

25 Populaton Entry Rate and Densty Varables. The entry rate for each cty s defned here by the number of new broadband provder entrants (offerng ether broadband cable or DSL access) that occurred n each cty n a partcular month. The populaton densty of each cty s the total number of actve ndustry partcpants n each cty n each month. Followng past practces n the lterature, frst- and second-order forms of populaton densty are used to reflect both legtmacy and competton processes n the populaton. Snce no sngle source exsts documentng the number of broadband compettors n the populaton, a number of sources were used to determne the hstory of entres and exts n the ndustry, ncludng varous regulatory, trade, and news sources. The frst task n ths data collecton effort nvolved generatng a lst of broadband servce provders n the state of Texas. The Report to the 77th Texas Legslature on the Avalablty of Advanced Servces n Rural and Hgh Cost Areas from the Texas Publc Utlty Commsson provded nformaton about the number of broadband servce provders and the broadband cable operators actve n each cty (Publc Utlty Commsson of Texas, 2001). Servce provders were also requred to fle documents wth the publc utlty commsson every year on the type of servces offered and the areas of servce avalablty. Although these sources do not provde a complete record of the populaton, ths nformaton was used to generate a lst of servce provders for further nvestgaton. Produced by the U.S. Department of Commerce and the Department of Agrculture, Advanced Telecommuncatons n Rural Amerca provded some nformaton on the broadband cable and DSL provders across the U.S. (Natonal Telecommuncatons and Informaton Admnstraton and the Rural Utltes Servce, 2000), whle CED Magazne 24

26 provded a much more comprehensve lst of broadband cable and natonal DSL servce provders (CED Magazne, 2001). After generatng a lst of broadband servce provders for further study, a comprehensve search of news sources and company nformaton on each provder produced a record of the entres and exts n each of the sx Texas ctes studed. The Lexs-Nexs database of news artcles was used to dentfy announcements of new servce ntroductons or wthdrawals from servce. Lexs-Nexs was deal for ths purpose because of ts coverage of both natonal and local news sources and ts ablty to access the full text of artcles. When an artcle was found whch dentfed an entry event, t was examned for the date of servce ncepton and the locaton n whch broadband access was made avalable; when no specfc date was ndcated, the artcle publcaton date was used to ndcate the month n whch the entry occurred. Company nformaton was also used to dentfy new market entres; the use of company web stes to promote products and servces to the publc and to communcate to external stakeholders about frm actvtes facltated the search for nformaton about foundngs and entres. Frequently, frms publshed press releases announcng entry events on ther web stes; some frms also released such nformaton n ther fnancal statements. In addton to the web stes of ndvdual companes, the Internet Archve was also used to dentfy changes n nformaton on company web stes that occurred over the years, or even to access corporate stes that no longer exst due to termnaton of the busness. 3 When announcements from company nformaton were found, the text was examned for the date of entry or ext; when a specfc date was not mentoned, the date of the announcement was used as the date of the entry or ext event. 3 The Wayback Machne at the Internet Archve s a collaboratve project, ncludng organzatons lke the Lbrary of Congress and the Smthsonan Insttuton, to allow access to web stes that have been publshed n the past but may no longer be avalable, ether because the nformaton was subsequently updated or because the companes that publshed them no longer exst. The archve provdes a hstorcal record of the nformaton that has appeared on Internet, recordng that nformaton for future revew. The Wayback Machne at the Internet Archve can be found at 25

27 The data from each cty sub-populaton are depcted n Fgure 1. The procedure employed n ths study dentfed a total of 55 entres across the sx ctes, by 17 companes provdng broadband cable or DSL servce durng the perod between January 1993 and Aprl Durng the perod of ths study, no frms ceased operatons n any of the ctes studed; subsequent analyss found market exts began n late 2000, outsde the scope of ths research project. The lack of exts from the populaton durng ths perod s the result of substantal rght-censorng n the data set; organzatonal exts and falures had smply not occurred wthn the hstorcal perod of the study. However, ths poses few problems for ths study snce left-censorng s typcally vewed as the most crtcal concern, and studes are often desgned to gather data on the earlest hstorcal perods possble n organzatonal populatons (Carroll and Hannan, 2000). The focus on the earlest perod n the hstory of the ndustry s actually benefcal to the study, snce the objectve s to examne the mpact of legtmacy on the creaton of new organzatonal populatons. Densty effects on foundng and entry can be attrbuted to the emergng legtmacy of ths form of busness organzaton because only the early hstory of the populaton s consdered, whle studes that nclude both early and late hstorcal perods can fnd densty effects that can confuse legtmacy effects wth processes of market power or resource parttonng (Baum, 1996: 89-90) Insert Fgure 1 About Here Whle the data s rght-censored, t does not exhbt left-censorng because the tme perod studed begns years before the frst entry actvty. The frst entrant began servce n May 1996 n Houston. However, much of the entry actvty started n late 1998 and contnued through early 26

28 2000. Despte lttle entry actvty n the early perod of the study, ndustry partcpants were actvely engaged n establshng what would eventually become the broadband access busness, ncludng development of the technologes used n provdng and usng the servce, nvestment n the creaton of entrely new organzatons and the expanson of exstng ones and the constructon of technologcal nfrastructure, and the lobbyng of both nvestors and regulators for support n takng advantage of the new opportuntes beng dentfed. To get a pcture of the extent of ndustry actvty gong on before the frst market entrant appeared, we must turn to other, non-densty-based measures of legtmacy. Meda Coverage Varables Data sources for meda coverage varables. In examnng ndustres nvolvng the Internet a number of news and nformaton sources are avalable for research purposes, many of whch are accessble n electronc form over the network tself as well as other tradtonal prnt forms. Onlne sources of Internet technology news, such as ZD Net ( and CNET ( are popular sources of onlne news and nformaton regardng many ndustres nvolved n Internet busness actvtes. Many tradtonal book publshers also publsh works on computer and networkng topcs, although even a casual browsng of the books avalable wll ndcate that many are techncal n orentaton rather than focused on characterstcs of the ndustres themselves. In addton to these sources, there are also many prnt perodcals that focus on Internet technology and busness actvty, lke Wred, Red Herrng, and Fast Company, whle more general nterest busness journals (e.g., Busness Week, Inc., etc.) also frequently publsh artcles n ths area. Wth ths wealth of nformaton avalable, a careful consderaton of the needs of the study s necessary n choosng the data sources to be used n a content analyss. The selecton of 27

29 documents used as the populaton of nterest depends on the hypotheses beng tested and the nferences ntended by the study as well as the dstrbuton of nformaton among content sources (Krppendorff, 1980: 65; Weber, 1990: 42). The nformaton necessary to test partcular hypotheses may not be evenly dstrbuted among dfferent sources; varous onlne nformaton sources, books, and prnt perodcals may not all supply the needed nformaton equally well for the purposes of the study. In addton, the needed nformaton must be easly accessble and avalable over the entre hstory of the ndustry under study. In measurng legtmacy, the concern n dentfyng the approprate populaton to draw content from s dependent on the ablty to access nformaton about the taken-for-granted character of organzatonal actvtes and the evaluaton and approval of those actvtes by nterested socal actors (such as entrepreneurs, nvestors, regulators, etc.). The content used should thus be a form of meda that extensvely reports on ndustry ssues and that nterested socal actors frequently access and depend on for mportant nformaton on the ndustry. Although populaton events n the broadband ndustry occur at the local level, the companes that provde access servces often operate across many areas, and the technologes they use are not specfc to any partcular regon. Because of ths, local nformaton sources (e.g., coverage n local cty newspapers) are not as crtcal to the legtmacy of the entre ndustry as s content that s applcable across all regons and to the ndustry as a whole; ths s also consstent wth the fndngs of Hannan and hs colleagues (1995) who found that legtmacy operates at a level above local actvty. The ABI/Inform database provded a good source of content for analyss n ths study, snce t contaned an extensve ndex to artcles n all areas of busness and management as well as access to ther full ctatons and abstracts. The ABI/Inform database covered over 1,800 professonal journals and trade publcatons, effectvely coverng the 28

30 majorty of avalable sources of busness knowledge that mght be relevant n the legtmaton process. Beyond the wde range of publcatons and subjects covered, ABI/Inform also provded access to content gong back to 1971, allowng for the collecton of data over the entre relevant hstory of the ndustry beng studed as well as easy to use searchng tools whch would facltate access to the nformaton. The dsadvantage of usng ABI/Inform was that t only provded comprehensve access to artcle abstracts rather than the full text of the artcles themselves, although ths problem was mnmzed by the use of the artcle abstract as the samplng unt and the fact that the nformaton needed by ths study s largely contaned n the abstracts makng the use of full texts unnecessary. To produce a body of text to use n the content analyss, a search for artcle abstracts n the ABI/Inform database was conducted and a multstage samplng procedure employed, utlzng both stratfed and systematc elements. Employng a seres of samplng technques n ths manner has been common practce n research usng content analyss, and helps captalze on the characterstcs of the content data tself to generate a representatve sample. As a startng pont n the content analyss, a combnaton of search terms was chosen to dentfy publshed artcles on the subject of broadband cable and DSL access technologes. A separate search was then conducted usng the chosen keywords for each month n the perod from January 1993 to Aprl 2000, generatng a lst of artcles appearng n the database each month on broadband access topcs. Fnally, n each month every fourth artcle abstract was systematcally selected from the lst of artcles startng from a randomly chosen pont, producng a 25% sample of artcle abstracts for subsequent analyss. The stratfed procedure compensates for varaton n the number of artcles publshed n each month, whle the systematc procedure facltates the selecton of a random sample drawn from a lst of artcle abstracts when the lst does not exhbt perodcty n 29

31 ts order (Budd, Thorp, and Donohew, 1967; Krppendorff, 1980; Rffe et al., 1998). Ths procedure produced a total of 1407 artcle abstracts for further analyss. General meda coverage varables. The volume of general meda coverage n a partcular month s the total number of artcles publshed each month, dentfed n the content analyss of the ABI/Inform database as pertanng to the broadband ndustry or to broadband cable or DSL technologes. Followng the gudelnes set out by Rffe et al. (1998), a content analyss procedure was formalzed n a content codng protocol. The protocol lsted all the categores used n the analyss and specfed how the coder was to assess the content n each artcle abstract. The protocol requred the coder to assess each artcle abstract n each month as to the types of technologes descrbed n the artcle (broadband cable, DSL, some other technology, or not about broadband), whether the artcle descrbed the ndustry and the opportuntes for expanson or growth postvely or negatvely, and f the artcle descrbed regulatory nterest n broadband. If the artcle descrbed any form of broadband access technology or the frms developng and sellng the technology t was vewed as contrbutng to the taken-for-grantedness of the ndustry and counted as ncreasng ts cogntve legtmacy. Fgure 2 shows the monthly general meda coverage of the broadband ndustry, ncludng coverage of both broadband cable and DSL technologes, over the 88 month tme perod of the study. The month-to-month level of dscourse on broadband technologes and servces starts at a low level and steadly ncreases over tme, ndcatng how broadband ncreasngly becomes recognzed and taken-for-granted among ndustry observers and partcpants. Interestngly, there s some level of dscusson about broadband even durng the earlest perod of the study, albet at a low level, durng a tme n whch there were no entrants or densty n the populaton. 30

32 Insert Fgure 2 About Here Postve and negatve ndustry coverage and governmental nterest coverage varables. Postve ndustry coverage represents the current volume of postve evaluatons of the broadband ndustry n the news meda and busness press, and encompasses nvestment actvty, publc offerngs of equty, ndustry growth, ncreasng levels of consumer demand, and venture captal nterest. Postve ndustry coverage s ndcatve of the socopoltcal legtmacy conferred by postve apprasals of the busness opportuntes present n the ndustry. The volume of negatve ndustry coverage s a measure of current negatve evaluatons of the ndustry, and s suggestve of the perceved potental for a downturn n the ndustry, poor prospects for busness growth, and of potental or actual frm falures. Governmental nterest coverage s measured by the current level of meda coverage of real or potental governmental regulaton of the Internet access market. Ths measure should ndcate the level of governmental nterest n the ndustry, and thus s representatve of the state's evaluaton and approval of the behavors of frms n the market. As n the case of the general meda coverage varable, the measurement of the ndustry evaluaton and regulatory meda coverage varables was guded by the content codng protocol. Each artcle abstract was examned for descrptons of growth and expanson of the ndustry or declne and frm falure, as well as any type of regulatory or governmental actvty. Snce news meda analyses of the broadband ndustry frequently dstngushed between broadband cable and DSL technologes, each technology was assessed separately for postve and negatve ndustry evaluatons. For example, a sngle artcle mght descrbe future growth and opportunty n the 31

33 broadband cable market but negatvely portray DSL; such an artcle would contrbute to both a measure of postve ndustry evaluatons as well as the negatve ndustry evaluatons measure. The total number of artcles descrbng the development of the ndustry postvely n each tme perod, coverng both DSL and broadband cable technologes, was used as a measure of postve ndustry evaluatons n that month. Conversely, the total number of artcles publshed each month descrbng the ndustry negatvely, regardng each form of broadband, was taken as an ndcator of negatve ndustry evaluatons. The total number of artcles publshed each month descrbng any form of governmental actvty or regulatory nterest n the broadband busness were used as a measure of the level of governmental regulaton n the ndustry. Fgure 3 depcts the volume of both postve and negatve evaluatons of the broadband busness n the news meda. Even as early as 1993, artcles appeared whch descrbed the opportuntes avalable n provdng broadband access servces or growth n the ndustry, ether n terms of growng demand for the servce or from ncreasng levels of nvestment by organzatons. Postve evaluatons grew at a steady pace over the entre tme perod n the study. Smultaneously, artcles negatvely descrbng the ndustry and ts prospects also appeared and steadly grew n number over tme, although the magntude of negatve coverage was never as large and that of postve coverage. Fgure 4 shows the monthly volume of coverage descrbng regulatory concerns or governmental nterest n the ndustry. Compared to postve and negatve evaluatons, regulatory coverage always appeared at a low level; however, the pattern of coverage showed the dscourse of the ndustry focused on the nfluence of the Telecommuncatons Act of 1996 and ts mplcatons for the broad deployment of telecommuncatons servces. 32

34 Insert Fgures 3 and 4 About Here Relablty of content analyss measures. In any study that employs a content analyss methodology, a major concern for the researcher s the relablty of the analyss procedure. Employng a hghly relable procedure enhances the objectvty of the study by dstngushng content analyss as a method of scentfc nqury from other methods used to analyze textual data, and nsures that the data produced by the procedure s consstent regardless of the researcher performng the analyss or when that analyss s performed (Berelson, 1954; Kassarjan, 1977). The formulaton of content categores and codng rules that dfferent researchers can consstently apply s a necessary ntal step n buldng a relable procedure (Kassarjan, 1977; Rffe et al., 1998), and that process s ted to the dentfcaton and operatonalzaton of the varables dscussed prevously. After ths takes place, the process shfts to focus on the assessment of how relably those procedures can be used n acqurng data. The frst step n assessng the relablty of a content analyss procedure s selectng a sample of the content materal to be used. Snce content analyss methods are often used to generate extremely large data sets, usng multple coders to replcate an entre content analyss study for an evaluaton of relablty s usually mpractcal. Usng addtonal coders to replcate only subsets of the content used n a study s vewed as suffcent n these crcumstances. When choosng the content for addtonal codng, Rffe et al. (1998) advocate usng a random samplng approach when performng a relablty test. In ths study, one of the authors performed the 33

35 prmary content analyss, whle a second coder was used to code a subset of the sample for the purpose of checkng the relablty of the codng procedure. 4 The second step n determnng the relablty of a content analyss study s the quanttatve measurement of the degree of consstency between coders. The most common method for determnng the relablty of a study s to calculate the percentage agreement between codng judges, whch s smply the proporton of agreement between two coders as a percentage of the total number of codng decsons (Kassarjan, 1977: 14; Kolbe and Burnett, 1991; Rffe et al., 1998: ). When the proporton of agreement method s used, a reasonable expectaton s that the procedure produces a percentage agreement near 80% or greater between codng judges. If substantally less than 80% agreement s reached, than the relablty of the study has not been establshed. Often researchers also report a relablty fgure for each category coded n addton to an overall measure; such category relablty evaluatons are better than an overall measure snce aggregatng multple relablty scores may dsguse poor performng categores. In ths study, none of the categores used had less than 88% agreement between coders, and many categores had over 90% agreement. Control Varables 5 Resdental populaton. The resdental populaton at the cty level (n hundreds of thousands of people) was ncluded to account for demographc factors that mght mpact the entry rate. In the course of ths study, t was found that companes often cted local resdental populaton level as a consderaton for entry nto a partcular cty. As stated earler, the 4 For ths study the second coder was chosen by selectng an ndvdual wth no background n busness educaton or n telecommuncatons technology. In choosng the second codng judge n ths fashon, the focus s on the codng procedure rather than the nature or characterstcs of the coder to accurately assess the text beng analyzed. The codng protocol, supplemented wth addtonal nformaton on the background of the ndustry and defntons of the technologes and termnology used n the ndustry, were collected together and used as a nstructonal tool or manual to tran the second coder n the use of the procedure and the meanng of the categores contaned n the protocol. 5 When partcular control varables were not avalable on a consstent monthly bass, well establshed lnear methods were employed to transform the avalable data nto a format sutable for subsequent statstcal analyss. 34

36 technologes underlyng broadband telecommuncatons servces depend heavly on access to hgh populaton densty levels wthn close physcal dstance to servce provder facltes to acheve scale economes. Broadband access s a range-dependent technology; customers can be provded access as long as they resde wthn a certan servce area. Hgher resdent populaton levels wthn potental markets allow for greater use of the communcatons nfrastructure nvestment, and could potental nfluence the pattern of entry rates n ways unrelated to legtmacy and competton processes. Annual populaton data was gathered from the U.S. Bureau of the Census. 6 GDP growth rate and unemployment rate. Economc varables are frequently used n foundng and entry rate studes as control varables; entry rates patterns could result from changes n economc condtons rather than from legtmacy or competton effects. The GDP growth rate was used to represent natonal economc condtons whle the cty-level unemployment rate was used to control for local economc factors. GDP measures were avalable from the U.S. Bureau of Economc Analyss on a quarterly bass, whle the unemployment measures were obtaned from the U.S. Bureau of Labor Statstcs on a monthly bass. 7 Internet growth rate. Fnally, to account for the growng sze of the Internet as a factor n the emergence of the broadband access ndustry, a measure was ncluded as a control whch represented the ncrease n the number of Internet users and the amount of nformaton avalable on the Internet. The Internet growth rate s the rate of ncrease n the number of hosts connected to the worldwde nformaton network. The measure s based on data provded by the Internet Software Consortum, and s generated by a methodology that systematcally looks for all the 6 Data from the U.S. Bureau of the Census are avalable onlne at 7 Data from the U.S. Bureau of Economc Analyss are avalable onlne at whle data from the U.S. Bureau of Labor Statstcs are avalable onlne at 35

37 computng devces connected to and assgned an address on the Internet. 8 Whle not representng the number of users of the Internet, t does represent n a relable way the number of systems on the network, and thus s a proxy for demand for access to and the amount of nformaton avalable on the Internet. Model Specfcaton and Statstcal Analyss In models where the entry or foundng rate s used as the dependent varable, consderaton must be gven to the fact that the rate varable s measured usng count data. Count data are characterzed by havng non-negatve, nteger values; frequently, count data also have many nstances of zero or very low values, whle hgh values are rare. When the data collected n a study take ths form, lnear regresson models are not the best choce snce the data are not normally dstrbuted. Instances where data are n the form of countng measures are wdespread n socal and economc research and frequently appear n organzatonal ecology studes, and statstcal models are avalable that more effectvely account for the dstrbuton of countng process varables. Typcally, count data are modeled usng varous forms of Posson regresson, where the dependent varable s drawn from a Posson probablty dstrbuton (Greene, 1997: 931; Hannan and Freeman, 1989: ; Long, 1997: 217; Neter, Kutner, Wasserman, and Nachtshem, 1996: 609). In ths applcaton, the rate of entry Y s assumed to follow a Posson process, and the probablty functon s gven as follows: Y µ f ( Y µ ) = exp( µ ), where Y = 0, 1, 2, Y! A dstngushng feature of the Posson dstrbuton s the equalty between the mean and the varance of the outcomes Y: 8 The data collected by the Internet Software Consortum are avalable onlne at 36

38 E ( Y ) = Var( Y ) = µ. A Posson regresson model can be set up where the mean entry rate of the th case s a functon of a set of predctor varables; the typcal formulaton of the Posson regresson model s the loglnear model: µ Y = E( X ) = exp( X β ), where = 0, 1, 2, n. In ths equaton, X s a vector of predctor varables for the th case and β s a vector of regresson coeffcents to be determned. By usng the log-lnear model, the mean response s constraned to postve values. Wth ths model, the probablty dstrbuton of the outcomes Y becomes condtonal on the characterstcs of each case as ndcated by the predctor varables X n the data: µ f Y X ) = exp( µ ). Y! ( Whle some studes n the feld of organzatonal ecology have used the standard form of the Posson regresson, most have appled more extended models n ther analyses whch address some of the restrctons requred wth ts use (Cameron and Trved, 1986; Long, 1997). Frequently, data used n practce by researchers feature dependent varables wth varances whch exceed the mean; ths s called overdsperson, and t results from unobserved heterogenety among observatons. Unobserved heterogenety among observatons occurs when dfferent cases are not equally lkely to produce occurrences of the counted varable; ndvdual observatons wth the same X may not have the same µ (Long, 1997: 231). The presence of overdsperson n the data makes the use of the Posson model problematc; overdsperson produces varances whch are larger than the mean, thus volatng a key assumpton of the model. Usng a Posson regresson model when there s overdsperson wll underestmate the number of Y 37

39 observatons at the extremes of the dstrbuton (.e., t wll underestmate the number of zeros and the number of large values) and produce overestmates of the sgnfcance of the ndependent varables. A more general extenson of the Posson model called the negatve bnomal model s frequently employed by researchers when t s necessary to relax assumptons n the Posson model regardng the form of the mean and varance of the underlyng probablty dstrbuton. The negatve bnomal regresson model adds a random error term δ to the mean µ ; the new mean λ s now a random varable that takes the followng form: λ = exp( X β + ε ) = exp( X β )exp( ε ) = µ δ. The mean of the error term δ s set to 1, so that the expected outcome s equvalent to the Posson process: E ( Y ) = E( λ ) = E( µ δ ) = µ E( δ ) = µ. The dstrbuton of Y s now condtonal on both X and δ : Y Y λ ( µ δ ) f Y X, δ ) = exp( λ ) = exp( µ δ ). Y! Y! ( In order to create a regresson model that s a functon of only the predctor varables X, then the dstrbuton of Y must be adjusted by the probablty dstrbuton of δ (whch s typcally a gamma dstrbuton): f ( Y X ) = f ( Y X, δ ) f ( δ ) dδ. 0 The probablty functon of Y now follows a negatve bnomal dstrbuton, and takes the followng form: f Γ( Y + ν ) ( Y X ) = Γ( Y + 1) Γ( ν ) ν + µ ν + µ. µ Y ν ν 38

40 The mean of the negatve bnomal dstrbuton s the same as the Posson: E( X ) = = exp( X β ). Y µ However, the varance s now much dfferent from the Posson formulaton: 1 Var( Y X ) = µ 1 + µ = exp( Xβ ) 1+ β ν 1 where α =. ν [ α exp( X )] In ths specfcaton of the negatve bnomal model, α s a dsperson parameter ndcatng the level of overdsperson n the data. When the overdsperson s small, α goes to 0 and the model reduces to the standard Posson regresson model. Tests exst whch can ndcate the need for the negatve bnomal regresson rather than the Posson regresson procedure (Greene, 1997: ; Long, 1997: ); these tests center on evaluatng the hypothess H 0 : α = 0. For an extended dscusson of the dervaton of the Posson and negatve bnomal regresson models, Cameron and Trved (1986) provde an extensve dscusson and comparson of regresson models of count data whle Greene (1997) and Long (1997) provde more recent and easly accessble revews of the Posson and negatve bnomal models. Software to estmate such models, as well as the comparson of dfferent models for dfferences n ft, s wdely avalable n commercal statstcal packages such as SAS and Stata (Allson, 1999; Long and Freese, 2001; StataCorp, 2001). RESULTS Table 2 reports the negatve bnomal regresson estmates for models of the entry rate across all sx ctes between January 1993 and Aprl Snce the lkelhood tests of the estmated overdsperson parameter α demonstrated mxed results, t was decded to model the 39

41 heterogenety n the data by usng negatve bnomal regresson nstead of the Posson regresson model. Dong so produced a more conservatve test of the sgnfcance of the parameters, snce the tendency of the Posson model to overestmate sgnfcance levels n the case of overdsperson was avoded. The pseudo R-squared and the AIC of each model are also provded n Table 2 as ndcators of ft wth the data Insert Table 2 About Here Testng Hypothess 1 The Densty-Dependent Model of Entry Rates Hypotheses 1a and 1b test the predcton of an nverted U-shaped relatonshp between densty and entry rates. As shown n Table 2, the parameter estmates of the baselne model of densty-dependent legtmacy and competton effects on entry rate are reported under model 1. Both the frst-order and second-order forms of populaton densty are sgnfcant at the 0.01 level wth parameter estmates n the expected drectons, provdng support for both legtmacy and competton effects of densty. Interestngly, even n the earlest phases of ndustry creaton, densty-dependent legtmacy effects are stll strongly evdent, and competton effects are stll observed at a sgnfcant level, both n an envronment wth few frms yet n operaton. Model 1 thus provdes support for the nverted U-shaped relatonshp between densty and entry rates among broadband access provders, as predcted by the populaton ecology model, even n the formatve stages of a populaton's hstory. 9 Any one measure of ft only provdes a "rough" gude to the ft of the model (Long, 1997; Long and Freese, 2001). Focusng on a sngle ftness measure should not be the only method used to assess the performance of a model; such an approach results n the optmzaton of only one crteron, possbly at the expense of other equally mportant consderatons. Furthermore, most goodness-of-ft measures have weaknesses whch prevent them from beng completely representatve of the adequacy of a partcular model. Therefore, the statstcal results reported n ths study gve both the pseudo R-squared measure as well as an nformaton-based measure, Akake's nformaton crteron, or AIC. Assessment of the ftness of sngle model or the comparson of multple models should balance the results of multple ft crtera, whle also consderng the theory beng tested as well as past research as addtonal gudes. 40

42 Testng Hypotheses 2 and 3 Meda-based Model of Legtmacy Effects and Entry Rates General meda coverage model. The second model from Table 2 reports the estmates for the negatve bnomal regresson model used to test hypothess 2; hypothess 2 stated that the cogntve legtmacy evdent from the level of general meda coverage of the populaton postvely affects the entry rate. The parameter estmate for the general meda coverage varable was postve and sgnfcant n ts nfluence on entry rates (at the 0.01 sgnfcance level), demonstratng the legtmatng effect the current volume of communcaton and nformaton transmsson wthn the ndustry has on foundng and entry rates n the populaton. Postve/negatve and regulatory meda coverage models. Hypothess 3 stated that the socopoltcal effects of postve and negatve ndustry evaluatons and regulatory nfluences evdent through ndustry dscourse affected the entry rate nto the populaton. The thrd regresson model reported n Table 2 reports the sgnfcance of both postve and negatve evaluatons n the news meda as well as coverage of governmental and regulatory nterest n the ndustry on the entry rates. The parameter estmates of the model showed both postve ndustry coverage and the reportage of poltcal and regulatory ssues postvely affected the entry of broadband provders, supportng hypotheses 3a and 3c wth p-values at the 0.01 level, demonstratng the mportance of the emergence of these socopoltcal legtmacy dmensons on the creaton of the ndustry. Also found sgnfcant, negatve ndustry evaluatons were negatvely related to the entry rate, supportng hypothess 3b wth a p-value of 0.05 and provdng strong evdence for the de-legtmatng nfluence of certan forms of ndustry dscourse. 41

43 Comparng densty-dependent and meda coverage models. Drect comparson of the news meda-based measures n models 2 and 3 wth the densty-dependent measures n model 1 s not possble usng a lkelhood rato test snce the models are not nested wthn one another. 10 However, a comparson crteron that has been useful n many lmted dependent varable statstcal models (ncludng Posson and negatve bnomal regresson) s Akake's Informaton Crteron, or AIC, whch allows one to compare two non-nested models that cannot otherwse be compared usng the lkelhood rato test (Cameron and Trved, 1998; Long, 1997). The AIC also penalzes the model wth the larger number of parameters for the degrees of freedom that are lost; because of ths, t can be used to supplement the pseudo R-squared n comparng models. All thngs beng equal, the model wth the lower AIC s the preferred model. Comparng models 2 and 3 wth the densty-dependent effects dentfed n model 1, usng meda-based measures of legtmacy to predct the populaton entry rate compares very favorably wth densty-based measures. Examnng models 1 and 2, the general meda coverage model has a hgher pseudo R-squared and a lower AIC than the densty-dependent model (pseudo R-squared of versus ; AIC of versus ), showng that meda-based ndcators of cogntve legtmacy may ft the data better than densty. In models 1 and 3, the model of socopoltcal legtmacy effects revealed through the volume of news meda coverage compares well to the densty-dependent model wth a hgher pseudo R-squared and nearly dentcal AIC (pseudo R-squared of and AIC of ). 10 More correctly, these models overlap only n the control varables, but the theoretcal varables of nterest do not overlap at all. Models that have no shared ndependent predctor varables are labeled as "strctly nonnested" models (Cameron and Trved, 1998). 42

44 Combnng News Meda-based Legtmacy Effects In A Sngle Model of Entry Rates Model 4 ntegrates both general meda coverage wth the current volume of news reports of ndustry evaluatons and regulatory attenton nto a sngle model. The purpose of dong ths was to buld a meda-based model of entry rates that smultaneously combnes cogntve and socopoltcal legtmacy effects. General ndustry coverage and regulatory nterest varables were shown to have postve and sgnfcant effects at the 0.01 level. Negatve ndustry evaluatons agan affect entry rates negatvely, wth a sgnfcance level of However, postve ndustry coverage s no longer sgnfcant when combned wth general ndustry coverage. Ths may be the result of collnearty between these two measures (the correlaton between the two was 0.85), a problem whch could result n the nsgnfcance of one measure of legtmacy even though both are separately hghly sgnfcant. Agan the AIC s useful here, because t can be used to assess the mprovement n the model combnng two sets of varables over models that consder each on ther own. The pseudo R-squared of model 4 s hgher than that of ether model 2 (usng only general meda coverage) or model 3 (usng postve/negatve and regulatory meda coverage). However, t s well known that these types of ft measures typcally ncrease when the number of predctor varables ncrease; a hgher R-squared alone s not evdence that the combned model s the better one. Comparng the AICs of each model gves addtonal nsght nto whch more approprate because t penalzes aganst a reducton n parsmony n the model. The AIC of model 4 (AIC = ) s lower than ether model 2 or 3, ndcatng that postve ndustry coverage should be ncluded n the model. Snce the statstcal ft of the model mproves even though ths sngle varable becomes nsgnfcant suggests that t plays an mportant but ndrect role n nfluencng the populaton entry rate; n usng multple 43

45 ndcators of legtmacy n the model, the possblty that dfferent dmensons of legtmacy nteract n complex and nterestng ways to affect entry rates becomes evdent. Testng Hypotheses 4 and 5 Combnng Densty-Dependent and Meda-Based Varables n a Model of Entry Rates Combned densty and general meda coverage model. The negatve bnomal regresson models 5, 6, and 7 combne both the densty-dependent and the meda-based effects consdered separately n models 1 through 4, provdng a test for hypotheses 4 and 5. Hypothess 4 stated that both densty-dependent effects and general meda coverage contnue to have the relatonshps wth entry rates descrbed n hypotheses 1 and 2, but that those relatonshps wll reman sgnfcant smultaneously; hypothess 4 s tested n regresson model 5 descrbed n Table 2. The statstcal analyss showed that both frst- and second-order forms of densty are sgnfcant, wth the frst-order measure reflectng the postve legtmatng effects of ncreasng densty n the populaton whle the second-order varable showed negatve compettve effects (although second-order densty was only mldly sgnfcant at the 0.10 level). The parameter estmate for general ndustry coverage s postve and strongly sgnfcant at the 0.01 level. The sgnfcance of densty-based legtmacy and competton as well as cogntve legtmacy effects of general news coverage of the ndustry provde support for ther combned effect on entry rates descrbed n hypothess 4. Furthermore, compared to model 1, the effect szes for the densty varable coeffcents are reduced n sze, ndcatng that general meda coverage now represents some of the varance n the model prevously accounted for by populaton densty. Table 3 shows the result of the tests comparng the models combnng densty-dependent and meda-based legtmacy effects wth the models whch treat ther nfluence separately. Comparng model 5 to model 2 tests the sgnfcance of addng densty varables to a model wth 44

46 general meda coverage. Both the lkelhood rato test and the comparson of AICs show that the densty varables sgnfcantly add to the explanatory power of the model. Comparng model 5 to model 1 tests the sgnfcance of addng the general meda coverage varable to a model wth densty varables. As before, both the lkelhood rato test and the comparson of AICs ndcate that both the news meda-based measure of the cogntve legtmacy of the broadband ndustry and the populaton densty varables should be ncluded n the model Insert Table 3 About Here Combned densty, and postve/negatve and regulatory coverage model. Hypothess 5 takes both the densty-dependent and meda-based effects descrbed n hypotheses 1 and 3, and combnes ther nfluences n a model of market entry rates. Model 6 from Table 2 tests ths hypothess by ncludng both densty-based measures of legtmacy and competton and the three meda-based measures of socopoltcal legtmacy as ndependent varables. Both the denstydependent effects of legtmacy and competton as well as coverage of postve and negatve evaluatons of the state of the ndustry and regulatory nfluence are all hghly sgnfcant, provdng support for hypothess 5. The coeffcents for both postve ndustry evaluatons and regulatory coverage are postve and sgnfcant (p-values of 0.05 and 0.01), ndcatng that both are crtcal n the creaton of the broadband populaton. The parameter estmates of negatve meda coverage are also sgnfcant but negatve n drecton (at the 0.05 sgnfcance level), ndcatng that t remans as a de-legtmatng nfluence. Table 3 shows the results of the lkelhood rato tests and model AICs comparng models 6 and 3. The result shows that addng frst- and second-order forms of densty produces a 45

47 sgnfcantly better model of the entry rate nto the broadband populaton then the use of medabased measures alone. In movng from model 3 to model 6, the effect szes of negatve ndustry evaluatons ncrease whle postve and regulatory meda coverage slghtly decrease. Ths may suggest that the addton of densty varables clarfes a strong and drect nfluence from negatve nformaton on the development of emergng organzatonal populatons, and that some of the effects of the densty varables may also represent the nfluence of regulatory factors and postve ndustry evaluatons. Smlarly, comparng models 6 and 1 n Table 3 shows that addng meda coverage measures to the densty-dependence model produces a more favorable model than usng densty-based varables alone. Combned densty and meda coverage effects model. Model 7 combnes the nfluence of densty wth all meda coverage varables. Most nterestngly, both frst- and second-order densty varables are not sgnfcant when ncluded n a model wth meda coverage varables. In contrast, three of the four meda coverage varables are sgnfcant n the combned model, suggestng that modelng entry rates at the early stages of populaton formaton benefts consderably from the addton of meda coverage varables. General meda coverage s hghly sgnfcant at the 0.01 level, and the drecton of the parameter estmate s postve. As was seen n prevous models, negatve meda coverage of the ndustry s negatvely assocated wth entry nto the populaton at the 0.01 sgnfcance level. The volume of regulatory coverage was also sgnfcant (at the 0.05 level), wth a postve effect on entry rates. As n model 4, the effects of collnearty between general ndustry coverage and the volume of postve evaluatons are seen n model 7, where the combnaton of the two varables results n the nsgnfcance of the postve ndustry evaluaton as an explanatory factor. 46

48 Snce the combnaton of densty and all meda coverage varables n model 7 results n the lack of sgnfcant effects from densty varables, the queston arses as to whether or not densty should be retaned n the entry rate model. Agan the results n Table 3 dsplayng the lkelhood rato tests and model AICs comparng full and constaned models provde an ndcaton of the approprate model. Comparng model 7 to model 4 shows the sgnfcance of addng densty varables to a model wth meda coverage varables. Whle the full model has a lower AIC and a hgher pseudo R-squared than the constraned model (AICs of versus ; pseudo R-squareds of versus ), thus suggestng that the model contanng both densty and meda coverage varables s approprate, the lkelhood rato test shows that densty varables do not add sgnfcantly to the explanatory power of the model wth meda coverage varables alone. Comparng model 7 to model 1 shows the sgnfcance of addng meda coverage varables to a model based on densty varables; both the comparson of AICs and the lkelhood rato test pont to the value of addng meda coverage varables to a model wth only densty varables. By comparng these varous models the benefts of gong beyond denstydependence n analyzng entry rates becomes apparent. Control Varable Models The parameter estmates lsted under model 8 n Table 2 depct all the control varables used n the study, and provde a baselne for the rest of the statstcal models n the study. Among the control varables, only the local resdental populaton and economc envronment varables produced sgnfcant results, and only n models wthout populaton densty-based varables (models 2, 3, 4, and 8), whle growth rate of the sze of the Internet was never sgnfcant n any model. When sgnfcant, the sgns of the parameter estmates were n the expected drectons. Interestngly, the fact that these control varables were only sgnfcant n models wthout frst- or 47

49 second-order densty measures ndcate that both forms of densty medate the relatonshp between demographc and economc condtons and the entry rate. DISCUSSION AND CONCLUSION The densty-dependence model of legtmacy and competton proposed by the populaton ecology tradton n organzaton scence has receved sgnfcant support by many dfferent researchers lookng at a varety of organzatonal populatons. Ther work has provded a sgnfcant contrbuton to the understandng of the mportance of changes n densty on the dynamc processes gong on wthn populatons of organzatons (Hannan and Carroll, 1992; Hannan and Freeman, 1989). The dfferng vews offered by organzatonal ecology and nsttutonal theory perspectves, hghlghted n the ntroducton to ths paper, provded the motvaton for ths study to take more detaled look at the operaton of legtmacy processes n the formaton of new organzatonal populatons. The sgnfcant results of ths study contrbute to ths body of work by focusng on the legtmacy sde of the model, partally supportng the densty-dependence argument whle also demonstratng how meda coverage varables sgnfcantly nfluence entry rates, extendng the understandng of legtmacy and ts nfluence not only on the dynamcs wthn establshed populatons but also on the emergence of new ones. Implcatons for Theory Development n Management and Organzaton Scence The results of ths study have several mplcatons for the feld of management and organzatons, partcularly n the development of the densty-dependence model of legtmacy and competton. Frst, ths study fnds support for the densty-dependence model of legtmacy and competton, and extends ts applcaton to the nascent broadband Internet servce provder ndustry. Second, the study consderably advances the feld by extendng the measurement of 48

50 legtmacy through the use of meda-based measures of taken-for-grantedness and socopoltcal nfluence n the ndustry. Fnally, the study also hghlghted the mportance of ndustry dscourse and nformaton exchange as a factor n the development of legtmacy; the sharng of nformaton among ndustry partcpants and observers bulds greater awareness and understandng of the populaton, exposes a greater number of people to the opportuntes beng presented, and underles the process whch establshes or weakens support from crtcal stakeholders. Applyng the densty-dependence model n emergent populatons. Whle the fndngs from the analyss of the broadband servce provders and ther foundng and entry actvtes clearly supports the densty-dependence model of legtmacy and competton, what dstngushes ths study from others s the focus on the hstorcal perod startng long before the start of the populaton through the begnnng stages of ts formaton; even n ntal stages of populaton creaton the densty-dependent effects of legtmacy and competton are both hghly sgnfcant. Ths s an mportant fndng consderng Baum and Powell s (1995: 533) observaton that stronger support exsts for densty-dependence n the lterature n studes wth populatons that had long hstores; the support for the nverted U-shaped relatonshp between densty and foundng rates may be undermned by the ncluson of populatons wth low densty n both early and late hstorcal perods. The broadband access provder populaton dd not exhbt a late perod of low densty, but densty-dependent effects on entry were stll found to be sgnfcant. Carroll and Hannan (2000: 23) speculated that n "modern" ndustres, the hstorcal pattern of organzatonal densty would be smlar to that found n older ndustres but would occur at an accelerated pace; ther depcton of the hard dsk drve ndustry showed the same pattern of populaton densty found elsewhere but unfoldng over the course of just a few decades, 49

51 comparatvely short for ecology studes. The populaton of partcpants n the broadband access ndustry developed only over the last few years, demonstratng an even qucker pace of development. However, the lack of sgnfcance of densty varables n model 7 from Table 2 suggests that the concepton of densty drven legtmacy processes be reconsdered, especally wthn the context of nascent organzatonal populatons. The legtmatng effect of meda coverage on entry rates. In response to the debate on the measurement of legtmacy and the assessment of ts nfluence (Baum and Powell, 1995; Carroll and Hannan, 1989b; Hannan and Carroll, 1995; Zucker, 1989), ths study went beyond the densty-dependent model of organzatonal ecology by extendng the measurement of legtmacy wth news meda-based data sources, and tested the mpact of those varables on entry rates. The models used allowed an examnaton of the legtmatng nfluence of both densty-dependent processes and nformaton exchange amongst ndustry partcpants, producng a rcher pcture of the processes underlyng populaton entry rates. When used n place of densty varables, models usng meda-based measures reflectng cogntve and socopoltcal legtmacy effects performed as well as the densty-dependent model n predctng entry rates; even when combned together, they retaned ther sgnfcance and produced a much more detaled pcture of the processes occurrng n the broadband ndustry. Moreover, n comparng models 4 and 7 n Table 2, the result of the lkelhood rato test ndcates that densty varables do not add sgnfcantly to the model contanng only meda coverage varables. At least n emergng organzatonal populatons, meda-based legtmacy effects appear to exert a stronger nfluence on foundng rates than densty-dependent processes; early communcatons processes and nformaton dffuson among ndustry partcpants s more mportant than densty growth patterns. In comparng these two perspectves, the study responds drectly to the call by Baum and Powell 50

52 (1995: 530) to examne how dfferent aspects of legtmacy combne to contrbute to the legtmacy of organzatonal populatons. Drect and ndrect nfluences on entry rates. Typcal research strateges n the study of organzatonal populatons look at the drect effects of the key densty-based varables as well as the ndependent control varables approprate for the context of the study; the desred result s a parsmonous model of the densty-dependent effects of legtmacy and competton on the entry and ext rates of the populaton. By combnng multple measures of legtmacy nto populaton entry rate models, t becomes possble to see potental medatng relatonshps that would not be evdent f densty-based measures were used alone; a possble new model mpled by the results of ths study s presented n Fgure Insert Fgure 5 About Here Lookng at the parameter estmates across all the statstcal models lsted n Table 2, t s clear that general, negatve, and regulatory meda coverage have sgnfcant and drect effects on the rate of entry. Increases n the general level of dscourse among ndustry partcpants ncrease the taken-for-grantedness of ndustry actvtes, and thus postvely affect the entry of new market partcpants. Poltcal and regulatory nterests evdent n meda coverage of the ndustry also postvely mpact populaton entry rates, suggestng that governmental attenton lead to greater nterest n and support for the developng broadband ndustry. Increases n negatve evaluatons of the ndustry and ts prospects have the reverse effect; negatve nformaton exchange has a delegtmatng effect on the ndustry, and leads to lower entry rates. 51

53 A medatng relatonshp s mpled between the socopoltcal mpact from postve ndustry coverage and general meda coverage. Agan comparng the regresson estmates from models 3, 4, 6, and 7 lsted n Table 2, t can be seen that the addton of general meda coverage to the models results n the postve ndustry meda coverage becomng nsgnfcant. Earler, t was noted that even wth ths loss of sgnfcance t was approprate to keep the varable n the model. The lack of a sgnfcant drect effect when combned wth general meda coverage suggests that broad coverage of the ndustry medates the relatonshp between postve coverage and entry rates. The mplcaton s that postve nterpretatons of the potental of the ndustry lead to more general dscourse about the technology, ts applcatons n the market, and organzatons created to offer the servce; such dscusson among ndustry actors then leads drectly to market entry. The complexty of these relatonshps ponts out how ntertwned cogntve and socopoltcal dmensons of legtmacy are wth each other and wth growth n the populaton; further research s clearly needed to more fully specfy these relatonshps. Most nterestng here s not the drect and ndrect mpact of the meda-based measures, but the role densty-dependence plays n lght of those meda coverage effects. Examnng models 5, 6, and 7 n Table 2, t s evdent that densty varables are only sgnfcant when general meda coverage and evaluatve meda coverage varables are consdered separately. Whle the ntal analyss n model 1 supported the densty-dependence model, when general and evaluatve meda coverage s consdered together, as t s n model 7, the legtmacy and competton effects of densty dsappear entrely. The mplcaton of ths fndng s that medabased measures of legtmacy may ntervene n the relatonshp between densty effects and foundng and entry rate patterns. These results recall the competng densty-as-process and densty-as-proxy arguments regardng the legtmaton effects of densty on entry rates (Baum 52

54 and Powell, 1995; Baum, 1996). The densty-as-process argument favored by Hannan and Carroll (1992) vews densty as controllng legtmacy processes operatng n organzatonal populatons; legtmacy s a concept that lnks densty to entry rates rather than a characterstc to be measured tself. Addng nondensty-based measures of legtmacy should enhance the sgnfcance of densty-based effects. The densty-as-proxy explanaton portrays densty as nfluencng populaton rates through legtmacy processes; densty s vewed as an ndrect ndcator of a populaton s legtmacy. From ths perspectve, addng addtonal nondensty-based measures of legtmacy should lead to a reducton n densty effects. Snce meda-based varables lead to a reducton n the sgnfcance of densty n models 5 and 6, and the lack of sgnfcance of densty n model 7, these results support the densty-as-proxy argument and show that organzatonal densty s a proxy for cogntve and socopoltcal legtmacy effects n new populatons. Furthermore, demographc and economc control varables were sgnfcant only n the absence of densty-based varables. Demographc and economc varables were ntended to account for changes n market condtons that could nfluence entry beyond the effects of legtmacy and competton. Varables lke the local rate of unemployment and the GDP growth rate provde some ndcaton of dscretonary ncome and access to resources that affect the level of market demand or the avalablty of nvestment funds, whle varables lke the resdent populaton of the cty provde a measure of the potental sze of the market and the capacty of the market to sustan the organzatonal populaton. The fact that these control varables lose ther sgnfcance n models wth densty-based varables suggests that densty medates the relatonshp between market condtons and entry; a populaton s densty mght represent the legtmacy of the populaton, but t mght also represent other underlyng condtons n the 53

55 envronment. If ths s the case, nterpretng the sgnfcance of densty n predctng entry rates becomes dffcult snce t would no longer smply represent the legtmacy of the populaton. Usng only densty may produce a parsmonous model of entry rates n an organzatonal populaton, but t may produce an ncomplete pcture of the mpact of legtmacy on the process snce densty may confound legtmacy wth the nfluence of other economc and socal factors. Legtmacy and nformaton exchange. Whle densty varables are clearly mportant n predctng entry rates, the meda-based measures of legtmacy emphasze nformaton exchange as an mportant element n the constructon of legtmate organzatonal populatons. The mplcaton for research s that the nformaton exchange between dfferent socal actors n and around the ndustry has to be more fully explored. The basc meanngfulness and acceptablty of an new organzatonal form to those n the ndustry, as evdenced by the rse s dscourse regardng that form, s a crtcal factor n supportng foundng events n the populaton. Dscusson among ndustry actors creates a shared dentty and bulds a common understandng of new organzatonal forms, frm strateges, and products and servces; t creates awareness of new organzatons to those who mght support or mtate them. In broadband, before the ndustry could be establshed the technology tself had to be developed, ndustry standards created, and ts usefulness determned; basc facts lke who mght provde broadband servces needed to be resolved (e.g., cable companes, regonal Bell companes, compettve telecommuncatons companes, or some other type of company, just to name a few). Indeed, the very meanng of the term "broadband" had to be settled before new entrants could be started to provde servces based on t. The establshment of a routne understandng and dentty of new organzatonal forms through the ongong conversaton among ndustry partcpants was a necessary step before new entrants could be founded. 54

56 Beyond a fundamental understandng of the organzatons n the populaton, actve support and approval by those n the busness envronment s also a key factor n the development of the organzatonal populaton. Organzatons and those who mght start them exst n a complex socal envronment, consttuted by sets of values of varous stakeholders and embedded n networks of relatonshps wth other organzatons. The meda communcaton studed n ths research project dentfed the postve and negatve nterpretatons of the broadband ndustry by varous nterested partes wthn that ndustry. Increasng amounts of nformaton exchanged descrbng the ndustry postvely lead to an ncrease n populaton entres drected toward captalzng on those opportuntes. As people and organzatons nvolved n the telecommuncatons and networkng ndustres developed broadband technology, a broader recognton of the growth potental of the market for hgh speed Internet access occurred. Broadband technologes were promoted, busness models were created, and nvestment resources flowed nto the ndustry to captalze on the technology and satsfy customer demand. As access provders entered the market, partcpants and observers of the ndustry contnued ther nterpretaton and evaluaton of ndustry actvtes. Negatve evaluaton of the ndustry also strongly affected the ndustry, but n ths case t played a de-legtmatng role, reducng the rate of entry. Even as captal flowed nto the ndustry, some level of concern or cauton regardng the true sze of consumer demand and the growth n the market appeared n meda coverage. Interestngly, ths delegtmatng nfluence occurs even n the earlest moments n the hstory of the ndustry, even before there s an establshed populaton. Regulatory ssues were also the subject of much communcaton n the meda, partcularly commentary on the meanng and mplcatons of potental and actual legslatve actons and regulatory agency rulngs, and t should be no surprse that entry rates ncreased as the meanng of regulatory 55

57 actvtes and the role of the government were developed and the scope of ther nfluence became less ambguous. All these varous facets of the legtmaton process are dependent on the nformaton exchange occurrng among ndustry partcpants. Dfferent dmensons may have dfferent relatonshps wth densty, entry rates, and wth each other. Furthermore, the pattern of dscourse n the ndustry and ts effect on ndustry actvtes may be subject to pecular ways n whch managers and entrepreneurs attend to nformaton and ts use n generatng organzatonal actvtes and entrepreneural efforts. Ths study takes a cautous frst step n the drecton of explorng these factors n organzatonal ecology models. Terreberry (1968) noted that legtmacy s ted to nformaton exchange and communcaton regardng the value of organzatonal practces and the wllngness to nvest resources. The populaton densty provdes ndustry partcpants wth some knowledge about actvtes occurrng wthn the populaton, but changes n densty provde nformaton that may seem only dstantly lnked to many crtcally mportant ssues n an ndustry. As a sngle measure, t may also obscure a number of nformaton exchange processes whch have dfferent effects on entry rates (Delacrox and Rao, 1994). Recent work on the formaton of new organzatonal forms has hghlghted the mportance of developng of a clearly defned and focused dentty wthn a populaton, whch s only partally nfluenced by patterns of organzatonal densty (McKendrck and Carroll, 2001; McKendrck, Jaffee, Carroll, and Khessna, 2003). Also mportant n the establshment of a legtmate dentty for a new populaton s the ongong dalogue and nformaton exchange occurrng wthn the ndustry. The actvtes of organzatons and entrepreneurs were clearly related to the level and content of ndustry dscourse; as an nfluence on the actvtes and practces of organzatons, ths communcaton content s evdence of the extent of both the cogntve and socopoltcal 56

58 legtmacy of the populaton. Examnng nformaton exchange provdes a wndow on the socal processes n whch legtmacy emerges and develops over tme. Aldrch and Fol (1994) noted that legtmacy emerges as an aspect of the socal context of ndustry creaton, provdng both a restrctve nfluence on behavor but also an arena n whch entrepreneurs work toward establshng the legtmacy of new organzatonal forms. Ths study clamed that ths socal context emerges through nformaton exchange and the sharng of conceptons of what broadband access mght provde and who mght provde t, and the constructon of legtmate status of new organzatonal forms for the provson of broadband servce can be seen n analyss of the busness press. Legtmacy and entrepreneurshp. As s evdent from Fgures 2 through 4, a great deal of communcaton actvty occurred wthn the broadband ndustry, about factors rangng from the development of the underlyng technologes to how the government should or should not be nvolved n regulatng the ndustry, even before the frst provder entered the market. The examnaton of communcaton actvty between ndustry partcpants before the appearance of the organzatonal populaton allows for an assessment of the nfluence ntal organzng actvtes have on ndustry creaton. Ths refocuses attenton on acton and nterest n the development of organzatonal populatons; the emergence and growth of populatons can be seen as the outcome of many competng nterests, wth organzatons aggressvely attemptng to nnovate and change ther strateges and technologes whle entrepreneurs buld new organzatons to explot emergent opportuntes. As Aldrch and Fol (Aldrch, 1999; Aldrch and Fol, 1994) ponted out, the creaton of legtmacy s a crucal problem for entrepreneurs engaged n the formaton of a new ndustry. Whle densty-based measures provde key nsghts nto the actvty level of establshed 57

59 populatons, the growth of legtmacy evdent n the nformaton exchange among ndustry partcpants gves mportant clues about populatons stll n the formatve stage of development, where the opportunty may stll be n doubt, the ultmate organzatonal form s unclear, and the best strateges and technologes may by unknown. Aldrch and Fol (1994) outlned a number of strateges that organzatons and entrepreneurs can take to establsh legtmacy. The mplcaton for entrepreneurs and managers s that they take an actve role n the communcaton processes defnng new organzatonal forms and lnkng pre-populaton actvtes to the formaton of the new ndustry. Carefully crafted busness models, coherent explanatons of technologcal benefts, the push toward standards and domnant desgns, and the creaton of ndustry consorta and trade assocatons are all actvtes that can rase the level of legtmacy of the ndustry; the contrbuton of nformaton exchange n creatng the socal context n whch ths occurs requres that ndustry actors drect ther efforts toward accessng the news meda, nfluencng publc opnon and nvestor concerns, ganng the attenton of nfluental opnon leaders, and lobbyng regulatory enttes. Ths work s both an ndvdual and a collectve effort; the burden of establshng legtmacy falls dsproportonately on those nvolved n the ntal moments of ndustry creaton, and those that follow beneft from the work of those that went before them (Delacrox and Rao, 1994). Methodologcal Implcatons In order to assess the nformaton exchange and dscourse nvolved n constructng the legtmacy of the populaton of broadband access provders, ths study employed a content analyss technque for data collecton. The content n news meda coverage has the capacty to both transmt and nfluence the values n a socal group (Alper and Ledy, 1969; Chen and Mendl, 1991; Gerbner, 1969; McCombs and Shaw, 1972; Namenwrth and Weber, 1987). 58

60 Furthermore, the prevalence of detaled archves of news meda coverage on a number of ndustres gve researchers a valuable resource n the exploraton of legtmacy (Baum and Powell, 1995; Hannan et al., 1995). Researchers are just begnnng to use content analyss technques to assess legtmacy (e.g., Coombs, 1991; Deephouse, 1996; Pollock and Rndova, 2003), and greater applcaton n populaton ecology research has the potental to further develop the understandng on how dfferent dmensons of legtmacy nfluence populaton processes. Lmtatons and Drectons for Future Research Whle ths study took a crtcal step n extendng the densty-dependence model of legtmacy and competton by examnng the content of communcaton and nformaton flows among populaton partcpants, much more work n ths area needs to done n order to fully understand how the socal context n whch legtmacy s constructed emerges and develops wth the growth of organzatonal populatons. Frst, whle the meda-based measures of postve and negatve ndustry evaluatons were desgned to assess concerns regardng the economc prospects of the ndustry, dfferences between the nfluence of economc factors versus wdely held perceptons about those economc factors was not dstngushed n ths study. Snce the subject of ths study was an ndustry n ts formatve stages, uncertanty and ambguty about ts future prospects and a lack of a past hstory to provde performance benchmarks caused entrepreneurs and other ndustry actors to base ther decsons on factors other than objectve economc nformaton. Socally-constructed belefs about the ndustry may be more mportant n such stuatons rather than quanttatve measures of ndustry growth. However, future studes could more formally address ths ssue by addng more objectve measures of ndustry growth to entry rate models. 59

61 Second, snce ths study proposed that nformaton exchange provdes the socal context n whch legtmacy s establshed and mantaned, further development of ths research could look at the structure of the mass communcaton process tself on legtmacy. There may be systematc dfferences n how organzatonal actvtes are covered n the meda. For nstance, t s reasonable to assume that large frms receve the most news coverage; as a consequence, they may also have greater nfluence on legtmacy and entry rates even n new ndustres. If true, the mplcaton would be that the creaton of new ndustres may nvolve the formaton of new frms, but may also requre the support of large ncumbent frms to ultmately acqure legtmacy. Bases n communcaton processes may have other consequences for the establshment of legtmacy, such as a focus on hghly successful (or unsuccessful), but unrepresentatve, entrants n the ndustry. Furthermore, dfferent sets of publcatons (e.g., weekly news magaznes wth a wde audence versus ndustry trade publcatons wth a focused consttuency) may reflect the nterests and conceptons of dfferent sets of ndustry actors whch may nfluence the development of new organzatonal forms. Future studes should examne further the dfferences between frms and new meda sources n ther ablty to nfluence the recognton and reputaton of the ndustry. Thrd, the results of ths study may not be generalzable to later perods n the hstory of ndustres. The scope of ths study was the tme perod leadng up to the frst entres nto the market and through the early phases of populaton development. However, the patterns of dscourse may change as the ndustry matures. The ntent of studyng an emergent ndustry s a focus on the mpact of legtmacy over other factors; after the ndustry becomes establshed, the nfluence of densty and other varables on the pattern of entres and foundngs may reflect other 60

62 populaton processes. However, t may be useful to study how the patterns and content of communcaton change over tme wth the maturaton of an organzatonal populaton. Fnally, ths study recognzes the mportance of legtmacy on the rates of entry nto the populaton. However, a major thrust of populaton ecology research s organzatonal survval. The arguments generated n ths study address market entry; a dfferent, albet complementary, set of arguments should be developed for organzatonal falures. Applyng the approaches developed here for the analyss of entry rates to an analyss of market ext rates would be a substantal contrbuton to the ecologcal lterature. Concluson Ths study set out wth the objectve of expandng our understandng of the effects of legtmacy on entry and foundng rates n populatons at a formatve stage of development. Content analyss methods were devsed to evaluate legtmacy effects usng news meda-based data sources, whch were combned wth densty-based measures n a model of entry rates nto the populaton. By usng both densty-dependent and meda-based measures of legtmacy, a more complex pcture of the process of ndustry creaton and development emerged, wth both meda- and densty-based measures sgnfcantly related to the rate of foundng and entry. Furthermore, t was found that meda-based measures of legtmacy produced entry rate models as good as those usng densty-based legtmacy and competton measures; combnng the two types of varables produced substantally better fttng models than those whch treated them separately, and n one nstance resulted n densty becomng nsgnfcant. The use of medabased measures allowed the models to account for pre-populaton actvtes, whch began to establsh the legtmacy of the ndustry before the frst entry appeared and supported the subsequent development of the populaton, whle the use of multple measures of legtmacy 61

63 allowed an examnaton of the complexty of the relatonshps between dfferent dmensons of legtmacy and entry and foundng rates. 62

64 REFERENCES Abrahamson, E Manageral fads and fashons: The dffuson and rejecton of nnovatons. Academy of Management Revew, 16: Aldrch, H. E. and Fol, C. M Fools rush n? The nsttutonal context of ndustry creaton. Academy of Management Revew, 19: Aldrch, H. E.; Zmmer, C. R.; Staber, U. H.; and Beggs, J. J Mnmalsm, mutualsm, and maturty: The evoluton of the Amercan trade assocaton populaton n the 20th century. In J. A. C. Baum and J. Sngh (eds.), Evolutonary Dynamcs of Organzatons: Oxford: Oxford Unversty Press. Aldrch, H. E Organzatons Evolvng. London: Sage. Alper, S. W. and Ledy, T. R The mpact of nformaton transmsson through televson. Publc Opnon Quarterly, 33: Ashforth, B. E. and Gbbs, B. W The double-edge of organzatonal legtmaton. Organzaton Scence, 1: Baum, J. A. C Organzatonal ecology. In S. R. Clegg, C. Hardy, and W. R. Nord (eds.), Handbook of Organzaton Studes: London: Sage. Baum, J. A. C. and Olver, C Insttutonal embeddedness and the dynamcs of organzatonal populatons. Amercan Socologcal Revew, 57: Baum, J. A. C. and Powell, W. W Cultvatng an nsttutonal ecology of organzatons: Comment on Hannan, Carroll, Dundon, and Torres. Amercan Socologcal Revew, 60: Baum, J. A. C. and Sngh, J. V Organzatonal nches and the dynamcs of organzatonal foundng. Organzaton Scence, 5: Berelson, B Content analyss. In G. Lndzey (ed.), Handbook of Socal Psychology, 1 st ed.: Cambrdge, MA: Addson-Wesley. 63

65 Berger, P.r L. and Luckmann, T The Socal Constructon of Realty: A Treatse n the Socology of Knowledge. Garden Cty, NY: Doubleday. Boardwatch Magazne Boardwatch Magazne's Drectory of Internet Servce Provders, 12th ed. Golden, CO: Penton Meda. Budd, R. W., Thorp, R. K., and Donohew, L Content Analyss of Communcatons. New York: Macmllan. Budros, A Analyzng unexpected densty dependence effects on organzatonal brths n New York's lfe nsurance ndustry, Organzaton Scence, 5: Cameron, A. C. and Trved, P. K Econometrc models based on count data: Comparsons and applcatons of some estmators and tests. Journal of Appled Econometrcs, 1: Regresson Analyss of Count Data. Cambrdge: Cambrdge Unversty Press. Carroll, G. R. and Hannan, M. T. 1989a Densty dependence n the evoluton of populatons of newspaper organzatons. Amercan Socologcal Revew, 54: b On usng nsttutonal theory n studyng organzatonal populatons. Amercan Socologcal Revew, 54: The Demography of Corporatons and Industres. Prnceton, NJ: Prnceton Unversty Press. Carroll, G. R. and Swamnathan, A Densty dependent organzatonal evoluton n the Amercan brewng ndustry from 1633 to Acta Socologca, 34: Carroll, G. R. and Wade, J Densty dependence n the organzatonal evoluton of the Amercan brewng ndustry across dfferent levels of analyss. Socal Scence Research, 20: Carroll, G. R.; Presendoerfer, P.; Swamnathan, A.; and Wedenmayer, G Brewery and Brauere: The organzatonal ecology of brewng. Organzaton Studes, 14: CED Magazne CED Indepth: Gettng Connected. Hghlands Ranch, CO: CED Magazne. Chen, C. C. and Mendl, J. R The constructon of leadershp mages n the popular press: The case of Donald Burr and People Express. Admnstratve Scence Quarterly, 36:

66 Coleccha, A Defnng and measurng e-commerce: A status report. OECD Workng Papers, 7(78). Coombs, W. T The falure of the Task Force on Food Assstance: A case study of the role of legtmacy n ssue management. Journal of Publc Relatons Research, 4: Deephouse, D. L Does somorphsm legtmate? Academy of Management Journal, 39: Delacrox, J. and Carroll, G. R Organzatonal foundngs: An ecologcal study of the newspaper ndustres of Argentna and Ireland. Admnstratve Scence Quarterly, 28: Delacrox, J. and Rao, H Externaltes and ecologcal theory: Unbundlng densty dependence. In J. A. C. Baum and J. Sngh (eds.), Evolutonary Dynamcs of Organzatons: Oxford: Oxford Unversty Press. Dempsey, J.; Frsan, G.; Mehrotra, R.; Rao, N. L.; and Whte, A. O The last mle to the Internet. McKnsey Quarterly, 1998 ssue 4: DMaggo, P. J. and Powell, W. W The ron cage revsted: Insttutonal somorphsm and collectve ratonalty n organzatonal felds. Amercan Socologcal Revew, 48: Dowlng, J. and Pfeffer, J Organzatonal legtmacy: Socal values and organzatonal behavor. Pacfc Socologcal Revew, 18: Duncan, D. F Health psychology and clncal psychology: A comparson through content analyss. Psychologcal Reports, 68: Edelman, L. B Legal envronments and organzatonal governance: The expanson of due process n the Amercan workplace. Amercan Journal of Socology, 95: Legal ambguty and symbolc structures: Organzatonal medaton of cvl rghts law. Amercan Journal of Socology, 97: Elsbach, K. D Managng organzatonal legtmacy n the Calforna cattle ndustry: The constructon and effectveness of verbal accounts. Admnstratve Scence Quarterly, 39:

67 Franzos, R Labor unrest n the Italan servce sector: An applcaton of semantc grammars. In C. W. Roberts (ed.), Text Analyss for the Socal Scences: Mahwah, NJ: Lawrence Erlbaum Assocates. Gerbner, G Toward cultural ndcators : The analyss of mass medated publc message systems. In G. Gerbner et al (eds.), The Analyss of Communcaton Content: New York: John Wley and Sons. Gddens, A Central Problems n Socal Theory: Acton, Structure and Contradcton n Socal Analyss. Berkeley, CA: Unversty of Calforna Press The Consttuton of Socety: Outlne of the Theory of Structuraton. Berkeley, CA: Unversty of Calforna Press. Greene, W. H Econometrc Analyss, 3rd ed. Upper Saddle Rver, NJ: Prentce Hall. Hannan, M. T Compettve and nsttutonal processes n organzatonal ecology. In J. Berger, M. Zeldtch Jr., and B. Anderson (eds.), Socologcal Theores n Progress: New Formulatons: Newbury Park, CA: Sage Publcatons Theoretcal and methodologcal ssues n analyss of densty-dependent legtmaton n organzatonal evoluton. Socologcal Methodology, 21: Hannan, M. T. and Carroll, G. R Dynamcs of Organzatonal Populatons: Densty, Legtmaton, and Competton. New York: Oxford Unversty Press Theory buldng and cheap talk about legtmaton: Reply to Baum and Powell. Amercan Socologcal Revew, 60: Hannan, M. T. and Freeman, J The ecology of organzatonal foundng: Amercan labor unons, Amercan Journal of Socology, 92: The ecology of organzatonal mortalty: Amercan labor unons, Amercan Journal of Socology, 94: Organzatonal Ecology. Cambrdge: Harvard Unversty Press. Hannan, M. T.; Carroll, G. R.; Dundon, E. A.; and Torres, J. C Organzatonal evoluton n a multnatonal context: Entres of automoble manufacturers n Belgum, Brtan, France, Germany, and Italy. Amercan Socologcal Revew, 60: Haveman, H. A Follow the leader: Mmetc somorphsm and entry nto new markets. Admnstratve Scence Quarterly, 38:

68 Hlgartner, S. and Bosk, C. L The rse and fall of socal problems: A publc arenas model. Amercan Journal of Socology, 94: Hoffman, A. J. and Ocaso, W Not all events are attended equally: Toward a mddle-range theory of ndustry attenton to external events. Organzaton Scence, 12: Holst, O. R Content analyss. In G. Lndzey and E. Aronson (eds.), Handbook of Socal Psychology, 2 nd ed.: Cambrdge, MA: Addson-Wesley. Hybels, R. C On legtmacy, legtmaton, and organzatons: A crtcal revew and ntegratve theoretcal model. Academy of Management Journal Best Papers Proceedngs, 1995: Kassarjan, H. H Content analyss n consumer research. Journal of Consumer Research, 4: Klennjenhus, J.; Rdder, J. A. de; and Retberg, E. M Reasonng n economc dscourse an applcaton of the network approach to the Dutch press. In C. W. Roberts (ed.), Text Analyss for the Socal Scences: Mahwah, NJ: Lawrence Erlbaum Assocates. Kolbe, R. H. and Burnett, M. S Content-analyss research: An examnaton of applcatons wth drectves for mprovng research relablty and objectvty. Journal of Consumer Research, 18: Krppendorff, K Content Analyss: An Introducton to Its Methodology. Beverly Hlls, CA: Sage Publcatons. Long, J. S. and Freese, J Regresson Models for Categorcal Dependent Varables Usng Stata. College Staton: Stata Press. Long, J. S Regresson Models for Categorcal and Lmted Dependent Varables. Thousand Oaks, CA: Sage Publcatons. McCombs, M. E. and Shaw, D. L The agenda-settng functon of mass meda. Publc Opnon Quarterly, 36:

69 McKendrck, D. G. and Carroll, G. R On the geness of organzatonal forms: Evdence from the market for dsk arrays. Organzaton Scence, 12: McKendrck, D. G.; Jaffee, J.; Carroll, G. R.; and Khessna, O. M In the bud? Dsk array producers as a (possbly) emergent organzatonal form. Admnstratve Scence Quarterly, 48: Meyer, J. W. and Rowan, B Insttutonalzed organzatons: Formal structure as myth and ceremony. Amercan Journal of Socology, 83: Meyer, J. W. and Scott, W. R. (eds.) Organzatonal Envronments: Rtual and Ratonalty. Beverly Hlls, CA: Sage Publcatons. Moschovts, C. J. P.; Poole, H.; Schuyler; T.; and Senft, T. M Hstory of the Internet: A Chronology, 1843 to the Present. Santa Barbara, CA: ABC- CLIO. Namenwrth, J. Z. and Weber, R. P Dynamcs of Culture. Wnchester, MA: Allen and Unwn. Natonal Telecommuncatons and Informaton Admnstraton and the Rural Utltes Servce Advanced Telecommuncatons n Rural Amerca: The Challenge of Brngng Broadband Servce to All Amercans. Washngton, DC: Unted States Department of Commerce and the Unted States Department of Agrculture. Neter, J.; Kutner, M. H.; Wasserman, W.; and Nachtshem, C. J Appled Lnear Regresson Models, 3rd ed. Chcago: Irwn. Olver, C Strategc responses to nsttutonal processes. Academy of Management Revew, 16: Sustanable compettve advantage: Combnng nstutonal and resource-based vews. Strategc Management Journal, 18: Organzaton for Economc Co-operaton and Development Measurng electronc commerce. OECD Workng Papers, 5(97). Parsons, T Structure and Process n Modern Socetes. Glencoe, IL: Free Press. Pollock, T. G. and Rndova, V. P Meda legtmaton effects n the market for ntal publc offerngs. Academy of Management Journal, 46:

70 Publc Utlty Commsson of Texas Report to the 77th Texas Legslature on the Avalablty of Advanced Servces n Rural and Hgh Cost Areas. Austn: Publc Utlty Commsson of Texas. Rao, H The socal constructon of reputaton: Certfcaton contests, legtmaton, and the survval of organzatons n the Amercan automoble ndustry: Strategc Management Journal, 15: Rffe, D.; Lacy, S.; and Fco, F. G Analyzng Meda Messages: Usng Quanttatve Content Analyss n Research. Mahwah, NJ: Lawrence Erlbaum Assocates. Rowan, B Organzatonal structure and the nsttutonal envronment: The case of publc schools. Admnstratve Scence Quarterly, 27: Schneberg, M. and Clemens, E. S. forthcomng The typcal tools for the job: Research strateges n nsttutonal analyss. In W. W. Powell and D. L. Jones (eds.), How Insttutons Change. Chcago: Unversty of Chcago Press. Scott, W. R. and Meyer, J. W The organzaton of socetal sectors. In J. W. Meyer and W. R. Scott (eds.), Organzatonal Envronments: Rtual and Ratonalty. Beverly Hlls, CA: Sage Publcatons. Scott, W. R Insttutons and Organzatons. Thousand Oaks, CA: Sage. Selznck, P TVA and the Grass Roots. Berkeley, CA: Unversty of Calforna Press. 1984/1957 Leadershp n Admnstraton. Berkeley, CA: Unversty of Calforna Press. Sewell, W. H., Jr A theory of structure: Dualty, agency, and transformaton. Amercan Journal of Socology, 98: Sngh, J. V.; Tucker, D. J.; and Menhard, A. G Insttutonal change and ecologcal dynamcs. In W. W. Powell and P. J. DMaggo (eds.), The New Insttutonalsm n Organzatonal Analyss: Chcago: Unversty of Chcago Press. StataCorp Stata Statstcal Software: Release 7.0. College Staton, TX: Stata Corporaton. 69

71 Suchman, M. C Managng legtmacy: Strategc and nsttutonal approaches. Academy of Management Revew, 20: Tarrow, S Democracy and Dsorder: Protest and Poltcs n Italy, Oxford: Clarendon Press. Terreberry, S The evoluton of organzatonal envronments. Admnstratve Scence Quarterly, 12: Tucker, D. J.; Sngh, J.; Menhard, A. G.; and House, R. J Ecologcal and nsttutonal sources of change n organzatonal populatons. In G. R. Carroll (ed.), Ecologcal Models of Organzatons: Cambrdge, MA: Ballnger. Weber, R. P Basc Content Analyss, 2 nd ed. Newbury Park, CA: Sage Publcatons. Whaples, R A quanttatve hstory of the Journal of Economc Hstory and the Clometrc revoluton. The Journal of Economc Hstory, 51: Zald, M. and Denton, P From evangelsm to general servce: The transformaton of the YMCA. Admnstratve Scence Quarterly, 8: Zucker, L. G Combnng nsttutonal theory and populaton ecology: No legtmacy, no hstory. Amercan Socologcal Revew, 54:

72 Fgure 1. Entry rate and densty n sx largest ctes n Texas, January 1993 to Aprl 2000 (by month) (a) Entry rate and densty n Austn (by month) Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan-2000 Densty Entres (b) Entry rate and densty n Dallas (by month) Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan Densty Entres 71

73 Fgure 1 (contnued). Entry rate and densty n sx largest ctes n Texas, January 1993 to Aprl 2000 (by month) (c) Entry rate and densty n El Paso Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan-2000 Densty Entres (d) Entry rate and densty n Fort Worth Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan Densty Entres 72

74 Fgure 1 (contnued). Entry rate and densty n sx largest ctes n Texas, January 1993 to Aprl 2000 (by month) (e) Entry rate and densty n Houston Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan-2000 Densty Entres (f) Entry rate and densty n San Antono Densty Entry Rates Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan Densty Entres 73

75 Fgure 2. Volume of general meda coverage of the ndustry, January 1993 to Aprl 2000 (by month) Monthly Volume of Meda Coverage Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan-2000 Fgure 3. Volume of postve and negatve meda coverage of the ndustry, January 1993 to Aprl 2000 (by month) Monthly Volume of Meda Coverage Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan-2000 Postve Evaluatons of the Broadband Access Industry Negatve Evaluatons of the Broadband Access Industry 74

76 Fgure 4. Volume of governmental and regulatory meda coverage of the ndustry, January 1993 to Aprl 2000 (by month) Monthly Volume of Meda Coverage Jan-1993 Jan-1994 Jan-1995 Jan-1996 Jan-1997 Jan-1998 Jan-1999 Jan

77 Fgure 5. Revsed model of meda coverage-based and densty-dependent effects on entry rates. 76