Founder-CEO Succession and the Paradox of Entrepreneurial Success
Noam Wasserman Harvard Business School, South Hall 310, Boston, Massachusetts 02163 [email protected]
Abstract control the company’s strategy and structure (Woodward In the last few decades, we have developed a substantial body 1965, Lawrence and Lorsch 1967, Thompson 1967). of knowledge about CEO succession. However, except for More specifically, CEOs make material strategic choices some studies of familybusinesses that lack direct applica- that can influence firm performance (Child 1972), and bilityto nonfamilyCEO succession, the past studies of suc- the qualityand performance of an organization’s top cession have not examined the veryfirst succession event managers is often the single most important determi- in a firm, when the Founder-CEO is replaced, on a large- nant of both the success and survival of the organization scale basis. The critical differences between later-stage suc- cession and Founder-CEO succession include the higher level (Drucker 1954). In short, the CEO of an organization of attachment between Founder-CEOs and the firms theycre- is a critical factor in its direction and performance. As ate, the much larger equityholdings of Founder-CEOs (which a result, changes in CEOs—“CEO succession events”— give them much more control of the firm), the fact that many are critical junctures for organizations. Founder-CEOs remain in the firm (even though it is being In the last few decades we have developed a sub- run bytheir successors), and the fact that nearlyall early- stantial bodyof knowledge about CEO succession. For stage succession events involve outside successors (in contrast example, studies have examined the performance conse- to later-stage succession research, which has focused on the quences of CEO succession (e.g., Khurana and Nohria insider-outsider distinction). These differences make it hard to 1999), the institutional/symbolic view of CEO suc- extrapolate from later-stage succession findings to Founder- cession (e.g., Gamson and Scotch 1964, Pfeffer and CEO succession. Therefore, in order to examine Founder-CEO Salancik 1978), and stock market reactions to CEO succession, I used field research and grounded theorybuilding to studythe factors that should affect Founder-CEO succession turnover (e.g., Furtado and Karan 1990). However, in Internet start-ups. I find that there are two central intertem- nearlyall past studies of CEO succession have focused poral events that mayaffect Founder-CEO succession: The on large, public companies. These companies are more completion of product development and the raising of each visible than their smaller brethren, and it is much round of financing from outside investors. I develop testable easier to get data on their histories, executives, and hypotheses about how each of these events affect the rate of succession events. However, even though “research on succession, and then test these hypotheses using an event- the effects of managerial succession has focused on historyanalysisof a unique dataset containing the succession large bureaucracies” (Haveman 1993, p. 864), there are histories of 202 Internet firms. Myfindings point to multiple some marked differences between large bureaucracies “paradoxes of success” in which the Founder-CEO’s success and small companies that prevent us from extrapolating at achieving critical milestones actuallycauses the chance of Founder-CEO succession to rise dramatically. from research findings about large-companysuccession (Entrepreneurship; Founder-CEO Succession; CEOs; Founding Teams; to small-companysuccession. CEO Succession; Top Management Teams; Venture Capital; Entrepreneur- This is particularlytrue for the first succession event ial Finance; Private Companies; Information Technologies) in an organization, where the Founder-CEO is replaced bya “professional” successor. Although succession is especiallycritical “for smaller organizations that are in the process of moving from founder to professional man- Introduction agement” (Kesner and Sebora 1994, p. 363), large-scale Chief executive officers (CEOs) are critical players in studies have not focused on Founder-CEO succession. their organizations. From their perch at the top of a com- The past “early-stage” succession studies were either pany, CEOs are able to direct their companies in the case-level explorations of succession issues (McGivern active pursuit of opportunities (Barnard 1938), and can 1978, Tashakori 1980, Hofer and Charan 1984, Marino
1047-7039/03/1402/0149$05.00 Organization Science © 2003 INFORMS 1526-5455 electronic ISSN Vol. 14, No. 2, March–April 2003, pp. 149–172 NOAM WASSERMAN Founder-CEO Succession and Dollinger 1987) or examinations of small pub- to do so. However, from an agencyperspective (Jensen lic companies, where it is rare to find a CEO who and Meckling 1976), because managers own little of founded the company(Dalton and Kesner 1983). In the the company, their interests can diverge markedly from family-business literature, studies have focused on the the interests of the owners. This makes it necessaryfor intrafamilytransition from one generation to the next owners to take actions that will align the incentives of (Levinson 1971, Dyer 1986, Handler and Kram 1988, the agents with their own. When the interests diverge Lansberg 1988, Handler 1990), but not on the non- too much, one of the major mechanisms that owners can familytransitions that are the core focus of research use to achieve better alignment is replacement of the on CEO succession. It is therefore particularlyprob- CEO with one whose interests are better aligned with lematic to extrapolate from these small-sample, family- the interests of the owners. business, and later-stage studies to overall Founder-CEO However, at the time of companyfounding, founder- succession. executives usuallyown all of the company’s equity, This paper seeks to begin filling this gap. It first which does not fit the Berle and Means model. The outlines some of the most central ways in which founders craft a vision, attract employees, and develop Founder-CEO succession differs from succession in products based on that vision, and perform the man- large companies. It then uses field research and grounded agement tasks necessaryto grow the business. These theorydevelopment to fill in our knowledge of the vari- founders continue to own all of the equityuntil the ables and keyevents that might affect Founder-CEO first time theyaccept outside investors, which often succession, and derives testable hypotheses about those comes manymonths after the companywas founded. antecedents and events. Finally, it uses a unique dataset Even in more-mature entrepreneurial firms, the Founder- containing the succession histories of 202 Internet start- CEO often still owns a large percentage of the company ups to test the hypotheses using an event-history analy- (Wasserman 2001), there is much greater inside owner- sis, and explores some of the limitations of the findings ship than in large firms, and the separation of owner- and the implications for future Founder-CEO succession ship and control that is typically the concern of agency research. theorists does not exist to the same extent. This makes it much harder to argue that owner-manager interests have diverged markedlydue to differences in their own- Founder-CEOs vs. Past ership of the company. Therefore, the core situational Succession Research characteristic of Berle and Means’ (1932) large compa- Given the growing importance of entrepreneurial firms nies does not applyto the small-companysettings of in the world economy(Sahlman et al. 1999), the fact Founder-CEOs. that small organizations employa large proportion of Second, as individuals, Founder-CEOs can be all workers (Granovetter 1984), and the importance of markedlydifferent from later-stage, “professional” founders for the growth of their companies (Schein CEOs. In contrast to people who join an organization 1985), it is important for us to gain a better under- after its founding, the identityof organizational founders standing of the factors that drive Founder-CEO succes- is “tightlylinked” to that of the organization (Dobrev sion in entrepreneurial firms. However, given the stark and Barnett 1999). The same mayhold for the “psy- differences between succession in large firms and in chological bonds” that link individuals to their orga- entrepreneurial firms, there is ample reason to believe nizations (O’Reillyand Chatman 1986). In addition, that Founder-CEO succession will differ from large-firm a large-scale studyof entrepreneurial compensation in succession, making it difficult to extrapolate from the the Internet industry(the same industryexamined in existing literature on succession. In particular, large- this study) found that Founder-CEOs differ from profes- companystudies neglect at least three critical aspects of sional CEOs in several other ways (Wasserman 2001). Founder-CEO succession. Professional CEOs are older, have more years of prior First, large-companystudies often take as their start- work experience, are paid higher salaries but own dra- ing point Berle and Means’ (1932) concept of the sepa- maticallyless of the company’s equity,and are locked ration of ownership and control. In the large companies into longer vesting schedules (which control how long in which CEO succession has been studied, owners and it takes for the CEO’s equityto become effective, and managers are, indeed, two verydifferent groups. This therebyindicate that the investors seek to ensure that the is necessarybecause the owners often do not have the professional CEO remains at the companyfor a longer skills, specific knowledge, or time necessaryto man- period of time). Furthermore, because theyhave much age the company, forcing them to hire “agent” managers larger equitystakes than do public-companyexecutives,
150 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession
Founder-CEOs (and anycofounders) have a greater per- 1985), especiallywhere the new executive came from centage of seats on the board of directors, giving them outside the firm and the former CEO departed. Larger much more potential control of the succession event organizations experience succession events more fre- than large-companyCEOs with smaller equityhold- quentlythan do smaller firms (Grusky1961) because ings would have. Reinforcing this is the fact that the their bureaucracies are more capable of handling the “revered founder” of a companycan have an extraor- disruptive effects of a succession event, and because dinaryinfluence on the keydecisions made within the succession is one of the means bywhich an organi- company, even outside of any other instrumental influ- zation can adapt to a changing environment. Similarly, ences (Zaleznik and Kets de Vries 1975). from a Weberian perspective, while founders relyon Third, according to a recent review paper (Kesner and charisma to build their companies, organizations have Sebora 1994), the most central distinction drawn in past become a lot more routinized bythe time a professional succession work has been that of successor origin: Did CEO joins the firm (Burton and Khurana 2000). This the successor CEO come from outside or inside the orga- formalization constrains individuals’ actions, imposing nization? Outside successors have been found to make discipline such that those actions remain consistent with more changes, to be more highlycompensated, and to organizational goals and purposes (Weber 1946 trans.). achieve higher interorganizational status than inside suc- Furthermore, “sensemaking has freer rein” in younger cessors (Grusky1963). In more competitive industries, organizations where more innovative, nonroutine deci- successors tend to come from inside their firms and to sions are made, affecting how critical high-level deci- remain in the CEO position for longer than in less- sions are made (Weick 1969). A classical example of competitive industries (Pfeffer and Leblebici 1973). In such high-level decisions in younger organizations is the larger companies, boards of directors are usuallyreluc- issue of whether to replace the founding CEO. tant to appoint an outside CEO (Mace 1971, Dalton Past research has also indicated that Founder-CEO and Kesner 1984) unless the companyhas experienced succession maybe the most critical succession event in serious underperformance (Khurana and Nohria 1999). the life of most firms: “After the starting difficulties have In addition, firms are more likelyto change strategy been overcome, the most likelycauses of business fail- when theychoose successors from outside the com- ure are the problems encountered in the transition from pany(Wiersema 1992), making outside CEO succes- a one-person, entrepreneurial style of management to a sion a wayfor boards to effect strategic changes in functionallyorganized, professional management team” large companies. However, as indicated bymyfield (Hofer and Charan 1984, p. 2), and the departure of research and confirmed byanalysisof the data collected a founder has an disproportionate negative impact on for this paper, the successors to Founder-CEOs almost the likelihood of organizational survival (Carroll 1984). always come from outside the firm, rendering moot the A final reason for studying Founder-CEO succession core insider-outsider distinction examined in past suc- is its potential for helping us enrich our alreadydeep cession research. This supports past speculation that understanding of CEO succession in general. Small firms while large firms usuallyturn to inside successors, small are usuallymore focused than are their larger brethren, firms, which have fewer candidates, and young firms, compete in fewer industries, and have simpler resource whose executives have accumulated less firm-specific endowments. Therefore, theyprovide “a more controlled experience, more often turn to outsiders (Helmich 1977, setting to study[succession’s] antecedents and effects” Reinganum 1985). As I show below, instead of outsider- than do the larger, more established firms of past CEO versus-insider, there are other critical distinctions— such succession research (Fiet et al. 1997, p. 364). as whether inside executives have equitycontrol, or For all of these reasons, it is important for us to whether critical intertemporal events are occurring—that gain a better understanding of what occurs within the maybe much more salient for Founder-CEO succession. “black box” of Founder-CEO succession. However, a In short, there are significant reasons whyFounder- dearth of data on private companies introduces chal- CEO succession should differ from later-stage succes- lenges for studying Founder-CEO succession, and is sion events. Although theyhave not explicitlyaddressed probablyone of the main factors in our lack of knowl- the issue, several other studies have hinted that Founder- edge about it. Given that nothing systematic has been CEO succession in small firms should follow rules dif- written about the process or determinants of Founder- ferentlyfrom those followed in large firms. For instance, CEO succession, I used a two-stage research design to although succession has no impact on the stock market investigate Founder-CEO succession. First, I conducted performance of larger companies, it has a positive impact field research, consisting of separate case studies of each on the performance of small companies (Reinganum of 20 private Internet firms, to perform grounded theory
Organization Science/Vol. 14, No. 2, March–April 2003 151 NOAM WASSERMAN Founder-CEO Succession building (Glaser and Strauss 1999). Theoretical sam- require the companyto hire a new CEO soon after the round pling and comparative analysis of these firms enabled closed. Although theyknew that this could drive the company me to gain an understanding of the Founder-CEO suc- founder to look elsewhere for funding, the VC firm believed cession process and of the critical events that might that it was better to raise the issue at the outset of the discus- sions. This would enable the VC to assess, prior to investing, affect this process. I then used myfindings to craft whether the founder reallywas open to the need for a change, hypotheses about the intertemporal factors that should and to avoid putting the companythrough an unpleasant fight affect Founder-CEO succession. Second, I collected a sometime in the future. unique dataset containing the histories of 202 Internet An executive who was working for the Founder-CEO at companies and then used event-historytechniques to test the time said that as the discussions with the VC continued, myhypotheseson the dataset. In the sections below, I the founder began talking about “doing as well as I can from describe each of these elements of the studyand discuss an equityperspective [and] what will be required for the companyto be long-run successful.” Theysoon hired an expe- its results. rienced CEO from a related industry, and the Founder-CEO’s relationship with the board changed dramatically. He felt like he “had disappeared.” The VCs did not call him anymore Hypotheses when theywanted to discuss critical companyissues. Instead, In this section, I describe the hypotheses that I derived theynow looked to the successor CEO for leadership on such from myfield research. In order to generate hypotheses issues. Over the next few months, he realized that, “Less and using grounded theory-building techniques (Glaser and less do I get referred to for myindustrybackground and knowl- Strauss 1999), I reviewed companydocuments and con- edge. In manyways,I’m not crucial to the company.” Instead, as product development was completed successfullyand the ducted interviews with 31 founders, professional CEOs, company’s sales began to rise, his role became that of the executives, and investors in 20 private Internet firms. I “external face of the company,” promoting the company’s work selected these companies based on theoretical-sampling through published articles and conference speeches, and by considerations (Glaser and Strauss 1999), with the goal helping plot the company’s long-term strategy. of illuminating the full range of core factors that might Echoing several other former Founder-CEOs I spoke to, he affect Founder-CEO succession on a wider scale. Toward said that, “We’re trying hard to find a role for me that keeps me operational and active. When and if I don’t have much this aim, I interviewed Founder-CEOs who were still the value to give, I might decide to move onto something else. I’m CEO, others who had been replaced and had left the sure that will come to pass eventually.” companies theyhad founded, and Founder-CEOs who had remained with their companies after leaving the From myfield research, I have found that there are CEO position. Theyincluded both experienced Founder- two critical events in a company’s development that CEOs and young, inexperienced ones. Some of their mayaffect Founder-CEO succession: the completion of product development and the raising of a new round companies had been quite mature at the time of suc- of financing. As these “inflection points” occur, the cession, while other companies were quite young at companychanges in several important waysthat may that point. In addition to these interviews, I also drew affect Founder-CEO succession. These changes include upon myexperiences from working for three months the achievement of milestones that both show if the inside a Boston-based venture capital firm as an asso- Founder-CEO has the abilityto lead the company,and ciate, assessing potential investments, negotiating with change the contingencies faced bythe firm. In addi- entrepreneurs, and participating in firmwide delibera- tion, these events change the company’s ownership and tions about the Founder-CEOs with whom we dealt. decision-making structures, and the amount of resources The following case vignette of one of the Founder- available to develop the company. The sections below CEO succession events I studied includes most of the describe how these events mayaffect Founder-CEO suc- keyissues that will be explored in the sections below, cession, and derive hypotheses that will be tested later and will laythe groundwork for deriving myhypotheses. in this paper. Where appropriate, I also reflect on how At an Internet companyfocusing on the health care arena, the the hypotheses either corroborate or conflict with general plan was for the M.D. who had originallydeveloped the busi- findings from the organizational and CEO-succession lit- ness concept “to be the CEO until the job outgrew me.” This eratures. In addition, where I found contradictoryevi- desire was reinforced bythe experiences of a friend of his, dence, I craft competing hypotheses that I test in an an experienced businessman who had founded a companyand event-historyanalysisat the end of this paper. recentlycontinued as its CEO until after the company’s suc- cessful initial public offering. However, the first venture capi- talist who showed an interest in the companytold him that he Shippingthe First Product was “fine to start the companyand get it going,” but that as a A high level of uncertaintymarks the earlystages of condition of his firm investing in the company, the VC would a company’s life. One of the critical milestones in
152 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession the growth of a technology-based company is finishing typically on the technical issues involved in developing development of the initial product, which helps resolve the company’s initial product or service (Greiner 1972). much of the technical uncertaintyregarding the feasibil- Leading the organization requires a relativelynarrow ityof developing the product. Even more important, in range of skills, for the tasks that must be accomplished almost all of the companies I studied, the completion of are predominantlyin just one or two areas (e.g., research product development was closelylinked to the shipping and development), making it less critical for top execu- of the first product to customers.1 Therefore, reaching tives to have a broad range of skills. this milestone enabled each companyto prove the prod- In contrast, according to executives and investors I uct’s viabilitywith customers and to begin generating interviewed, once a companybegins selling its initial revenues. product or service, the range of tasks required to be suc- Given this, successfullyreaching this milestone is an cessful expands dramatically. Now, the company must early, concrete indicator about the success and future worryabout marketing the product, building a sales prospects of the company. Many Founder-CEOs I stud- force, supporting the product, and managing the complex ied believe that bysuccessfullyleading their companies finances that come with the onset of incoming revenues. through the product development phase, theyshould be The complexitythat must be managed—and therefore assured of retaining the CEO position. As one Founder- the demands of the position—change dramaticallyat CEO stated, “If I’ve gotten us to that stage, man—that this point. Similarly, organizations that are growing usu- should reallysayto them that I can lead this company allyreach a stage in their evolution where a centralized for a while.” Theyargue that investors have few tangible decision-making approach and the level of complexity signs of how well the companyis performing, and that require a change in leadership (Greiner 1972). The com- the completion and shipping of the first product is the pany’s critical contingencies (Salancik and Pfeffer 1977) first real sign that the Founder-CEO is capable of lead- have shifted radically, from technical contingencies to ing the companyto future successes. Furthermore, they marketing and sales ones, requiring verydifferent skills believe that theyare uniquelyqualified to lead the com- from the company’s CEO. pany. As one young founder stated, “I’m the one with As a result, it maybe the case that the best-performing the vision and the desire to build a great company—I companies mayhave to replace the Founder-CEO sooner have to be the one to run it. The people here were my than worse-performing companies. As a venture capital- hires, and the vision was mine from the beginning.” ist stated: A founding CEO who leads his companythrough this milestone should indeed be able to make a stronger argu- The toughest time to change CEOs is when the CEO has been reallysuccessful at developing the company.But those fast ment for retaining his position, for he has proven his growth companies outstrip the CEO’s skills the fastest, and skills in an area critical to companysuccess. This is that’s when we have to push the hardest for a change. With supported bythe findings in large-companysuccession a good “story,” it is best to add a professional CEO before the studies, which consistentlyshow that the rate of succes- scale of operations might logicallyjustifyit. You must stay sion is much lower when the CEO’s companyis per- ahead of the curve to drive momentum. forming well or is meeting board expectations than when Along these lines, I pose the following hypothesis that firm performance is low or falls short of board expec- competes with Hypothesis 1A: tations (Grusky1963, Allen et al. 1979, Jensen 1986, Puffer and Weintrop 1991, Boecker 1992, Cannella and Hypothesis 1b. The rate of Founder-CEO change Lubatkin 1993, Useem 1993). Therefore, I pose the fol- will increase after the company finishes development of lowing hypothesis: the initial product. Hypothesis 1a. After the Founder-CEO has success- Raisinga New Round of Financing fully led his or her company through the successful The second critical event is the raising of new rounds of completion of product development, the rate of Founder- financing. Because start-ups usuallylack the resources CEO succession will decrease. to invest in product development, hire employees, and However, myfield research indicates that in develop keyfacets of their business models, theyusually entrepreneurial firms, the opposite maybe true: Success- have to relyon outside investors for capital. In early- fully completing product development mayactually stage technologycompanies, these outside investors increase the chances that the founding-CEO will be predominantlyinclude venture capitalists (professional replaced. This is because the company’s needs shift dra- private-equityinvestors investing on behalf of their lim- matically. Early in a company’s history, the focus is ited partners), but theyoften also include angel investors
Organization Science/Vol. 14, No. 2, March–April 2003 153 NOAM WASSERMAN Founder-CEO Succession
(wealthyindividual investors) and corporate investors, that it needed “an experienced executive, experienced who usuallyinvest for strategic reasons (Robinson and at managing other executives and a major development Osnabrugge 2000). From the outside investor’s perspec- project.” Along with two other experienced executives tive, the decision about whether to invest in a particular within the companywho agreed that there was a need to start-up is fraught with a high level of uncertainty. Early- change CEOs (“we’re in the big leagues now, we need stage companies rarelyhave substantial tangible assets someone serious”), the VC put pressure on the founder that the investors can assess (Gompers and Lerner 1999, to step aside and refused to invest unless he did so. Sahlman et al. 1999). Therefore, potential investors seek The building cash crunch helped increase the willingness to reduce uncertaintybefore investing byassessing the of the founder to step up to the “honorific” chairman’s skills of the executive team (in particular, those of the role in the company, opening the way for a professional founding CEO), the qualityof the business idea and CEO to take over. Once the founder had “assured the business model, and the “trajectoryof the market” that VCs he would get out of the wayand let profession- the companyis targeting. Opening up the firm to the als run the company,” the VCs completed the investment “due diligence” questioning of potential investors usu- that enabled the companyto develop a verysuccess- allysubjects companyplans and personnel to much ful second generation of its product. According to the more intense and impartial scrutinythan theyhave professional CEO who succeeded the founder, within received in the past (Wasserman 1999). As a result of three months it was alreadyapparent that the founder’s the findings from these due diligence inquiries, outside “operational-involvement days were long past,” though investors often require the company’s management team the company’s early investors still looked to him for to address perceived weaknesses. Sometimes, the list of advice on major business issues. In this and several other weaknesses includes the Founder-CEO himself or her- cases I examined, the need to raise a new round of self, in which case the investors maychoose to push for financing helped force a change in CEOs. a change in CEO. In contrast to the family-firm litera- These cases are consistent with the resource- ture (e.g., Handler and Kram 1988), which has reached dependencyliterature. From a resource-dependencyper- “the general conclusion that the need for the initial suc- spective (Pfeffer and Salancik 1978), a critical challenge cession is often ignored and seldom planned” (Rubenson for entrepreneurial firms is the raising of capital with and Gupta 1996, p. 26), Founder-CEOs in companies which to grow the business. Without capital, the firms with outside investors must deal with parties that push cannot invest in product development, hire the execu- for succession to be an explicit topic of discussion, as tives and people necessaryto build the business, and described further below. Because preinvestment research always leaves many create the infrastructure to be a successful company. In questions unanswered, investors also tryto reduce the addition, high-profile investors bring to their companies uncertaintyover time by“staging” their capital invest- critical nonfinancial resources such as credibility, exper- ments across multiple, smaller “rounds” of investment tise, and contacts, enabling them to attract additional that are months or years apart (Gompers 1995). Start-ups high-qualityemployees,partners, and other investors therefore typically have to go through multiple rounds of (Sahlman et al. 1999). Therefore, when firms become negotiating with their investors. In each of these rounds, desperate for capital, theyare more willing to accede to the investors can make demands that affect the leader- the demands of outside resource providers (Pfeffer and ship of the company. Each round of investment is tied Salancik 1978). to the expected achievement of keymilestones in the However, the cases above conflict with a different company’s development. As they get more information part of the literature. The existing literature suggests about the Founder-CEO’s abilities, the market, and the that companyfounders usuallyhave a “honeymoon company’s prospects, the investors are able to make bet- period” (Choi and Shanley2000) during which new ter decisions about the companyand the role of the stakeholders—such as investors—assess the company’s Founder-CEO. performance against their expectations. During the initial For instance, in an Internet telephonycompanyI stud- “exploration” period (March 1991), stakeholders gain ied, the young founder had already raised an initial information about the founder’s abilities and the com- round of funding from angel investors. However, the pany’s prospects and begin to judge the company’s companywas now getting low on cash, and needed future viability. In contrast, my field research indicates a large capital infusion in order to build a second- that outside investors, usuallyVCs, often tryto force a generation system that would be able to handle high CEO change much earlier in the process. A VC stated volumes of users. A potential VC told the company that, “Our default assumption when we first look at a
154 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession companyis that the Founder-CEO can’t lead this com- round, or a little bit of capital (Wasserman 1999). By panygoing forward,” and that theytherefore push for raising a lot of capital (“doing a big round”), a Founder- an immediate change in CEOs. This is more in line CEO can get the resources to increase the company’s with other research that found that CEOs do not have rate of growth, as one Founder-CEO stated: “The pri- the luxuryof a honeymoonperiod: “Instead of facing a maryreason to raise a lot of capital is to grow the honeymoon period, a newly appointed CEO begins with business more quickly, both by organic growth and by a period of extreme vulnerability.” (Fredrickson et al. acquisition.” However, as described further below, CEOs 1988, p. 258). As a result, Fredrickson et al. found have to balance a desire to do a big round with the need that a disproportionate number of CEOs have veryshort to give up more of the company’s equity and control in tenures. Similarly, in the early-stage companies that are exchange for the additional capital. the focus of this study, VCs know that they will have the In one company, a Founder-CEO wanted to raise $5M most leverage to effect a change when the companyis in his first round of financing in order to increase the rate coming to them for capital in a new financing round, and at which his companycould grow. In his negotiations therefore push for a change immediatelywithout waiting with potential investors, he found that, as a condition for a “honeymoon” period to pass.2 The VC will onlybe of investing so much moneyin the company,investors willing to provide capital to the companyonce the VC would insist on bringing in a professional CEO once is satisfied that the necessarychanges will be made soon the round was completed. However, he said, “I was not after the round of financing is complete. (In such cases, going to hand the companyover to someone else before I VCs often include an explicit CEO-succession clause in had to.” Therefore, instead of the higher figure, he raised the term sheets theyproffer to companies before they $2M during the first round, and therebyavoided having invest.) For example, a general partner at a large Boston- to give up his CEO position until the second round of area venture capital firm explained: financing. In two other companies, the Founder-CEOs made the Upfront, I ask founders to level with me. If theyare interested in working with me on the basis of being a big shareholder, opposite decision, preferring to raise a large amount then I am interested. If theyare interested in working with me of moneyall at once “at the expense of” giving up because theyhave to run the company,then it’s probablynot their positions as CEOs. One reason theygave was that going to make sense for us to work together. theydisliked raising capital from investors, and therefore In a similar vein, board beliefs and assumptions are one wanted to avoid the “pain” of having to go out to raise a new round sooner than theyhad to. Another reason of the most critical sociopolitical factors in CEO dis- was that theybelieved that their companywould miss missal (Fredrickson et al. 1988). If the VCs, who will a “fleeting window of opportunity” if they did not have be serving as board members once their investment has the resources to make keyinvestments or acquisitions been finalized, believe that Founder-CEOs often have soon, motivating them to raise all of the cash necessary to be replaced immediately, they raise the issue at an to do so. However, theyacknowledged that this made earlypoint in time when theyhave the most bargain- them more susceptible to demands for change from their ing power: when the Founder-CEO’s companyis raising investors. Similarly, from a resource-dependency per- moneyfrom the VCs. Therefore, I pose the following spective (Pfeffer and Salancik 1978), CEOs who want hypothesis: to raise a larger round put themselves more at the Hypothesis 2. The rate of Founder-CEO change will mercyof capital providers than do CEOs who raise increase after each round of outside financing. smaller rounds, increasing their susceptibilityto investor demands for a change in CEOs. RaisingBigVersus Small Rounds From the investor’s perspective, when a company However, rounds of financing are not created equal. raises a small amount of capital, the investor knows that For instance, while companies often choose to raise a the companywill have to raise additional capital sooner large amount of moneyin some rounds, other rounds than if theyraised a larger amount of capital. There- might be a lot smaller. In addition, in some rounds, only fore, there will be a shorter amount of time before the the company’s existing investors participate, while other investor will be able to once again make demands that rounds include some new investors. The next two sets of are backed up bythe power to withhold capital. As one hypotheses examine the impact that each of these factors experienced West Coast investor told me, “If I can keep could have on Founder-CEO succession. [Founder-CEOs] on a tighter leash byputting less in now When theyare raising new rounds of financing, com- and having them come back for more soon, I might not panies can choose to raise a lot of capital during the push as hard for changes now.”
Organization Science/Vol. 14, No. 2, March–April 2003 155 NOAM WASSERMAN Founder-CEO Succession
Hypothesis 3. Controlling for the number of previ- 1987) in which VCs invite other VC firms with whom ous rounds of financing, the more money raised by a theyhave developed a relationship to invest in their best company in its latest round of financing, the higher the companies, in exchange for future reciprocated invest- rate of Founder-CEO succession. ing opportunities. To build good syndication relation- ships, VCs usuallytryto avoid dragging other VCs Bringing in New Investors into their “lemons” (Gompers and Lerner 1999). There- Another wayin which financing rounds differ is in fore, firms in whom new VCs are investing maybe the mix of existing versus new investors. Each fund- more likelyto have excellent Founder-CEOs than firms ing round, a companycan either raise moneyfrom its into which VCs do not invite their syndication part- existing investors or it can seek to add new investors. ners. In addition, there often are information asymme- It is often easier to raise moneyfrom existing investors, tries between the new VCs and the existing “insider” VC for theyare alreadyfamiliar with the company,have investors, who are more familiar with the firm (Jensen relationships with the existing executives, have a strong and Meckling 1976). Because theyknow less, new VCs interest in seeing the companysucceed, and maynot maybe less able to trigger a reexamination of whether wish to have their equitypositions diluted (Gompers and to replace the Founder-CEO. Furthermore, new investors Lerner 1999). However, it is often necessaryto raise in each round (and especiallythe “lead” investor) are moneyfrom new investors in subsequent rounds, for as a often added to the board of directors once the round companygrows and its capital requirements increase, the is closed (Wasserman 2002), enlarging the board. Past company’s existing investors may be unable (or unwill- research has shown that large boards usuallycontain ing) to provide for all of its capital needs. a more diverse range of perspectives, which makes it For Founder-CEOs, a benefit of onlyhaving exist- harder to reach a consensus about major issues, such as ing investors involved is that when executives can form whether to replace a CEO (Goodstein et al. 1994, Fiet social ties with outside board members, theygain added et al. 1997). Adding new investors also spreads control abilityto influence decisions about their future (Main of the companyacross more people, further preventing et al. 1995). This mayhelp Founder-CEOs stayin their anyone investor from having disproportionate control positions as long as new board members or investors do of the Founder-CEO succession decision. These reasons not enter the picture. In contrast, the involvement of new suggest that the rate of Founder-CEO succession may investors can be detrimental to the Founder-CEO’s abil- fall after a round that includes new investors. Therefore, ityto retain his position. These new investors, who have I also pose the following hypothesis that competes with not developed the same social ties with the Founder- Hypothesis 4A: CEO, mayhave less confidence in his or her abilityto Hypothesis 4b. When a company adds new investors, continue to run the companyand maytherefore cause there will be a lower chance that the Founder-CEO will the board to reexamine whether to replace the founder. be replaced than when it does not add new investors in In addition, new investors can add credibilityto the a new round of financing. companyand bring new skills and networks of contacts that the companycan leverage (Bygrave 1987). Founder- Changing Ownership Structure CEOs are often willing to accede to the demands of such As mentioned above, each new round of financing also high-qualityinvestors to gain their involvement with the changes the percentage of the companythat is owned by company. Referring to one of the oldest and most pres- outside investors. The more capital raised in each round, tigious of the Silicon Valleyventure capital firms, one the higher the percentage of the companythat the inside Founder-CEO stated, “I’d step down in a minute if that’s executives have to give up to outside investors, which what would get them to invest in us.” is the keytrade-off in a Founder-CEO’s quest to raise a Hypothesis 4a. When a company adds new in- large amount of capital. The likelihood of CEO dismissal depends on the vestors, there will be a higher chance that the Founder- power of the incumbent CEO (Fredrickson et al. 1988), CEO will be replaced than when it does not add new and a critical factor in that power is the amount of equity investors in a new round of financing. held bythe top executives of the company(Rubenson On the other hand, the other cases I studied indicate and Gupta 1992). Conversely, the higher the percentage that existing investors often will not agree to bring new owned byoutside investors, the more decision-making investors into a companyunless theyhave a lot of con- control the outside investors should have about impor- fidence in its prospects and its management team. Much tant issues (Pfeffer and Salancik 1978). For critical deci- of VC investing occurs within “syndicates” (Bygrave sions, such as whether to replace a Founder-CEO, the
156 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession keyturning point is whether outside investors own more Hypothesis 5. In insider-controlled firms, the rate of than 50% of the companyand therefore have “equity Founder-CEO succession will be lower than in outsider- control.” One venture capital firm I studied that invests controlled firms. large amounts in early-stage companies usually insists on having at least 50% ownership after the first round Alternative Hypotheses: Individual, Company, and of financing (“the money buys half the company”) in External Factors order to have more control over each company’s critical The hypotheses above focus on the completion of prod- decisions. However, according to an executive consultant uct development and the raising of each new round who specializes in entrepreneurial firms, “You usually of financing, which seem to be the two most impor- have to go two or three rounds before the executives tant intertemporal events that might affect Founder- have less than 50% of the equity. Once you go south CEO succession. However, past succession work has of having equitycontrol, the moneyinterests can band also examined succession contingencies at the level of together and force them out.” individual-CEO characteristics (e.g., Kotin and Sharaf Similar indications of the importance of equitycon- 1967) and organizational characteristics (e.g., Gouldner trol come from studies of familybusinesses, where gen- 1954). To accuratelytest the hypothesesdeveloped erational transitions within a business are not complete above, I included these factors as control variables in my until the voting stock is passed down as well (Handler event-historyanalyses. 1990). In a striking example of how important equity Individual Characteristics. Individual-level character- control can be in nonfamilybusinesses, too, in one fast- istics that might affect Founder-CEO succession include growing start-up I studied, the young founder brought the Founder-CEO’s years and breadth of prior work in a CEO who had been the head of a major public experience, the existence of “ally” cofounders, the software company. The founder retained majority own- Founder-CEO’s own percentage of equityowned, and ership of the companyand became an outside direc- whether the Founder-CEO left the CEO position volun- tor. However, within two months, the outsider-founder tarilyor involuntarily. was disillusioned with the professional CEO’s “ability First, a powerful recurring finding from studies of to execute our vision” in spite of the professional-CEO’s later-stage CEO succession is the fact that CEOs with “tremendous blue-chip credentials.” Three months later, more years of work experience are able to hold onto their with the backing of the rest of the board of directors, the positions longer (Kesner and Sebora 1994). According outsider-founder fired the high-profile professional CEO to human capital theory(e.g., Becker 1964, Carroll and and began a search for a CEO who better matched what Mosakowski 1987, Dobrev and Barnett 1999), direct the founder believed the companyneeded. Had the pro- work experience provides skills that are not easily fessional CEO had a large equitystake in the company, learned byother means. When these skills and actions the outsider-founder’s abilityto oust the CEO “would help address the most critical problems faced bythe have been severelyhampered.” firm, the executive gains power (Kanter 1977). The ben- Reinforcing this point, past research has found that in efits of previous work experience maybe particularly externallycontrolled firms, outside investors use selec- salient for entrepreneurs, especiallywhen the experience tion and retention to control CEOs (Salancik and Pfeffer is directlyapplicable to the building of youngfirms. For 1980). In owner-managed firms, where the CEO and the example, among early-stage companies around Boston’s rest of the top-management team own controlling stakes Route 128, experienced executives were more effec- in the firm, the tenures of top managers can be more than tive at conveying a sense of confidence to potential twice as long as the tenures of top managers in exter- partners and resource providers, therebyfinding it eas- nallycontrolled firms (McEachern 1975, Finkelstein and ier to attract them to their new companies (Nohria Hambrick 1989). However, these large-companystudies 1988). Byexperiencing multiple roles or multiple orga- have not had to use a longitudinal approach to studythe nizations, executives build expertise applicable to the effects of ownership structure, given that the structure in entrepreneurial setting, such as operational best prac- large companies rarelychanges dramaticallyover short tices and skills negotiating with multiple constituencies amounts of time. In contrast, in early-stage companies (Dobrev and Barnett 1999). like those in this study, the changes are dramatic during Most applicable to this study, when the Founder-CEO each of the rounds of financing (which maybe onlya is young and inexperienced, there may be a higher prob- few months apart), making the intertemporal changes in abilitythat outside investors will not have confidence ownership structure an important factor to assess.3 in his or her abilityto lead the company,and will
Organization Science/Vol. 14, No. 2, March–April 2003 157 NOAM WASSERMAN Founder-CEO Succession therefore seek to replace the Founder-CEO with a more had been founded, he attributed his verysolid position experienced person. However, if the Founder-CEO has within the companyto his preexisting relationships with a long track record that investors can assess, the out- the team members, who were all still employed at the side investors might be able to gain more confidence in company.Therefore, in myevent-historymodels I also the founder and be less inclined to replace him or her. included the number of companyfounders. One founder I interviewed reflected on the advantages Fourth, we can extend this analysis of the impact of that experienced CEOs have over younger competitors: “allyfounders” to the equitycontrol issues discussed “Would a young CEO work more hours and be more above. Hypothesis 5 focused on how the amount of manic about the business? Yes, but he wouldn’t work equityheld bythe founding team might affect Founder- as smart, and smart is what we need in a CEO.” There- CEO succession. Even when there are multiple founders, fore, in testing the hypotheses described above, I also Founder-CEOs own a disproportionate share of the included the Founder-CEO’s years of prior work experi- equityheld byinsiders (Wasserman 2001). However, ence in myevent-historymodels. Founder-CEOs do not have direct control over the equity A second related factor is the Founder-CEO’s breadth held byother insiders. Therefore, ceteris paribus, where of prior work experience. On the one hand, before the Founder-CEO herself owns more of the equity, she founding their current companies, Founder-CEOs may should have more control over a succession event than have spent their careers in a functional area (e.g., if she owned a smaller percentage. Therefore, in my finance, business development, sales and marketing, models, I also controlled for the percentage of equity technologydevelopment) and developed deep expertise personallyheld bythe Founder-CEO. in that particular area. On the other hand, a Founder- Fifth, in developing Hypothesis 2, I focused on how CEO mayhave achieved a broader “general manage- a VC’s abilityto force a succession event is highest ment” background in business. Having previous general when a companyis raising a new round of financing, management experience can help a CEO adapt to new which should cause the rate of succession to rise after organizational demands and therebyretain his or her each round of financing closes. This assumes that all position for a longer period of time than if the CEO succession events are involuntaryand initiated bysome- has a narrower background (Hambrick and Mason 1984, one other than the outgoing CEO (Friedman and Singh Rubenson and Gupta 1996). Therefore, in mymodels 1989). However, some succession events maybe volun- I also included a control for whether the Founder-CEO tary. This is important because how the predecessor CEO had a narrow, functional background or a broader, gen- leaves office influences who is selected as a successor eral management background. CEO (Sonnenfeld 1986). Most pertinent to the central Third, research on large-companysuccession has issues of this paper, voluntarysuccession events may shown that CEOs who enter the position with allies are speed up the rate at which the CEO position changes more secure than are CEOs who enter without allies hands, for it mayfacilitate the search for, selection, and (Grusky1969). In early-stagecompanies, an analog to integration of a new CEO. Past studies have grappled this would be whether the Founder-CEO has founded with how to identifywhether a succession event is vol- the companyalong with one or more cofounder “allies,” untaryor involuntary(Kesner and Sebora 1994), using or whether the Founder-CEO started the companyon his such proxies as whether the predecessor CEO was older or her own (“without allies”). Founder-CEOs maybe than 65 years old. Aside from the problems introduced able to build a broader base of power byattracting other bythe use of such proxies (Kesner and Sebora 1994), talented executives to help them start the company. A these metrics don’t applyto the subjects of this paper, young founder stated that, “With Mark [his cofounder] for nearlyall of the Founder-CEOs were in their 30s and bymyside, I felt like we were bulletproof and could take 40s and none was older than 49. Given this, is it possible anything that came our way.” A Founder-CEO who has in this setting to separate CEO “dismissal”—i.e., when cofounder allies maytherefore be able to retain his posi- the CEO’s departure is ad hoc and against his or her will tion for a longer period of time, especiallyif the skills (Fredrickson et al. 1988)—from voluntarysuccession? of those cofounders complement his own. Another com- On the one hand, there is a strong psychological link- panyI studied had a verylarge founding team of eight age between founders and the companies theycreate people. The CEO had used his personal network to pull (Dobrev and Barnett 1999), indicating that Founder- the team together, recruiting each founding team mem- CEO succession maybe predominantlyinvoluntary.In ber for his deep expertise in a functional area critical fact, the vast majorityof the Founder-CEO successions I to the company’s success, and the team as a whole for studied were involuntary. On the other hand, there were its complementaryskills. Two yearsafter the company four cases I studied where the Founder-CEO was the
158 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession one who realized that a change was necessary. These directors (Fiet et al. 1997). On boards that have a high were typically inexperienced founders who realized that percentage of insiders, the Founder-CEO’s position may investors would be unwilling to invest capital in a com- be more secure because board members who have been panythat did not have a seasoned executive in charge, handpicked bythe CEO are often less vigilant than are and therefore proactivelyinitiated a CEO change. For outside directors who are on the board because they instance, one founder I interviewed recognized that, are major shareholders (Fiet et al. 1997). Outside board “The VCs wanted a celebrityCEO as part of the pack- members tend to focus on organizational performance in age when we went to them” and initiated a professional- assessing CEOs and are more inclined to dismiss CEOs CEO search that he planned to complete before the next of low-performing companies (Mizruchi 1983). At the financing round. Another founder realized the need for a same time, while outsiders are less beholden to the CEO, change after a series of meetings with accounting firms theyalso lack the firm-specific knowledge that inside and potential investors where directors would have (Fredrickson et al. 1988). To con- trol for this heterogeneitybetween boards, I included I did not feel theywent well. Theywere asking questions with acronyms I didn’t know, and I felt I was hurting the company. in mymodels a control for the percentage of the board I thought I might be in over myhead, and those meetings consisting of inside directors. reinforced that I did not know all I needed to run the company. The second company-level factor is the company’s primarybusiness segment. A recent studyshowed Both of these founders “voluntarily” brought in profes- that a CEO’s abilityto influence companyperfor- sional CEOs shortlybefore their next financing round. mance varies markedlybybusiness segment (Wasser- Even when interviewing the people involved in a man et al. 2001), which can affect the rates of CEO succession event, it can be hard to tell when a particu- succession across segments. In the Internet indus- lar succession event trulywas voluntaryor involuntary. try, segments can differ in structure, capital inten- Even so, the cases described above suggest a possible sity, maturity, growth rates, and other factors. In this wayto control for whether the succession was voluntary: data set, the segments were as follows: business-to- that “voluntary” succession may occur more frequently business e-commerce (“B-to-B”), business-to-consumer in the months before a financing round.4 In between e-commerce (“B-to-C”), content/new media, infrastruc- financing rounds, when outside investors do not have ture, and services. In mymodels I created a dummy the power to withhold capital in order to effect a CEO variable for each of these segments, which would allow change, convincing the founder to step down as CEO the succession rates to varybysegment. would be a painful and drawn-out process of persuasion. As described above, large-companysuccession stud- This makes it more likelythat succession events in the ies have consistentlyfound that good recent company time before a financing round maybe voluntaryones, performance—e.g., a large increase in sales (Fiet et al. in contrast to the investor-driven succession events of 1997)—significantlydecreases the probabilityof CEO Hypothesis 2, which happen upon the closing of a new dismissal (Kesner and Sebora 1994). Past studies have round of financing. Therefore, to control for this possi- predominantlyused profitability-or sales-based metrics bility, I included in my models a time clock measuring of performance. However, in the early-stage companies the amount of time until the next round closed. that are the focus of this study, these metrics would be Company Level. The company-level factors include problematic because in the time period before compa- the characteristics of the company’s board of directors, nies begin selling products to customers, theyhave no the company’s primary business segment, its recent level revenues and estimates of their profits or losses can be of performance, and the number of people employed by unreliable at best. The best alternative that applies to the company. start-ups is probablythe company’s most recent valua- CEO dismissal is usuallyinitiated bythe board (Mace tion. When a private companyraises a round of financing 1971, Mintzberg 1983, Mizruchi 1983, Lorsch 1989), from investors, those investors place a valuation on the making board characteristics a potentiallyimportant fac- company, akin to a public company’s market value on tor in Founder-CEO succession. Further, complex and the stock market (Wasserman 1999). When the company ambiguous decisions—such as those about whether to has completed keymilestones and succeeded in ways replace the CEO—intensifythe differences between that gain the confidence of its investors, the company’s individual board members (Mintzberg 1983), especially valuation will usuallyincrease accordingly,while failure in industries that are highlycompetitive (D’Aveni 1994). to achieve milestones can result in a decrease in valua- The most frequentlystudied typeof board diversity tion (Sahlman et al. 1999). Therefore, as an indicator of is the mix of outsider/investor versus insider/executive recent companyperformance, I used the valuation placed
Organization Science/Vol. 14, No. 2, March–April 2003 159 NOAM WASSERMAN Founder-CEO Succession on the companybyinvestors in the most recent round because it included detailed questions on a broad array of financing. of issues, the surveycould onlybe completed bya top Finally, the number of employees can also play officer of the company(i.e., the CEO, CFO, or head of an important role in the occurrence of Founder-CEO human resources with the help of the CEO or CFO). succession. For instance, Greiner argued that the more Pretesting of the surveyshowed that the surveywould people employed by a firm, the more acute the com- require at least 20 minutes to complete, an amount of munication and coordination challenges, the more dif- time that CEOs and CFOs might not be willing to invest. ferentiated the functions within the firm, and the more Further, surveys of CEOs—even those of stable, estab- complex the problems faced bythe entrepreneur-CEO lished companies—“have historicallysuffered from low (Greiner 1972). Therefore, I used the number of employ- response rates” (Zajac 1990, p. 223). ees, a robust metric of companysize, in mymodels. I pretested mysurvey(with 10 people from a vari- etyof Internet companies) in order to learn how to External Factors. Even broader, the performance of avoid some of these problems. Results of the pretest- individual companies depends at least in part on external ing and of follow-up interviews with the participants market factors (Porter 1980). For example, an indus- enabled me to refine four of the surveyquestions and try’s stage of development affects how much the indus- to reduce the amount of time that would be required to try“paradigm” has developed (Porter 1980), which complete the questionnaire. In addition, discussions with affects the variance in performance and the number of pretesters indicated that the top executives of Internet firms against which each firm competes (Fredrickson companies severelylacked information about compen- et al. 1988) and can influence the size of the primary sation in the industry, and believed that they suffered executive-level talent pool (House et al. 1985). Within increased turnover within their companies because of the Internet industry, while the growth was consistently it. This was true both for successful companies, which strong throughout most of mydata set, the crash of wanted to make sure that theycould continue to attract April 2000 caused dramatic shifts in the dynamics of new employees and that their current employees could the industry(Barrett 2000) with possible implications not be “poached” bycompetitors, and for their less- for Founder-CEO succession. Therefore, I included a successful counterparts, for whom attraction and reten- dichotomous variable indicating whether the succession tion were vital. Therefore, the main inducement that I event took place before or after the crash. used to get top executives to fill out the surveywas the promise to provide respondents with a copyof the Data and Methods aggregate compensation results. (As an indication of the effectiveness of this inducement, of the respondents, all Sample but one requested a copyof the aggregate results.) I As described above, past studies of CEO succession have accompanied this promise with a concise statement that focused on organizations, such as large companies or ensured complete confidentialityregarding specific com- major sports teams, whose leaders were not founders of panysubmissions. Of the six pages in the final sur- the organization. Data on these organizations are pub- vey, three were compensation related, two were finance liclyavailable, facilitating research on them. However, related, and onlyone page directlypertained to Founder- in order to studyFounder-CEO succession, I needed CEO succession. As a result, the title of the survey data that were not publiclyavailable, for private compa- was the “Information TechnologyCompensation Sur- nies do not have to disclose such data and rarelydo so vey,” which helped reduce the chances that companies voluntarily. Therefore, I conducted a survey of Internet that had experienced messyFounder-CEO succession start-ups to collect data for this studyand other related events would be sensitive to filling out the survey. projects. To build a list of candidate companies, I began with On one hand, financial data are relativelyunambigu- the Venture One database, creating a list of all com- ous and their reliabilityis strong (Antle and Smith panies whose primaryline of business involves the 1985), so theylend themselves to collection via a sur- Internet. After eliminating the companies with invalid vey. On the other hand, doing a survey introduced two addresses or that lacked contact information, I ended up major problems. The first problem was the fact that with a list of 1,037 companies. Questionnaires were sent some of the data I sought—for example, the equityhold- to all of them. I received valid surveyresponses from ings of insider executives versus outsider investors at 202 of the companies, or 20% of myinitial list of candi- each stage of financing—was verysensitive. The sec- date companies. The 20% response rate was reasonably ond problem in collecting the data was the fact that, high considering the sensitivityof the questions and the
160 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession level of executives targeted (Finkelstein 1992, Waldman proportional (an assumption I test below). A particu- et al. 2001). Of the respondents, 59% were the chair- lar strength of Cox models for this analysis is the fact man, CEO, or CFO of the company, with an additional that theyallow us to include the time-dependent covari- 21% being the vice president of human resources. ates (Blossfeld and Rohwer 1995) necessaryto test my Most important for this paper, the surveyincluded hypotheses. questions about when development was completed on In this data set, the risk set at each time point tconsists the company’s first product (Hypotheses 1A and 1B), of all Founder-CEOs who are still in the CEO position at the timing of financing rounds (Hypotheses 2 and con- the time point, and therefore “at risk” of being replaced. trols), financial details about company-financing history The hazard rate ht is the probabilitythat a Founder- (Hypotheses 3 and 5), and executive, board, and founder CEO will be replaced between consecutive time points, demographics (control variables). In addition, to test given that the Founder-CEO is at risk of replacement at Hypotheses 4A and 4B about the effect of having new the earlier of the two time points. The destination state investors, I used data from Venture One on the rounds of interest is the exit of the Founder-CEO from the role in which each venture capitalist began investing in each of CEO, with the subsequent entryinto the firm of the company. 5 To test the representativeness of mysample, I obtained second CEO. To measure Founder-CEO succession, I from Venture One the distributions of venture-backed modeled the likelihood that the founding CEO would be Internet companies across three keydimensions. I used a replaced. This variable was coded 1 at the time point Kolmogorov-Smirnov two-sample test to assess whether when the Founder-CEO was replaced, and 0 otherwise. respondents were statisticallydifferent from the Ven- Firms where the Founder-CEO was still the company’s ture One universe. In all three dimensions—industry CEO as of the end of the data collection period (July31, segment, geography, and company stage (product-in- 2000) were treated as censored observations, as were development versus shipping)—there were no statis- cases where the companywas acquired (two compa- ticallysignificant differences, indicating a lack of nies) or went out of business (three companies) before response bias. In addition, for those questions that could there was a Founder-CEO succession event. Because I be verified, I checked randomlyselected data from many was able to get complete data on companyhistoryfrom of the responses against Venture One to reassure myself time of founding, left truncation is not a problem in this that the surveyanswers were accurate. data set. Geographically, of the 202 responding companies, With regard to the variables I used to test myhypothe- 47% were headquartered in California, 17% in the Mid- ses, the grounded-theoretical discussion above suggests Atlantic states, and 17% in the Northeastern states. With that there are both time-constant and time-varying fac- regards to business segments, 36% of the companies tors that influence the Founder-CEO succession event. focused on business-to-business e-commerce, 26% on Therefore, in mymodels I use both constant and time- enabling software and infrastructure, 13% on business- varying predictor variables. Variables that are constant to-consumer e-commerce, 10% on Internet services, 7% over time include the controls for the number of people on content and new media, and 8% were categorized as who founded the company, the Founder-CEO’s years of “other.” Almost all of the companies in the data set were work experience before joining the company, whether founded between 1995 and 1999. the Founder-CEO had a narrow functional background before founding the company, and the company’s busi- Methods and Variables ness segment. For the first two of these variables, I took To test myhypotheses,I modeled Founder-CEO suc- cession rates using a Cox event-historymodel (Cox the data on founders and Founder-CEO work experience 1972). An event historyis a record of when a focal directlyfrom the surveyresponses. For the breadth of the event occurred to a sample of individuals or organi- Founder-CEO’s prior experience, I coded a dichotomous zations. In general, event-historymethods enable us to “narrow experience” variable as 1 if the Founder-CEO make causal inferences about how changes in one vari- had worked in a functional area (e.g., finance, busi- able affect the focal event (Blossfeld and Rohwer 1995). ness development, sales and marketing, technology, or More specifically, Cox models—“unequivocally the best human resources) without having served in anygeneral- all-around method for estimating regression models with management positions, anda0iftheFounder-CEO continuous-time data” (Allison 1984, p. 35)—use par- had held a general-management position before found- tial likelihood to estimate hazard rates, while assuming ing the current company. For the business segments, that transition rates for different values of covariates are I created one dummyvariable for each of the five
Organization Science/Vol. 14, No. 2, March–April 2003 161 NOAM WASSERMAN Founder-CEO Succession segments (business-to-business e-commerce, business- companies in the sample. Table 1 shows summarydata to-consumer e-commerce, content/new media, infras- and a correlation matrix for the predictor variables used tructure, and services). in the event-historymodels. Among the predictor vari- To be able to include time-varying covariates, I used a ables, there were high correlations between the num- spell splitting technique (Tuma et al. 1979, Gould 1999). ber of rounds of financing completed and the amount For each unit of time (here, each month) that a Founder- raised (r = 0 48, p<0 001; companies tend to raise CEO was at risk, I created a separate observation record. larger rounds as theymature), between the amount raised For each person-month, I coded the dichotomous depen- and the number of new VCs participating in the round dent variable as 1 if the Founder-CEO was replaced in (r = 0 30, p<0 01; in order to raise more money, more that month, and 0 otherwise. Finally, I pooled all of the VCs have to participate in the round), and between the person-months into a single sample, then estimated Cox number of rounds of financing completed and whether models for the data set. Because this procedure treats insiders control a majorityof the equity(r =−0 55, each month at risk as a separate observation, I could p<0 001; each round, more equityis sold to outsiders, include in the model the time-varying explanatory vari- diluting insiders’ equityholdings). ables hypothesized to affect Founder-CEO succession.6 To test the hypotheses, I built a series of three nested The time-varying variables include information about Cox models that estimated the effects of the indepen- the latest round of financing and about the company’s dent and control variables on Founder-CEO succession. status with regard to product development. First, financ- Table 2 presents these models.8 Model 1 is a “baseline” ing rounds are well-defined events that are accurate Cox model that predicts the likelihood of Founder-CEO down to the month in which theyoccurred. 7 To ensure succession on the basis of the control variables for the the accuracyof the round dates, I checked everyround 202 companies. Model 2 adds to this baseline the effects against the Venture One database and against company of completing product development, and the full Model press releases. I then calculated time clocks for both the 3 adds the variables relating to financing rounds. With time until the next round of financing and the time since regards to assessing individual predictors, I focus on the the last round of financing (for company-months after significance of the t-statistics for the null hypothesis that the first financing round). Time-varying data that change each coefficient is equal to zero (Allison 1995). each financing round include the amount raised (in $M), The “baseline” model shows that with regard to the most recent valuation (in $M), the percentage of the individual-level controls, whether the Founder-CEO the companyowned byoutside investors, the percentage had a narrow background before founding the com- personallyowned bythe Founder-CEO, how manynew panyis highlysignificant at p<0 005 with a positive investors participated in the round, and the percentage coefficient, indicating that having a narrow background of board members who were inside executives. Second, significantlyincreases the hazard of Founder-CEO suc- the company’s status with regard to product development cession. The percentage of equitypersonallyowned by was captured via a dichotomous indicator of whether the the Founder-CEO is significant at p<0 05 with a nega- initial product had begun shipping and a time clock that tive coefficient, suggesting that the higher the percentage tracked the amount of time since product shipment had owned, the lower the chances of succession. At the com- begun. As with the dates for financing rounds, I checked panylevel, the time-since-founding time clock is highly surveyresponses for product shipping dates against com- significant at p<0 005 with a positive coefficient, indi- panywebsites, press releases, and press articles writ- cating that as a companyages, the hazard of Founder- ten on the companies. Because the time distributions of CEO succession increases significantly(before we add financing rounds and time elapsed since initial product anyother time clocks to the model). The company’s shipping were both lognormallydistributed, I logarithmi- most recent valuation is significant at p<0 10 with a callytransformed these time clocks in myevent-history negative coefficient, suggesting that the higher the com- models. pany’s valuation, the lower the hazard of succession. No other variables were significant in this model,9 which had a log-likelihood of −195 64. Results Model 2 adds the time clock that tracks the (log) The spell-split data set included 5,930 total person- number of months since the completion of product months. Of the 202 companies included in the data development. This time clock is highlysignificant at set, 60 experienced a Founder-CEO succession event. the p<0 01 level and indicates that the rate of CEO Figure 1 shows a graph of the overall Kaplan-Meier succession is high immediatelyafter the completion of survival estimates (with 95% confidence band) for the product development and decreases after that. Among
162 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession
Figure 1 Graph of Kaplan-Meier Survival Estimates for All 202 Founder-CEOs in Sample
Kaplan-Meier survival estimate 95%, pointwise confidence band shown
1 11 1.00 22 33 18 79 165 25 33 5 14 2 0.75 3 6 3 2 3 4 4 4 0.50 5 3
2
0.25
. 1
0.00 0 50 100 150 analysis time
Table 1 Summary Statistics and Correlation Matrix for Cox Model Variables
Std. Mean Dev. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
(1) After April 2000 0 12 0 33 1 00 (2) Number of founders 2 46 1 29 0 01 1 00 (3) Most recent valuation 16 34 90 85 0 16 0 12 1 00 in $M (4) Number of 29 65 58 01 0 21 0 07 0 72 1 00 employees (5) Pct. of board who 0 81 0 20 0 04 0 17 0 14 0 08 1 00 are insiders (6) Founder-CEO’s 0 47 0 30 0 02 0 51 0 15 0 20 0 29 1 00 equity pct. (7) F-CEO has narrow 0 49 0 50 0 14 0 15 0 05 0 01 0 21 0 05 1 00 background (8) Time since product 11 65 21 20 0 08 0 20 0 04 0 19 0 06 0 09 0 14 1 00 shipping (9) Rounds to date 1 05 1 15 0 05 0 07 0 46 0 36 0 19 0 44 0 05 0 34 1 00 (10) Number of new VC 0 98 1 45 0 03 0 05 0 12 0 20 0 14 0 21 0 09 0 05 0 08 1 00 investors (11) Amount raised in last 6 02 13 78 0 13 0 13 0 55 0 48 0 10 0 46 0 04 0 02 0 48 0 30 1 00 round (12) Time since last round 6 25 6 45 0 04 0 02 0 18 0 08 0 11 0 07 0 10 0 29 0 19 0 05 0 14 1 00 (13) Time to next round 13 27 17 57 0 24 0 09 0 13 0 17 0 03 0 04 0 15 0 11 0 05 0 06 0 09 0 18 1 00 (14) Insiders control 0 76 0 43 0 06 0 19 0 06 0 09 0 40 0 46 0 05 0 28 0 55 0 30 0 39 0 01 0 10 equity
Organization Science/Vol. 14, No. 2, March–April 2003 163 NOAM WASSERMAN Founder-CEO Succession
Table 2 Event History (Cox) Models for Founder-CEO Succession
Model 1 Model 2 Model 3 Control Variables Completion of Product Dev’t. Product Dev’t. and Financing
Hazard Hazard Hazard Ratio Coef. (Std. Err.) Ratio Coef. (Std. Err.) Ratio Coef. (Std. Err.)
Control Variables Founder-CEO’s equity % 0 172 −1 761 0 791 ∗∗ 0 571 −0 560 1 202 5 600 1 723 1 115 F-CEO’s years of prior exp’c 1 031 0 031 0 023 1 022 0 027 0 023 1 290 0 081 0 061 F-CEO has narrow backgrd. 3 539 1 264 0 362 ∗∗∗∗ 4 083 1 407 0 575 ∗∗ 2 136 0 759 1 249 Number of founders 0 873 −0 136 0 144 0 966 −0 035 0 202 2 195 0 786 0 483 Time since founding (L) 8 191 17 222 1 648 ∗∗∗∗ 7 421 14 883 8 624 1 401 1 303 3 181 Most recent valuation (S) 0 878 −0 130 0 073 ∗ 0 891 −0 115 0 097 0 853 −0 159 0 155 Number of employees 1 000 0 000 0 003 1 002 0 002 0 002 1 003 0 003 0 009 % of board who are insiders 0 646 −0 436 1 057 0 771 −0 260 1 367 0 866 −0 144 1 255 Segment 2 (Bus.-to-consumer) 0 944 −0 058 0 382 0 659 −0 417 0 495 0 009 −4 685 3 848 Segment 3 (Content/NM) 0 957 −0 044 0 581 0 664 −0 410 0 831 0 025 −3 695 2 423 Segment 4 (Infrastructure) 0 777 −0 252 0 714 0 159 −1 840 1 155 0 428 −0 848 3 225 Segment 5 (Services) 0 780 −0 248 0 457 0 312 −1 164 0 912 0 051 −2 976 2 158 After April 2000 1 168 0 155 0 493 1 338 0 291 0 538 8 042 2 085 1 329 Predictor Variables Time since product ship’g (L) 0 482 −0 730 0 262 ∗∗∗ 0 093 −2 379 0 843 ∗∗∗ Rounds to date 0 039 −3 247 1 636 ∗∗ Time since last round (L) 0 185 −1 688 0 852 ∗∗ Time to next round (L) 0 970 −0 030 0 254 Amount raised last rd. (L) 2 220 0 798 0 173 ∗∗∗∗ Number of new VCs 0 090 −2 406 0 671 ∗∗∗∗ Insiders control equity 0 030 −3 495 1 476 ∗∗ Log likelihood −195 64 −77 92 −10 57
Note. ∗p<0 10; ∗∗p<0 05; ∗∗∗p<0 01; ∗∗∗∗p<0 005 Variable transformations: L = natural log; S = square-root “Std. Err.” columns show robust standard errors. No. of companies 202. No. of Founder-CEO succession events 60. No. of Total observations 5930. Dependent variable: Whether a Founder-CEO succession in event occurred at time point t. the controls in this model, while the variable indicat- months since the companybegan shipping the prod- ing whether the Founder-CEO had a narrow background uct decreases the hazard of succession bya multiple loses some significance, it is still significant at p<0 05. of 0.193.10 This supports Hypothesis 1B (that complet- With the addition of the shipping time clock, the time- ing product development precipitates a succession event) since-founding time clock’s standard error increases to over Hypothesis 1A (that completing development solid- the point where the time-since-founding variable is no ifies the Founder-CEO’s position). longer significant. Model 2’s log-likelihood is −77 92. Moving to the financing variables, the time since last Model 3 is the full Cox model, which adds the round was significant at the p<0 05 level, support- financing-round variables. Focusing on the individual ing Hypothesis 2 that the rate of Founder-CEO suc- predictors in Model 3, I can assess the hypotheses devel- cession increases after the companyraises a round of oped in this paper. With regard to Hypotheses 1A and financing. In terms of practical significance, doubling 1B about the effects of the completion of product devel- the time-since-last-round time clock decreases the haz- opment on the rate of Founder-CEO succession, the ard of succession bya multiple of 0.310. 11 With regard time clock for the completion of product development to Hypothesis 3, the amount raised in the last round of is significant at the p<0 01 level. The more recently financing is significant at the p<0 005 level. The higher the first product began shipping, the higher the rate of the amount raised in the last round, the higher the rate of succession. More specifically, doubling the number of Founder-CEO succession, supporting Hypothesis 3 that
164 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession large rounds are related to higher turnover. In terms of strength of this factor overwhelms the power gained by practical significance, a doubling of the amount raised the Founder-CEO from having successfullyled the com- in the last round increases the hazard of succession by panyto a major milestone. a multiple of 1.738.12 With regard to Hypotheses 4A The second intertemporal event is the raising of new and 4B, the number of new VCs participating in the lat- rounds of financing. During these rounds, Founder- est round of financing was significant at the p<0 005 CEOs negotiate with potential investors in an effort to level (with a hazard multiple of 0.090 for each new VC), receive capital to build their companies. If such investors supporting Hypothesis 4B (that the involvement of new doubt the Founder-CEO’s abilityto lead the company VCs is linked to a lower rate of succession) over com- over the long run, theywill be unwilling to invest unless peting Hypothesis 4A (that the involvement of new VCs there is a change in CEO. As shown in this paper, is linked to a higher rate of succession). With regard investors have a particularlyhigh amount of leverage to to the effects of outside equityownership, the equity- effect a Founder-CEO succession event at the time that control variable is significant at the p<0 05 level. When Founder-CEOs come to them for a new round of financ- insiders own more than half of the equity, the hazard ing, for the investors can refuse to provide capital if their of Founder-CEO turnover is much lower (bya hazard demands are not met. The power to make demands is multiple of 0.030), supporting Hypothesis 5. greatest when the companyis raising a large amount of The next section discusses these findings in light of capital and when outside investors own a higher percent- the literature on CEO succession and describes potential age of the company, giving them additional control of future research possibilities. critical decisions within the company. These results enable us to compare and contrast suc- Discussion cession in larger companies, which has been the focus The results in this paper have shown how intertempo- of past CEO-succession research, with the Founder-CEO ral events affect the likelihood that a Founder-CEO will succession event that is the focus of this paper. As with be replaced. Using myfield results and a unique dataset later-stage succession events, equitycontrol can playa that included 202 companies and 5,930 person-months significant role in Founder-CEO succession. However, of data, the analysis provides strong evidence of a link- there are some critical ways in which Founder-CEO suc- age between Founder-CEO succession and the comple- cession departs from the later-stage succession findings tion of both product development and each round of from past organizational research. Most important, in financing. contrast to large companies, where delivering good per- The first intertemporal event examined here was the formance helps ensure CEO longevity(Grusky1963, completion of product development. A companyin the Allen et al. 1979, Jensen 1986, Useem 1993, Denis and midst of product development faces critical technical Denis 1995), success can be detrimental for Founder- challenges. Earlyon, Founder-CEOs who are adept CEOs who want to remain in their position. In this at solving such challenges are often able to attract “paradox of success,” Founder-CEOs who are successful high-qualitytechnical people, to manage the product either at leading their companies to successful comple- development process well, and to help their organi- tion of product development or at raising a round of zations succeed at developing the product efficiently. financing have a higher rate of turnover. As initiallyindi- However, once the initial product has been developed, cated bymyfield work, succeeding at leading a com- the CEO’s job broadens and gets much more com- panyto keymilestones often means that the company’s plex, for he or she has to begin selling the product needs outstrip the Founder-CEO’s skills faster. While to customers, building an organization to support the the Founder-CEO’s skills were a good “fit” for the con- product, and creating a marketing team. This dramatic tingencies faced bythe companybefore, enabling the change in the contingencies faced bythe firm often companyto reach its critical milestones, those skills are results in a mismatch between the skills of the techni- usuallymuch less important now that the companyfaces callyadept Founder-CEO—whose skills were the key radicallydifferent contingencies. As a VC from a large to success until now—and the new needs of the orga- Northeastern firm told me: nization. The fact that the rate of succession increases immediatelyafter the completion of product develop- In situation after situation we have found that if we are really ment suggests that companyowners proactivelyassess focused on building a big and important business, and we do the qualityof this skills-contingencies fit and make CEO a great job of launching a companyquickly,growth often out- changes before a mismatch would cause problems. The strips the founding CEO’s abilityto manage.
Organization Science/Vol. 14, No. 2, March–April 2003 165 NOAM WASSERMAN Founder-CEO Succession
Furthermore, succeeding at raising a large round of who were able to remain in their positions for a long financing also increases the chances of succession signif- period of time: icantly. By trying to raise a large round, Founder-CEOs Their businesses grew relativelyslowly,with relativelylittle put themselves at the mercyof capital providers, increas- VC backing, so theyhad time to develop management skills ing the hazard of succession. Founder-CEOs therefore whereas VC-backed startups tend to grow quickly, growing can face a critical dilemma, in which performing poorly faster than the founder’s bandwidth can expand. would have undesirable consequences, but performing well mayalso cause them to lose their positions. One In firms that either lack the resources to grow or are Founder-CEO who participated in the surveytold me in led byfounders who do not want the firm to grow, a postsurveyfollow-up interview whyhe had been able founders can maintain direct control of the critical orga- to retain the CEO position: nization decisions (Carland et al. 1984). Furthermore, such Founder-CEOs also have more time to increase The onlyreason I’m still CEO is that we haven’t been at either the value of the companybefore selling equityto out- end of the spectrum. We haven’t “cratered,” so the VCs haven’t side investors, which means that theywill be able to been banging the table calling for myhead. But the VCs also retain a larger percentage of the equityfor themselves. don’t see us turning into one of their homeruns, so theydon’t From a situational perspective, Founder-CEOs who want want to eat up their time and resources trying to find a superstar to run the company. to remain CEO might select business segments or busi- ness concepts where the critical contingencies faced by It is important to note that the impact of “success” their companies do not change markedlywhen prod- on the rate of CEO succession has been measured dif- uct development is complete, such as where the com- ferentlyin large-companystudies from how it was mea- pany’s advantages are built on relationships with key sured in this study. Past large-company studies have partners, both before and after product development is measured success using such metrics as the company’s completed. Founders interested in retaining the CEO profitabilityor sales growth (Kesner and Sebora 1994). position might also concentrate on proactivelybuilding However, as noted in the “Data and Methods” section the skills that will be necessaryonce such critical mile- above, these metrics do not applywhen companies are stones are reached, even if those skills are not needed not yet shipping products, which can be true for a sub- earlyin the company’s history. stantial period of time in new companies. Instead, for Looking to the broader literature, the intertemporal their metrics of success, young companies rely on the evolution of equitycontrol, in which each financing achievement of event-driven milestones (Sahlman et al. event changes the balance between ownership and man- 1999) such as the successful completion of a round of agement, can help give us a richer picture of agency financing or of initial product development. Myfield theory.As generallyapplied, agencytheoryassumes that research showed that these events can have important managers and owners are distinct groups whose inter- impacts on Founder-CEO succession, and were therefore ests diverge due to their differential ownership stakes the metrics I used in myanalyses.However, we should in the company. This assumption reduces agency the- also keep in mind that theyare not directlycomparable ory’s applicability to early-stage companies that have not to the “success metrics” commonlyused when studying accepted outside capital, or where the CEO still owns large-companysuccession. as large a share of the companyas do the largest out- The results of this studyalso help show what types side investors. However, from an agencyperspective, we of personal characteristics and situations might help can view entrepreneurship as the process bywhich own- Founder-CEOs retain their positions for a longer period ership and management become separated. At the time of time. For example, with regard to personal char- of companyfounding, when there are not anyoutside acteristics, if the founders bring their own financial investors, the people running the firm own all of the resources to the company(e.g., from their past suc- equityand the principals are the same as the agents. As cesses at founding companies), theyare less depen- each round of financing is completed and the manager- dent on outside investors and less susceptible to their owners sell more of the firm to outside investors, the firm demands for a change in companyleadership. Theycan begins to look more like the models studied in agency- wait longer before raising their first round of financing, based analyses. At the same time, at all stages of private- which enables them to prove—and improve—their abil- companydevelopment, Founder-CEOs own far more of ities for a longer period of time. A young Founder-CEO the firm’s equitythan do nonfounding CEOs (Wasser- whose companyfailed due to a lack of funding made man 2001). Interestingly, this means that replacing a the following observations with regard to Founder-CEOs Founder-CEO with a professional CEO who owns less of
166 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession the firm can, in fact, worsen the principal-agent problem, make the industryinto the “Mediterranean fruit flyof for the lesser amount of equityheld bythe professional succession research:” There are many“subjects” to study CEO could mean that the professional CEO’s interests and theydevelop veryquickly.However, the benefits are less aligned with the interests of the investors. In gained from focusing on the Internet sector do come addition, the outsider professional CEO often has to with some costs. For example, bylimiting myfocus to climb a steep learning curve, further increasing the costs the Internet sector, I mayhave introduced biases that are of succession. unique to this setting. Does the compressed time frame On the other hand, there are other theoretical per- of competing in an ultracompetitive industry(Eisenhardt spectives that might give us a richer understanding of 1989, D’Aveni 1994) force boards to take material short- the reasons whyFounder-CEOs are replaced. As men- cuts in the succession process? Do Founder-CEOs usu- tioned above, the “changing contingencies” perspective allystayat the firms theyfounded, or does the Internet (Hickson et al. 1971) gives us strong forward-looking industry have idiosyncratic characteristics of psycholog- reasons for whyFounder-CEOs might be replaced after ical attachment that lead its former Founder-CEOs to achieving critical milestones. In addition, another per- stayat their firms more often than in other industries? spective that might shed further light on Founder-CEO Are Internet companies more susceptible to the demands succession is institutional theory. Some of the quotes of outside investors, given their capital requirements and from myfield research indicate that venture capitalists rates of growth? Future research that extends myanal- and other keysuccession playersmake decisions guided yses to other entrepreneurial settings such a biotechnol- by “rationalized myths” (Meyer and Rowan 1977), fol- ogyor medical devices would help locate the potential lowing taken-for-granted assumptions, beliefs, and rules biases caused bya focus on the Internet sector. Other of thumb about Founder-CEOs. For example, one VC biases mayhave been introduced bythe typesof firms described above how his “default assumption” is that that responded to the compensation surveyfrom which I the Founder-CEO will not be able to lead the company obtained myquantitative data. For instance, as described for a sustained period of time, and he therefore seeks in the Data and Methods section, the survey’s respon- to remove the Founder-CEO much earlier in the process dents matched the distribution of private companies in than might be expected. However, when pushed for how the Venture One database. However, the sample was he had arrived at this rule of thumb, he admitted that it dominated bycompanies that had raised venture financ- was a widelyheld belief within the venture capital indus- ing, as is true of Venture One itself. These companies tryand one that he had not questioned. This maybe true are typically more successful than are similar compa- even when the best decision maybe to allow some time nies that have not raised venture money(Gompers and to learn about the Founder-CEO’s capabilities bygiv- Lerner 2001). Theymayhave different ownership struc- ing the founder a honeymoon period (Choi and Shanley tures, mayhave boards that are more outsider-heavy, 2000). Other such “institutions” or “conventions” that and maybe at a different stage of organizational evo- mayaffect Founder-CEO succession include the rules of lution. While some of these factors could decrease the thumb about how much equityoutside investors should hazard of Founder-CEO succession, most of them could receive in the earlyrounds of financing, the appropriate increase it. Future research could help illuminate the role for outside board members in making critical deci- actual impact of anybiases introduced bya venture- sions, and whether Founder-CEOs should remain at their heavysample. companies after being replaced. Even though the people Other fertile areas for research include gaining a richer who adhere to them rarelyassess the reasoning behind understanding of how CEO succession is affected by such rules, the rules help add legitimacyto the demands other characteristics of financing rounds, other Founder- made byVCs. CEO characteristics, the event of going public, and how As the first large-scale exploration of Founder-CEO the timing of Founder-CEO succession affects firm sur- succession, this paper probablyopens up at least as vival. First, within financing rounds, does Founder-CEO manyquestions as it answers. Our knowledge of CEO succession increase when the new investor in a round succession could benefit greatlyfrom future exploration is high status, compared to when the new investor is a of these questions. lower-status VC firm? Is Founder-CEO succession more For example, in this study, I focused exclusively on prevalent when one investor owns all of the outside Internet firms. The fast growth of these firms—in which equityand holds one board seat, or when the outside the time period for founding the firm, building the team equityis more evenlyspread across investors who col- and board, developing the product, and growing to be lectivelyhold more than one board seat? Second, the a large organization is compressed dramatically—helps Founder-CEO characteristics controlled for here were
Organization Science/Vol. 14, No. 2, March–April 2003 167 NOAM WASSERMAN Founder-CEO Succession years and breadth of prior work experience. Besides operating capacitywithin their organizations (Kesner prior work experience, what other Founder-CEO char- and Sebora 1994), more than half of the Founder-CEOs acteristics affect the rate of succession? For Founder- in this studyremained with their companies after the CEOs who worked in Internet companies before, does succession event. This has important implications for the the rate of succession decrease? For Founder-CEOs who process bywhich new executives “take charge” of their previouslyfounded another company(whether Internet- organizations (e.g., Ciampa and Watkins 1999). While focused or not), does the rate decrease? Are Founder- the former CEOs of large companies rarelyremain with CEOs who have alreadyaccumulated substantial wealth their companies, it is verycommon for former Founder- quicker or slower to step down? Does the gender or edu- CEOs to remain with their companies, either in the cational background of the Founder-CEO affect the rate chairman’s role or in a role below the CEO.13 Even after of succession? a succession event, manyinvestors—and even some suc- Third, while mydata set did not include manylater- cessor CEOs—want the Founder-CEO to staywith the stage companies, myfield research indicates that the companyin some capacity.As one investor told me, process of preparing to go public could also have a “You can replace an executive, but you can’t replace powerful effect on Founder-CEO succession, similar to a founder.” Once the professional CEO has entered the the effect of raising a large round of private financ- company, the deposed Founder-CEO has to negotiate a ing. After all, as one executive consultant commented new role, often from a position of weakness. For the to me, “Underwriters put the team under a microscope company’s new CEO, the continuing involvement of the and are ruthless about it,” much like VCs who perform Founder-CEO can introduce some verydifferent “tak- due diligence on a potential investment. Does the rate ing charge” challenges than when the previous CEO of Founder-CEO succession increase prior to going pub- does not staywith the company.This is especiallytrue lic, given that the firm’s financial contingencies should when the Founder-CEO remains in an operating position change dramaticallyonce it is a publiclyheld firm? Does within the company. However, even when the Founder- the rate of succession differ byhow large a public offer- CEO does not retain an operating role, the Founder-CEO ing the firm decides to pursue, just as the amount of often keeps his seat on the board of directors (often moneyraised in each financing round affects succes- as chairman), which also introduces critical board man- sion when the firm is private? Fourth, compared to large agement challenges for the new CEO. The former CEO companies, whose survival rates are relativelyhigh, the can have a major effect on an organization long after mortalityrate of small companies is veryhigh (Sahlman the CEO has left office (Gordon and Rosen 1981), par- et al. 1999). According to some researchers, organi- ticularlywhen the former CEO becomes chairman of zational survival is the ultimate indicator of success the board, able to look over the new CEO’s shoulder (Hannan and Freeman 1989). Interestingly, young news- (Fredrickson et al. 1988). paper firms have been found to have a disproportionate Board members’ perceptions of whether the Founder- likelihood of failure after publishers who were founders CEO will staymayeven affect the timing of the suc- resign (Carroll 1984). Further research that focused on cession event itself. As a venture capitalist expressed it, the survival rates of entrepreneurial firms that expe- board members often go through a “feeling-out process” rience Founder-CEO succession at different points in during which they“take the measure of the Founder- their development would enable us to perform a survival CEO” to see if the founder will staywith the company assessment of the intertemporal affects of Founder-CEO after a professional-CEO is brought in. Future research succession. Do firms that replace their Founder-CEOs could benefit from exploring whether this feeling-out earlyhave a higher or lower rate of survival? Do firms process affects the timing of Founder-CEO succession. whose Founder-CEOs stayafter the succession event For instance, if the board believes that it is critical have higher or lower rates of survival? to keep the Founder-CEO around, but realizes that Finally, almost all of the factors examined in this the Founder-CEO will leave the companyif forced to studyare “prearrival factors” in the CEO succession. vacate the CEO position, it maydelaypushing for However, succession research has also started exploring the succession. Support for this proposition is provided “postarrival factors,” such as the actions taken bythe byFigure 2, which shows that the survival profile of new CEO and the “sources of power and influence” in Founder-CEOs who will staywith their companies dif- the organization after the successor’s arrival (Gordon and fers markedlyfrom the survival profile of Founder-CEOs Rosen 1981). Founder-CEO succession adds an inter- who will not staywith the companyafter the succes- esting element to this exploration. In contrast to larger sion event. If a Founder-CEO can crediblythreaten to companies, where deposed CEOs rarelyremain in an leave the companyif replaced bya professional CEO,
168 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession
Figure 2 Graph of Survival Probabilities by Whether Founder-CEO Continued at the Company After Succession Event
fstay = 0 fstay = 1
5.02058 By Categories of fstay -Ln[-Ln(Survival Probabilities)]
-1.35888 1.38629 5.04343 ln(analysis time) the Founder-CEO maybe able to retain the position for at the time that his companycompleted its first round of financing, a longer period of time. A competing-risk analysis of “We knew to the month when we would need our second round. It Founder-CEO succession, with the two outcomes being always depends a little on market conditions, but you have a pretty whether the deposed Founder-CEO stays or departs from good idea of when the next round will be coming up.” the firm, would illuminate whether the “all or nothing” 5Although it is possible for companies to progress through multiple threat of leaving the companyis able to affect the pro- CEO succession events, restricting attention to the first transition is cess of succession and delaythe succession event. appropriate if we suspect that the process of replacing the Founder- CEO differs from that of later successions (Allison 1984). Acknowledgments 6Because each companyin the data set has multiple company- The author would like to thank Nitin Nohria, Peter Marsden, David Ager, months of observations, I used the “robust” method of calculating the Tom Knox, Mariko Chang, Bill Simpson, the anonymous Organization variance-covariance matrix (Lin and Wei 1989), clustering the obser- Science reviewers, and Senior Editor Hayagreeva Rao for their comments vations bycompany. on an earlier version of this paper. 7When a round of financing is completed, both sides to the transac- tion (the companyand the investors) sign documents containing the Endnotes final terms of the round and often issue a press release to publicize 1This was epitomized bythe CEO whose exhortation to his engineer- completion of the financing milestone. ing team was, “We’re not done with first-generation development until 8To check the adequacyof the full Cox model, I tested the cen- a customer order says we’re done!” tral Cox model assumption about proportional hazards. I examined 2In addition to this field evidence, I also ran auxiliaryanalysesof my the Schoenfeld and scaled-Schoenfeld residuals (Schoenfeld 1982) on large-scale data set to check whether the honeymoon effect exists in this arena. I included in mymodels honeymoonperiods of both two both a “global” (full model) and “detailed” (covariate-specific) level, and four months, but did not find them significant at anystandard with the null hypothesis that proportionality holds. In short, at both level of statistical significance. levels of testing, the Cox model assumption that proportionalityholds 3For this reason, a recent studyemphasized the need for small- appears to be well justified at anystandard level of significance. In the companyresearch to use longitudinal data and quantitative controls detailed tests, all but one covariate had chi-squared statistics smaller (Fiet et al. 1997). than their degrees of freedom, so I could not reject the null hypothesis 4The Founder-CEOs I interviewed had a solid idea of when the next that proportionalityholds. (The one exception was Rounds-to-date, funding round would occur, and planned accordingly. One stated that with rho = 0 37269, chi2 = 1 07, and Prob > chi2 of 0.3011 with 1 d.f.
Organization Science/Vol. 14, No. 2, March–April 2003 169 NOAM WASSERMAN Founder-CEO Succession
However, even for this covariate, I could not reject the null hypoth- Bygrave, W. 1987. Syndicated investments by venture capital firms: esis.) For the “global” model as a whole, I also could not reject the A networking perspective. J. Bus. Venturing 2 141. null hypothesis, for the global test’s Prob > chi2 was 0.9974 (16 d.f.). Cannella, Jr., A., M. Lubatkin. 1993. Succession as a sociopolitical 9Perhaps the most surprising is the fact that prior years of work expe- process. Acad. Management J. 36 763–793. rience do not affect Founder-CEO succession in this dataset. However, Carland, J. W., F. Hoy, W. Boulton, J. A. C. Carland. 1984. Differen- the youthof the Internet industrymaymean that few people have tiating entrepreneurs from small business owners: A conceptual- anypreexisting industry-specificknowledge or skills to applyin their ization. Acad. Management Rev. 9 354–359. positions as CEO. Carroll, G. R. 1984. Dynamics of publisher succession in newspaper 10Using the hazard ratio of 0.093 from Table 2, doubling the number organizations. Admin. Sci. Quart. 29 93–113. of months in the shipping-product time clock changes the hazard of , E. Mosakowski. 1987. The career dynamics of self employ- succession bya multiple of 0.093 ln 2 = 0 193. ment. Admin. Sci. Quart. 32 570–589. 11Using the hazard ratio of 0.185 from Table 2, doubling the number Child, J. 1972. Organizational structure, environment and perfor- of months in the last-round time clock changes the hazard of succes- mance: The role of strategic choice. Sociology 6 1–22. sion bya multiple of 0.185 ln 2 = 0 310. Choi, Y. R., M. T. Shanley. 2000. A new venture’s honeymoon period: 12Using the hazard ratio of 2.220 from Table 2, doubling the amount Knowledge, resources, and real options reasoning. Proc. Acad. raised in the last round changes the hazard of succession bya multiple Management. of 2.220ln 2 = 1 738. Ciampa, D., M. D. Watkins. 1999. Right From the Start: Taking 13In the data set used in this paper, of the Founder-CEOs who were Charge in a New Leadership Role. Harvard Business School replaced, 23% took a position below the CEO, while 40% moved Press, Boston, MA. into the chairman role. Thirty-seven percent of the Founder-CEOs Cox, D. R. 1972. Regression models and life tables. J. Royal Statist. left the companyupon the arrival of a professional CEO. Another Soc. Series B 34 187–202. studyof high-growth firms found that of the Founder-CEOs who were Dalton, D. R., I. F. Kesner. 1983. Inside/outside succession and orga- replaced, 50% remained on the board of directors for at least five nizational size: The pragmatics of executive succession. Acad. years after the succession event, and a little less than 25% left their Management J. 26 736–742. companies completely(Rubenson and Gupta 1992). , . 1984. Organizational performance as an antecedent of inside/outside chief executive succession: An empirical assess- References ment. Acad. Management J. 28 749–762. Allen, M. P., S. K. Panian, R. E. Lotz. 1979. Managerial succession D’Aveni, R. A. 1994. Hypercompetition. Free Press, New York. and organizational performance: A recalcitrant problem revisited. Denis, D. J., D. K. Denis. 1995. Performance changes following top Admin. Sci. Quart. 24 167–180. management dismissals. J. Finance 50 1029–1057. Allison, P. D. 1984. Event History Analysis: Regression for Longitu- Dobrev, S. D., W. P. Barnett. 1999. Organizational roles and tran- dinal Event Data. Sage Publications, NewburyPark, CA. sitions to entrepreneurship. Stanford UniversityResearch Paper . 1995. Discrete-time methods for the analysis of event histo- No. 1587, Stanford, CA. ries. H.-P. Blossfeld, G. Rohwer, eds. Techniques of Event His- Drucker, P. F. 1954. The Practice of Management. Harper & Row, tory Modeling: New Approaches to Causal Analysis. Lawrence New York. Erlbaum Associates, Mahwah, NJ. Dyer, W. G. J. 1986. Cultural Change in Family Firms: Anticipat- Antle, R., A. Smith. 1985. Measuring executive compensation: Meth- ing and Managing Business and Family Transitions. Jossey-Bass, ods and an application. J. Accounting Res. 23 296–325. San Francisco, CA. Barnard, C. 1938. The Functions of the Executive. Harvard University Eisenhardt, K. M. 1989. Making fast strategic decisions in high- Press, Cambridge, MA. velocityenvironments. Acad. Management J. 32(3) 543–576. Fiet, J. O., L. W. Busenitz, D. D. Moesel, J. B. Barney. 1997. Comple- Barrett, L. 2000. A year to remember and forget. Red Herring mentarytheoretical perspectives on the dismissal of new venture (December 29). team members. J. Bus. Venturing 12 347–366. Becker, G. S. 1964. Human Capital: A Theoretical and Empirical Finkelstein, S. 1992. Power in top management teams: Dimensions, Analysis. Columbia UniversityPress, New York. measurement, and validation. Acad. Management J. 35 505–538. Berle, A., G. C. Means. 1932. The Modern Corporation and Public , D. C. Hambrick. 1989. Chief executive compensation: A study Property. Macmillan, New York. of the intersection of markets and political processes. Strategic Blossfeld, H.-P., G. Rohwer. 1995. Techniques of Event History Mod- Management J. 10 121–134. eling: New Approaches to Causal Analysis. Lawrence Erlbaum Fredrickson, J. W., D. C. Hambrick, S. Baumrin. 1988. A model of Assoc., Mahwah, NJ. CEO dismissal. Acad. Management Rev. 13(2) 255–270. Boecker, W. 1992. Power and managerial dismissal: Scapegoating at Friedman, S. D., H. Singh. 1989. CEO succession and stockholder the top. Admin. Sci. Quart. 37 400–421. reaction: The influence of organizational context and event con- Burton, M. D., R. Khurana. 2000. The routinization of charisma: tent. Acad. Management J. 32 718–744. Employment systems, succession, and firm survival. Working Furtado, E. P., V. Karan. 1990. Causes, consequences, and shareholder paper, Harvard Business School, Harvard University, Cambridge, wealth effects of management turnover: A review of the empirical MA. evidence. Financial Management 19(2) 60–75.
170 Organization Science/Vol. 14, No. 2, March–April 2003 NOAM WASSERMAN Founder-CEO Succession
Gamson, W. A., N. Scotch. 1964. Scapegoating in baseball. Amer. J. , W. Meckling. 1976. Theoryof the firm: Managerial behavior, Sociology 70 69–72. agencycosts and ownership structure. J. Financial Econom. 3 Glaser, B. G., A. L. Strauss. 1999. The Discovery of Grounded 305–360. Theory: Strategies for Qualitative Research. Aldine de Gruyter, Kanter, R. M. 1977. Men and Women of the Corporation. Basic New York. Books, New York. Gompers, P. 1995. Optimal investment, monitoring, and the staging Kesner, I. F., T. C. Sebora. 1994. Executive succession: Past, present, of venture capital. J. Finance 50 1461–1489. and future. J. Management 20(2) 327–372. , J. Lerner. 1999. The Venture Capital Cycle. MIT Press, Khurana, R., N. Nohria. 1999. The effects of CEO turnover in large Boston, MA. industrial corporations: A studyof the Fortune 200 from 1978- , . 2001. The Money of Invention. Harvard Business School 1993. Working paper, Harvard Business School, Harvard Univer- Press, Boston, MA. sity, Cambridge, MA. Goodstein, J., K. Gautam, W. Boeker. 1994. The effects of board size Kotin, J., M. Sharaf. 1967. Management succession and administrative and diversityon strategic change. Strategic Management J. 15 style. Psychiatry 30 237–248. 241–250. Lansberg, I. 1988. The succession conspiracy. Family Bus. Rev. 1(2) Gordon, G. W., N. A. Rosen. 1981. Critical factors in leadership suc- 245–255. cession. Organ. Behavior and Human Performance 27 227–254. Lawrence, P. R., J. W. Lorsch. 1967. Organization and Environment. Gould, W. 1999. Stata FAQ: How do I estimate a Cox model with a Harvard Business School Press, Boston, MA. continuouslytime-varyingparameter? www.stata.com. Levinson, H. 1971. Conflicts that plague familybusinesses. Harvard Gouldner, A. W. 1954. Patterns of Industrial Bureaucracy. The Free Bus. Rev. 49(2) 90–98. Press, Glencoe, IL. Lin, D. Y., L. J. Wei. 1989. The robust inference for the Cox propor- Granovetter, M. 1984. Small is bountiful: Labor markets and estab- tional hazards model. J. Amer. Statist. Association 84 1074–1078. lishment size. Amer. Sociological Rev. 49 323–334. Lorsch, J. 1989. Pawns or Potentates: The Reality of America’s Greiner, L. E. 1972. Evolution and revolution as organizations grow. Boards. Harvard Business School Press, Boston, MA. Harvard Bus. Rev. 50(4) 37–46. Mace, M. 1971. Directors: Myth and Reality. Division of Research, Grusky, O. 1961. Corporate size, bureaucratization, and managerial Graduate School of Business Administration, Harvard University, succession. Amer. J. Sociology 67 263–269. Cambridge, MA. . 1963. Managerial succession and organizational effectiveness. Main, B. G. M., C. A. O’Reilly, J. Wade. 1995. The CEO, board of Amer. J. Sociology 69 21–31. directors and executive compensation: Economic and psycholog- . 1969. Succession with an ally. Admin. Sci. Quart. 14 155–170. ical perspectives. Indust. and Corporate Change 4(2) 293–332. Hambrick, D. C., P. A. Mason. 1984. Upper echelon: The organization March, J. G. 1991. Exploration and exploitation in organizational as a reflection of its top managers. Acad. Management Rev. 9 learning. Organ. Sci. 2 71–87. 193–206. Marino, K., M. Dollinger. 1987. Top management succession in Handler, W. C. 1990. Succession in familyfirms: A mutual role entrepreneurial firms: Cases from the computer industry. J. Man- adjustment between entrepreneur and next-generation family agement Case Stud. 3 70–79. members. Entrepreneurship Theory and Practice (Fall) 37–51. McEachern, W. A. 1975. Managerial Control and Performance. , K. E. Kram. 1988. Succession in familyfirms: The problem of Heath, Lexington, MA. resistance. Family Bus. Rev. 1(4) 361–381. McGivern, C. 1978. The dynamics of management succession. Man- Hannan, M. T., J. Freeman. 1989. Organizational Ecology. Harvard agement Decision—U.K. 16 32–42. UniversityPress, Boston, MA. Meyer, J. W., B. Rowan. 1977. Institutionalized organizations: Formal Haveman, H. A. 1993. Ghosts of managers past: Managerial succes- structure as myth and ceremony. Amer. J. Sociology 83 340–363. sion and organizational mortality. Acad. Management J. 36(4) Mintzberg, H. 1983. Power In and Around Organizations. Prentice- 864–881. Hall, Englewood Cliffs, NJ. Helmich, D. 1977. Executive succession in the corporate organization: Mizruchi, M. S. 1983. Who controls whom? An examination of the A current integration. Acad. Management Rev. (2) 286–294. relation between management and boards of directors in large Hickson, D. J., C. R. Hinings, C. A. Lee, R. E. Schneck, J. M. corporations. Acad. Management Rev. 8 426–435. Pennings. 1971. A strategic contingencies theoryof intraorgani- Nohria, N. 1988. Creating new business ventures: Network organiza- zational power. Admin. Sci. Quart. 16 216–229. tion in market and corporate contexts. Unpublished Ph.D. disser- Hofer, C. W., R. Charan. 1984. The transition to professional man- tation, Massachusetts Institute of Technology, Boston, MA. agement: Mission impossible? Amer. J. Small Bus. 9(1) 1–11. O’Reilly, C. A., J. Chatman 1986. Organizational commitment and House, R. J., J. Singh, P. J. Tucker. 1985. Some causes and effects of psychological attachment: The effects of compliance, identifica- executive succession: An ecological perspective. Presented at the tion, and internalization on prosocial behavior. J. Appl. Psych. 71 Academy of Management Meeting, San Diego, CA. 492–499. Jensen, M. 1986. The takeover controversy: Analysis and evidence. Pfeffer, J., H. Leblebici. 1973. Executive recruitment and the develop- Corporate Finance J. 4(2) 6–32. ment of interfirm organizations. Admin. Sci. Quart. 18 449–461.
Organization Science/Vol. 14, No. 2, March–April 2003 171 NOAM WASSERMAN Founder-CEO Succession
, G. R. Salancik. 1978. The External Control of Organizations: A Tuma, N. B., M. T. Hannan, L. D. Groeneveld. 1979. Dynamic anal- Resource Dependency Perspective. Harper and Row, New York. ysis of event histories. Amer. J. Sociology 84 820–854. Porter, M. E. 1980. Competitive Strategy. Free Press, New York. Useem, M. 1993. Executive Defense: Shareholder Power and Corpo- Puffer, S., J. Weintrop. 1991. Corporate performance and CEO rate Reorganization. Harvard UniversityPress, Cambridge, MA. turnover: The role of performance expectations. Admin. Sci. Waldman, D. A., G. G. Ramirez, G. Gabriel, R. J. House, P. Puranam. Quart. 36 1–19. 2001. Does leadership matter? CEO leadership attributes and Reinganum, M. R. 1985. The effect of executive succession on stock- profitabilityunder conditions of perceived environmental uncer- holder wealth. Admin. Sci. Quart. 30(1) 46–60. tainty. Acad. Management J. 44(1) 134–143. Robinson, R. J., M. V. Osnabrugge. 2000. Angel Investing: Match- ing Startup Funds with Startup Companies. Jossey-Bass, San Wasserman, N. 1999. Venture capital negotiations: VC versus Francisco, CA. entrepreneur. Harvard Business School Note #800-170. Harvard University, Cambridge, MA. Rubenson, G. C., A. K. Gupta. 1992. Replacing the founder: Explod- ing the myth of the entrepreneur’s disease. Bus. Horizons 35(6) . 2001. Inside the black box of entrepreneurial incentives. 53–57. Working paper, Harvard Business School, Harvard University, , . 1996. The initial succession: A contingencymodel Cambridge, MA. of founder tenure. Entrepreneurship Theory and Practice 21(2) . 2002. The venture capitalist as entrepreneur: Characteristics 21–35. and dynamics within VC firms. Unpublished Ph.D. thesis, Har- Sahlman, W. A., H. H. Stevenson, M. J. Roberts, A. Bhide. 1999. The vard University, Boston, MA. Entrepreneurial Venture. Harvard Business School Press, Boston, , N. Nohria, B. Anand. 2001. When does leadership matter? The MA. contingent opportunities view of CEO leadership. Working paper, Salancik, G. R., J. Pfeffer. 1977. Who gets power—and how theyhold Harvard Business School, Cambridge, MA. on to it: A strategic-contingencymodel of power. Organ. Dynam. 5 373–395. Weber, M. 1946 trans. From Max Weber: Essays in Sociology. Oxford , . 1980. Effects of ownership and performance on executive UniversityPress, New York. tenure in corporations. Acad. Management J. 23 653–664. Weick, K. 1969. The Social Psychology of Organizing. Addison- Schein, E. 1985. Organizational Culture and Leadership. Jossey-Bass Wesley, Reading, MA. Inc., San Francisco, CA. Wiersema, M. F. 1992. Strategic consequences of executive succession Schoenfeld, D. 1982. Partial residuals for the proportional hazards within diversified firms. J. Management Stud. 29 73–94. regression model. Biometrika 69 239–241. Woodward, J. 1965. Industrial Organization: Theory and Practice. Sonnenfeld, J. 1986. Heroes in collision: Chief executive retirement Oxford UniversityPress, New York. and the parade of future leaders. Human Resource Management 25 305–333. Zajac, E. J. 1990. CEO selection, succession, compensation and firm performance: A theoretical integration and empirical analysis. Tashakori, M. 1980. Management Succession: From the Owner- Founder to the Professional President. Praeger, New York. Strategic Management J. 11 217–230. Thompson, J. D. 1967. Organizations in Action. McGraw-Hill, Zaleznik, A., M. F. R. Kets de Vries. 1975. Power and the Corporate New York. Mind. Houghton Mifflin, Boston, MA.
172 Organization Science/Vol. 14, No. 2, March–April 2003