Paradigm of Turnover in the Dual Paths of CEOs CareerLiability of Newness, Oldness an Honeymoon

Paradigm of Turnover in the Dual Paths of CEOs CareerLiability of Newness, Oldness an Honeymoon Paradigm of Turnover in the Dual Paths of CEOs Career: Liability of Newness, Oldness an Honeymoon Zhaojun Gao (Peking University, Strategy Department. ) ABSTRACT: This study develops a paradigm of CEOs turnover in which both the influences of origin and succession types to the likelihood of dismissa are incorporated. By applying the “Red Queen” known in evolutionary theory into CEOs research, our study highlights the internal competition within top management teams. Drawing on the unique culture background of China known as the internecine struggle, we propose that inside CEOs both as the original and follower types increase the likelihood of dismissals. The finding of dual paths of CEOs? career facilitates our analysis. We also examine the liability of newness and oldness in CEOs? career. Our study provides evidence for the existence of the honeymoon between CEOs and boards. 1. INTRODUCTION Asian history is „for the most part, really unhistorical, for it is only the repetition of the same majestic ruin?, and: „China ... lie(s), as it was still outside the World?s History?. 1Hegel “The disastrous events of internecine struggle that followed are well known. The wounds caused by those events are so deep, a generation must pass for them to heal, ...” Ch?ao-Ting Chi. (1934: 9) China has a long history of feudalism, at least fifteen-thousand years of exchanging different dictators. The history of China is a record of internecine struggles through which the power and rights were transferred. The special history and circumestance foster the unique culture that draws many scholars? interest in the context of management research. Theories of organizational politics highlight interest conflicts and power struggles within organizations, especially among senior executives (Lazear & Rosen, 1981; Pfeffer, 1981). For example, Fama (1980: 293-294) proposed that competition among top managers plays an important role in the internal monitoring of firms. Ocasio (1994), drawing on Pateto?s (1968) circulation of power theory, reported that the likelihood of CEO turnover among companies with low performance increased with the proportion of inside directors. Wei Shen following a power perspective studied the antecedents of CEO dismissal followed by inside succession and suggested that non-CEO senior executives frequently play an important role in CEO dismissal (Wei Shen, 2002). Further, little research has directly focused on power contests within top management as an important cause of CEO dismissal (for the exception see Wei Shen, 2002). Most importantly, all the previous research negelict that the process of selecting CEOs is contigent with mechnism of firms? govervance. Some literatures (Wei Shen) do examine the infulence of the original and follower types as susession to the likelihood of dismissal. Their studies had the unavoidable limitation that put themselves out of the whole logical understanding about CEOs turnover. This shortage appeared apparently especially they focused on the CEOs succession which is the most important part of turnover. More details will be seen in the proceeding empirical part. The turenovers of CEO involve rountines, path dependence and organizational culture. Certain firm perfers to and follows very pattern of CEO selection both in the dismissal and sussession. To advance understanding of this paradigm, we develop the two related concepts: family and professional patterns. 1 Quoted in S. Avineri (ed) (1969) Karl Marx on Colonialism and Modernization (NewYork: Doubleday & Co), pp. 11–12. 1 / 16 2The “Red Queen” (Van Valen, 1973) is well known in evolutionary theory (Van Valen, 1973). We think that it is central, driving force behind the success and failure of CEO careers. To anticipate our conclusion, we find support for the idea that internal competition increases the likelihood of dismissal. The consideration here of both the family pattern and professinal pattern is unique in the CEO turnover literature, but facilitates the application of the “Red Queen” to internal competition within top management teams. Human capital and network drew more attentions to the field of mangemant research. From that perspective, this paper examines the age effects on the dual paths and finds that the effect of age, when increasing, depresses the baseline hazard rate. In other words, more younger the CEO more higer dismissal rate will . But the zazard rates increase after reaching the lowest point. So, more accurately speaking there is a U-shaped relationship between CEOs age and the correspondent dismissing rates. That trigers us thinking the underlying story that CEOs also endure a liability of oldeness in their careers beside the liabiltity of newness. Some researches had mentioned that manageners might have a period of honey time which was called the „honeymonth? in their careers (Wei Shen, 2002). Literatures in organizational research also highlight the existence of adolescence when covering the organizational mortality research. But few study had examined the existence of honeymoon or adolescence in CEOs? career. Given these significant limitations of past research, we think the likely idea that there exists „houneymoon? in CEO careers merits further testing. At the same time, the limitation of measuring CEO tenure motives our study. Some researches (Wei Shen, 2002) focus their study and constraints on the availability of financial information and divide the CEO tenure into fiscal years. But fiscal year may be a relatively coarse meansure, despite its frequent use as a metric in organizational research. We assess more accurate imformation about CEOs? tenure when the career terminated. That facilitates our research. Such a finer measure would be especially useful to test for the existence of the „honeymoon? between a new CEO and a board that is often discussed in the business press (Wei Shen, 2002). To demonstrate the boundaries of our study is significant important. The concept of dismissal in our research is a more broaden and universal one, which involves not only really dismissed, also CEOs that terminated their career before deadline of contracts. There are some reasons for the abnormal termiantion. They might come from the governance structure changes that drive those CEOs leaving their postions. So, we are more interested in the unexpected termination of CEO careers than the solely dismassial reported by the firms. The relationships and the underlying processes we explored in this study thus differ importantly from those that figure in the extant literature on CEO succession in general. 2. THEORY AND HYPOTHESES 2.1 Theoretical Framwork In a qualitative study, Gabarro (1987) suggested that the process of CEOs? establishing authority in office generally takes from 30 to 48 months. So it asways takes 2.5 to 4 years for a CEO to gain adiquate power and trust from other memebers of management team. Wei Shen (2002) selected five years of tenure as the cutoff in their coding when design the variable: early years of tenure. Drawing on their studies and our preliminary analysis, we distinguish two pathes of CEO career as can be shown in the following Graph 1. 2 Van Valen (1973) coined the term Red Gueen when introducing this idea to the study of biological evolution. He was referring to Lewis Carroll?s Through the looking Glass, where Alice notices that she appears to be stationary even through she is running a race. The Red Queen?s response is that Alice must be from a slow world, since in a fast world one must run just to stay still. 2 / 16 Smoothed hazard estimate .12.14.16.18.1 051015analysis time Graph 1 Our explortary analysis demenstrates the finding consistent with those above mentioned theoretical and empirical studies. The smoothed hazard estimate appears two different patterns. We must emphasize that the honrizantal line represents the duration which is caculated by the tenure before the CEO is dismissed. In our sample the longest tenure of CEO is about seventeen years. Initally the hazard rate is increasing and after it reaches a point then decreases to the lower point at which the tenure is about five years. After that, the hazard estimate almost 3monotonically increases to the highest peak and decreases. So, the interesing finding is that there exists dual paths of CEO career: one is less than five year, the other is more than ten years. Both careers of CEOs in each path experences its peak of hazard rate. It indicates that the hazard rate of dismissals is monotonically increasing to its peak then decreasing in both pathes of the careers. 2.2 New concepts Family and Professional Patterns In order to better inllustrate the whole processes of CEOs? turnover, we develop two new concepts, family pattern and professional pattern, in our research. This point had been negelected in previous research (See Wei Shen, Academic Management Journal，2000). They mainly focused on the effects of original types when followed by either outsiders or indiders. But for faciliating their analysis, the followers and were seperated into two datasets: one for outsiders and one for insiders. With this procedure, those authors lost the comparison between diffirent types of followers. They can only compared them based on the results coming from the datasets they splited. In that case, the comparison shows little statistical power only because the it happened between two seperated subsumples. The conceptualization of the following two patterns gain the advantages not only from a promoted usnderstanding of CEOs? turnovers which is taken as a rountine, but also from the statistical facilitation of comparision. So, the family pattern and professional pattern are introduced here to compensate the limitation exiting in previous research. The family pattern is coded 1, if origin as insider and also followed by an insider; and 0, when it is followed by an outsider. So positive family patterns mean the firms are more likely to prefer insider CEOs both in orgin and follow types. Professional pattern is coded 1, if origin as outsider and also followed by an outsider; and 0 if followed by an insider. So positive professional pattern indicates that firms are more likely to prefer CEOs coming from outside. The two terms, family pattern and professional pattern we introduced here can better interpretate the paragram of CEOs? turnover. They bring us more precisely understanding of CEOs? types both from the origin and succession. Our research overcomes the limitation in precious literatures by conceptulizing a theoretical diagam which can be better used to understand the whole process and mechanism of CEOs turnovers. 3 As we can see in the graph 1, before the hazard rate reaches the highest point it experences a relative high peak which is ignored in our analysis because of its narrow rage coved and inconsiderable influence to the hazard rate in the dual paths of caeer. 3 / 16 Followers: Insider 1 Outsider 0 Original: (1, 0) Insider 1 Cell 2 (0, 1) Outsider 0 Cell 3 Path 2: More than ten years Path 1: Less than five years Graph 2 With the above diagam, the underling story of dismissals is uncovered. Although most research focuses on the dismassal of CEOs, they only partially touched the whole changing process in CEOs? careers. For instance, the original types of CEO are examined by coding as the dummy variable and the followers are also coded as a dummy when being involved. When the researchers compared the effects of original types with different followers on the dismissal rates, splitting the sample into two separate parts: one for inside followers and one for outside followers, is always performed by them(Wei Shen, 2002). We believe that the above commenly adopted meathod can not capture a clear picture of the changing process of CEOs and just a snapshot of the dynamic events. So, aiming at solving this probem we design the following patterns which depict the whole process of CEOs turnovers. Family patten and professinal patters are two new concepts in our analysis. The conceptulized concepts make our analysis different and advance to previous research. The logic can be better understood with the help the above graph 2. The triangle represents family pattern and professinal pattern. If the original and following types are all the insider, the pattern is the so-called family patten indicating that the firm is more likely to choose insiders as CEOs both as orignin and follower. And its hyprid pattern means that the CEO came into his/her position as insider but was followed by an outsider. In graph 2, the cell at the end of the first row represents the relative event happended in CEO?s turnover. The same terms are designed for the pair of concepts: professinal and its hybrid pattern. When origin as ourside CEO who is dissmissed is followed by outside successor, the pattern of the firm is called the professial pattern which indicates that the firm is more likely to select outside CEOs. We call it professinal patter because most of the outsiders are professinal managers who are renderring their services in the open markets. By the same token, if an origin as outsider is followed by an insider successor it will be the hybrid pattern coded 0. Actually, we are trying to generate some variables capturing the charateristics of firms in their managerial mechanisms. The family and professinal patterns we developed in fact are the reflection of firms? unique favorate or routine in their managerial fields. By carrying out the above concepts, the research question is put into a more solid analytical foundation on which we can conduct our analysis. At this point, our reseach is also making the contribution to the understanding of top managers careers both from the dismissal and sucession perspectives. Fama (1980) maintain viewpoint that a top manager is disciplined by competition from other managers. The main two restrictions are manager reputation and efficiency wage mechanism. In the competitive market, wages that manager earn are based on his/her past performance because a new employer has no new information about his/her ability (Jiangguo Y. etal., 2008). Conveying efficiently the ability information to the labor market makes managers working hardly in pursuing a higher wage in their careers. Therefore, market competition plays an important role on managerial behavior. However, the management market in China has some special characteristics. There are more 4internal competitions between the same management team. Within a same management team, senior offiers are 4 The author had been working in a State-Owned Enterprise for about 15 years from 1995 to 2009 and has the deep understanding 4 / 16 competing for promoting to a higher level of management. Especially, the professinal market for CEOs in China is relatively immuture comparing with that in West countries so that the mechanism of professional CEO markets cannot function properly. In a research of the effects of manager compensation and market competition on financial fraud in pubilc companies, the authors also find the inefficency of mechanism in China professional CEOs? market. Their results suggest that underdeveloped manager market can't discipline Chinese top managers efficiently. Hence the mechanisms of reputation and efficiency-wages, which tend to reduce the incidence of fraud, do not function properly in China. They do not provide an incentive for managers to prevent fraud in the firms (Jiangguo Y. etal., 2008). The mixture of immaturity of professional market and fierce competition internal companies labeled as the cultural production of the internecine struggles characterizes the essence of CEOs turnovers. To make our analysis consistent with previous research, we also conduct the same prosedure with Wei Shen (2002). For years, reseachers in strategy and organizations have been intrigued by various aspects of the Red Gueen (Barneit, 1996). Their models allow for competitive interaction among organizations that are limited in their ability to strategize. Reseach about organizations shows that an organizaiton facing competition is likely to respond, but that its response is likely to be limited, i.e., merely „satisficing ? through a localized search and decision process (March and Simaon , 1958). In this paper we argue that the same logic can be equally applied to the competition between top managers, especially the chief exacutive officers. Family pattern is termed here to represent the fierce competion within the top management team. When family pattern is favored by a certain firm, it can be treated as a simble of „arms races? (Dawkins, 1982). Looking into the family pattern process, we might easily accept that it is self-reinfocing, and probably is an important stimulus to managers turnovers. From an evolutionary perspective such self-reinforcing processes can result in important long-term developments even through the effect of each CEO change may be small. „Incremental? change usually is thought to mean small change, but each small change within the Red Queen triggers the next, accumulating over time into a potentially large evolutionary difference (Barneit, et al., 1996). When the family pattern processed by a firm, it will be more likely to increase the likelihood of dismissal because the so called Red Queen and „arms races? effects.Armed with the above logic, we posit the following hypothesis. Hypothesis 1. 5Familiy pattern compared with its hybrid pattern increases the hazard rate. By the same token, when professional pattern was chosen, the likelihood of dismissal of CEO will be relatively lower than that of firms choosing family pattern, only because outside CEOs are keeping themselves far from the internal „arms races? competition. So, we posit the following hypothesis. Hypothesis 2. The professional pattern decreases the hazard rate with comparision to its hybrid pattern. 2.2 The Liability of Newness and Oldness Two famous papers in studying the organizational mortality are worth mentioning here. One is about liability of newness (John Freeman & Carroll & Hannan, 1983); and the other one is about liability of adolescence (Josef Bruderl & Rudolf Schussler, 1990). These research takes an organization as human-beings. But as far as we know that few literatues that fucuse on executives, executive careers and leadership had examined the liability of newness and adolescence. There is not a vague relationship between CEOs age and dismissal risks. CEOs with less managerial experinences and connections in social networks are more likely to be exposed to higher risk of dismissing. Accumulating knowledge and experiences is a time consuming process. And most of the magament skills is built through practices through which tacit knowledge is transferred. It seems quit reasonable that with one? age increasing the lower likelihood of dismissing is expected. While with age increasing, the CEO might be lagged behind the wave of social development only because of the inability of catching up newness of invovation. Acctually, scholars (Timothy G. Habbershon, Mattias Nordqvist and Thomas M. Zellwege) see entrepreneurial performance not as a manifestation of the firm?s need to overcome liability of newness, but to overcome „liability of 6 oldness.?After the lowest hazard rate was researched, it will be increasing with CEO becoming more older. When and experiences of the so-called internecine struggle internal the firms. At the same time, interviews with managers also strength our belief of the fierce internal competition within groups of arm-length connected people. 5 We compare the effect of the family pattern to hazard rate of dismissal with its hybrid pattern. The same comparison happened to the professional pattern and its hybrid pattern. Here we don?t pay the attention to the different effects brought out by comparing the family pattern and professional pattern, although we can do that easily. The reason is that many firms don?t make the decision between the pure patterns, instead of that between one of the pure patterns and its hybrid pattern. 6 The liability of oldness is firstly defined as the liability faced by established firms to keep up with changes in their environmental 5 / 16 investigating the entrepreneurial opportunites and ethical dilemmas presented by technological turbulence, Hall et al. (Jeremy Hall & Philip Rosson. 2006) find that unlike new entrants, incumbents must abide by the currently accepted norms, and thus suffer from “liabilities of oldness”. The human capital involves the CEO?s industry experience (Wei Shen, 2002), education bankgrounds and social networks. The age to some extent can reflect the experiences of CEO. Usually, an elder CEO has a longer time of cumulating management knowledge and embeded in the network of management. Of cause, the younger gains more experences than elder if he/she entered the CEO career earlier than the elder. But their management skills and experience increase with their practices and accumulated with their age. So, we choose the CEOs age as the measurement of variables that infulence the hazard rate of dissmissal. A CEO?s competence can rise as time goes by when he/she acquires more experience and more wisdom (Fan & Lan; ). For example Murphy and Zimmerman (Murphy& Zimmerman,1993) find a strong relation between CEO turnover and CEO age. However, the rise of competence will eventually take a downturn as the CEO ages and the firm will find it beneficial to remove the incumbent CEO if the firm starts to doubt the ability of CEO, especially coupled with poor performance. In addition, anticipating the difficulty in replacing aged CEOs, shareholders may put a retirement policy in place in order to remove aged and incompetent CEOs (J. W. Fredrickson & D. C. Hambrick, et al. 1988). That is to say, initially CEOs age will lower the likelihood of dismissal in the dual pathes of career, but it will increase the likelihood of dismissal in the later stage of CEOs? career. So, we have the following hypothesis. Hypothesis 3. A U-shaped relationship between CEOs age and the likelihood of dismissing rates can be expected. 2.3 The Honeymoon Does the honeymoon period exist in the dual paths of CEOs careers? If it exists, how long it will be lasted? New CEOs confront significant chanllenges posed by the demands of new positions. Given this situation, it is important for new CEOs to have a “honeymoon period,” during which they have the latitude to acquire needed task knowledge and skills without worrying about being dismissed because of mistakes in their learning process (Wei Shen, 2003). Although lots of scholars hold that truth to be self-evident that CEOs start their jobs with such a honeymoon period, the reality is always the opposite: they begin with a period of high vulnerability (Fredrickson et al., 1988; Shen & Cannella, 2002). Many CEOs lose their jobs in less than three years, which is barely enough time for them to complete the process of taking charge (Gabarro, 1987). The risk of dismissal is likely to especialy high for CEOs who are appointed under conditions of financial stress and who are expected to turn around firm performance quickly (Hambrick & D?Aveni, 1992). In the business press, the existence of the “honeymoon” between a new CEO and a board draw many attentions but never been examied (Wei Shen, 2002). In our study, we mainly test our hypotheses about honeymoon through the methodological approach. The hypothesis on the honeymoon states that risks rise for an initial period of a CEO tenure?s existence and decline afterwards. The decline of the dismissal risk is assumed for the same reasons that led to the liability of newness hypothesis. If the so-called honeymoon does exist, we can expect the curve of hazard rates has an increasing period before reaching the peaks, which is assumed the honeymood between CEOs and boards. Hypothesis 4. There is an invert U-shaped relationship between CEOs? tenure and dismissal rate in the dual career paths. 3. METHODS 3.1 Sample The data source is from the famous GTA Information Technology Company who provides comprehensively relative data about all the listed firms in China. Govenance Structure Dataset is suitable for collecting data of CEOs? turnovers. The tureovers of top managers including directors of board and chief executive officers are reported each yea for 1,347 companies. For the purpose of our study, we delete the information about the board directors and only make the CEOs observations remained in our final sample. And in our sample, the reasons why a very tenure was ended is all precisely termed so that we can identify whether a perticular tenure is terminated normally or not. After deleting missing values, 3996 observations remained in our dataset. 3.2 The Defination of Dismisal We define the dismiss of CEO by a more broaden concept. If the tenure of a CEO is terminated before the end and organizational setting. Here we apply the concept to CEOs tenure, which means the liability faced by an CEO to keep up with changes in the environmental circumstances and situations where he/she are taking in charge. 6 / 16 of contract, it is defined as a dismissal case. Such observations involve the dismissal reported in the annual news, the depature because of the managerial structure change, etc. In other words, the depature because of the sharehold 7changes is deemed as dismissal. We mainly focused on the abnormal termination of CEO tenure. Including those cases make the sample can be better serverd for our purpose. Actually, based on the Chinese management circusmtances, they cannot be negelected and are the essence of our research. 3.3 Predictor variables Family Pattern. If a CEO origin as insider and also, when dismissed, is followed by insider, it is called the family pattern which is coded 1. In graph 2, the cell 1 represents the conrespond cases. And if the insider CEO is followed by an outsider succession, it is called the hybrid pattern which is coded 0. The cell 2 in the graph 2 represents the relevent cases. Professional Pattern. If a CEO origin as outsider and is followed also by an outsider after he/she was dismissed, it is called the professional pattern which is coded 1. And it hybrid pattern is the one that original outsider CEO followed by insider succession, which is coded 0. The cell 3 and 4 in graph 2 are the conresponded representions. Insider CEO To keep consistent with previous research, we also conduct the same procedure with other scholars like WeiShen (2002) in which we did not employ the concepts of family and professional pattern. And also at that step we didn?t distingush the dual pathes careers. We just want to make the comparsion between our procedure and that what the previous research had adopted. CEO origin as insider is coded 1, and outside 0. Dismissal Age Dismissal age is measured when the CEO leaving his/her position. And we also include the squared age as a comparision in our analysis. In human capital literatures, the age is widely accepted as one of the most important measurement among the education backgrounds and social network variables such as the embededness and centrality and so forth. Control variables Early years of tenure We create a dummy variable called the early years of tenure which is coded 1 if a CEO? tenure is less than 5 years, and 0 if lager than five years. Our method is consistent with that adopted by Wei Shen (2002) and applicable with what we had found in the previous exporing analysis. Also it is supported by a qualititive study in which Gabarro (1987) suggested that the process of CEOs? establishing authority in office generally takes from 30 to 48 months. Duality Research had found that the duality significant infulences the possiability of dismissing. We control the effect of duality in our model in which it takes value one if the CEO is also the director of board, and zero otherwise. Agent We also control the agent situation in the early stage of CEOs? career. It is coded 1 if the CEO is an agent and 0 if not. Concorrent Social network structure infulences the CEO?s embededness in the top management team. Many CEOs take another position in the abranch of the parent company. This will excert the infulences on its authority of management which in turn effect the likelihood of dismissal. We control the concorrent infulences in our model in which the variable concorrent is coded 1 if the CEO is holding another position in the company and 0 otherwise. 3.4 Analytical Methodology For the necessity of considerring longitudinal studies in CEO?s turnovers research, we selected a continous-time event history analysis to test our hypotheses. Event history analysis is an established methodology for analyzing dynamic process when the outcome of that process is a discrete event (Tuma & Hannan, 1984). We conducted our analysis using the software pakeage Stata 10, which provides maximum likelihood estimation of both monotonic and nonmonotonic parametric models of duration dependence (Stata, 1999). Choosing the appropriate parametric model for the data in event history analysis is of critical important (Yamaguchi, 1991). We started with the exponential model for its ability and flexibilty of analizing the data in different periods (Adding reference here). We define CEO dismissal (failure in the data) as occuring when the tenure was terminated obnormally 7 The dismissal cases include the following six main types. The tenure of CEO was terminated by the reasons of: (1) the change of control of firms; (2) personal resign; (3) dismissal by firms; (4) personal reasons; (5) improving the corporate body governance structure; (6) law suits. 7 / 16 compared to the retirement ect.. When obnormal termination didn?t happen in a CEO?s career, it was treated as right censored (Hannan and Freeman, 1988). We modeled dismissal failure using r(t), the instantaneous rate of failing. This hazard rate of failure is defined as the limiting probability of a failure between t and t+?t, given that the repective CEO was operating at t, calcultated over ?t: 错误～未找到引用源。 (1) Parametric estimates of the hazard rate require assumptions about the effect of time (in our models, the years of tenure)on failure. The exponential model are more flexiable and allow us to see the different effects of variable on the hazard rate in the dual paths of CEO careers. Therefore, we used a piecewise exponential model which allows the rate of failure to vary in an unconstrained way over preselected time ranges (Tuma and Hannan, 1984; Barron, West, and Hannan, 1994). In our models, the time range is divided at k points (α,α, ..., α), which with α = ?, 12kk+1 creates k time periods; I = {t | α?t?α}, l=1, ..., k. The baseline failure rates (constants) are estimated for each time lll+1 period. Thus, the piecewise exponential model we estimate is of the form: (2) 错误～未找到引用源。 Where X represents the vector of covariates, β the associated vector of coefficents and ais a constant l coefficient associated with the lth time period. The reported results are maximum-likelihood estimates obtained using the statistical package Stata 10. Our analysis is preceded with Gompertz model to examine hypotheses about liabilities about newness and honeymoon or adolescence because the index shows that the Gompertz model is more better for exponentioal model. The survival and density functions for the three-parameter generalized gamma model are 错误～未找到引用源。 if k ? 0; or 错误～未找到引用源。 if k= 0 (3) where I(k, σ) is the incomplete gamma function and z = [ln(t) - λ]/σ. This model is implemented by parameterizing 错误～未找到引用源。 and treating the k and σ as ancillary parameters to be estimated from the data (Stata, 1999: 440). The hazard function of the generalized gamma distribution is extremely flexible, allowing for a large number of possible shapes, including as special cases the Weibull distribution when k =1, the exponential distribution when k = 1 and σ =1, and the log-normal distribution when k=0. We then estimated other functional forms, including Weibull, log-normal, and log-logistic models. Our analysis suggested that the generalized gamma model provided the best fit and had the largest log-likelihood with the smallest value on the Akaike information criterion (AIC; Akaike, 1974). It has been consistently demonstrated in previous research that CEO dismissal is frenquently the outcome of a power struggle between a CEO and other members within the top management team such as outside directors (Boeker, 1992; Boeker & Goodstein, 1993; Cannella & Lubatkin, 1993). To test our theory and convincingly demonstrate that turnover patterns as a refection of firm rountine simultaneously infulence CEO dismissal and subsequent inside and outsie succession, we conducted our analysis not only on the paradigm patterns, but also on CEO dismissal followed by both inside and outside succession without examing the dismissal in the dual paths of career. The first reason of performing this steps is that we want to show the difference between the anaysis of us and that used by others. The second reason is to find further implication of the infulence and extent to which the internal competiton has. 4. RESULTS Table 1 reports variable means, standard deviations, and Pearson correlation for all the observations in our sample (all CEO-years). Table 2, 3 and 4 present the results from the event history analysis. Table 2 is designed for our hypotheses about the family and professinal patterns. The exponential model is employed here with the varibles 8 / 16 like age, age2 and duality as controls. We proposed that family pattern would increase the likelihood of CEO dismissal. In model 1 and 5 of table 2 each of the coefficents of the family pattern displays a positive and significant impact. While for the professional patterns, we hypothesized a negotive impact on the hazard rates. The results surport our hypotheses both for the family and professional pattern. Model 1 and 2 present the results in the five-years path and the model 5 and 6 is for the longer career path which is about more than ten years. In the dual pathes of careers our hypotheses are strongly surported. We also conduct the same procedure of previous research (Wei Shen, 2002) and get the results reported by table 3. The purpose of doing that is to compare the results and show the advantages of our procedures. Without considerring the family and professional patterns, the results can only reflect the influence of original types on dismissal rates when the CEO is followed by different successions. As we already mentioned above, the separation of the sample into two independent subsamples: one for inside followers and one for outside followers makes the results lost the comparison. But the results here also bear a deep implication. Insider CEO increases the likelihood of dismissal both for the inside and outside successions. It is a quiet different story about CEOs careers in Chinese context. --------------------------------------------- Insert Table 1 Here --------------------------------------------- --------------------------------------------- Insert Table 2 Here --------------------------------------------- --------------------------------------------- Insert Table 3 Here --------------------------------------------- So, the conclusion here is that the familiy pattern compared with its hybrid pattern increases the hazard rate, while the professional pattern decreases the hazard rate. The impacts of CEO age can be got a better understanding from the graph 3 in which the survial rate of elders are much higher than yongers. Then less likelihood of dismissal for an older CEO. The results of Table 4 from Gampertz models surport our assumption. In the both pathes of careers, CEO age is negotively related to the hazard rates, which means that with the age increasing the hazard rates of dismissal decreases. That is to say, the age will lower the baseline hazard model and as they are depicted in the graph 4 and 5. So, the yonger the CEO the more likehood of dismissal. The hypothesis 3 about the liability of newness is surported. 9 / 16 Kaplan-Meier survival estimates 0.000.501.000.250.7505101520analysis time agef = 1agef = 2agef = 3agef = 4 Graph 3 In Graph 6 we combine the hazard rates in the dual pathes of CEO careers together. It clearly show there are two peaks in both of the path of CEO careers. We caculate the peaks which measures the highest hazard rates. In the five-year career path, CEO reaches the peak after working for about 2.2 years. And in the more-than-ten-year career path, the peak comes at 7.2 years. So, in the both pathes, the hazard rates first increases and reaches its peak and then drops down monotonically. That is to say, CEOs has a certain time period in which they could mentain their positions. This kind of periods might be called as the widely accepted concept adolescence in CEO careers. Also it is the honeymoon between CEOs and boards. Although the dual paths of CEOs career provide different periods of honeymoon, they do foster a certain periods of time in which CEOs enjoy their less dismissing rates. --------------------------------------------- Insert Table 4 Here --------------------------------------------- .5.6.7.8.9 012345t hthtage 10 / 16 Graph 4 .6.7.8.9 05101520t hthtage Graph 5 .6.7.8.9 05101520 t2 htage1htage2 Graph 6 Among the control varibles, the early years of tenure and the duality show the significant impacts. For the early years of tenure coded 1 if the tenure is less than five years, the effects are positive and significant. That means that CEO faces high dismissal whem the career is shorter, which is consistent with our ordinary understanding that longer working solid the position of CEO. Duality also show the positive influence on the dismissal rate. 5. DISCUSSION By taking the CEOs? turnover as a routine-based behavor of firms, this study helps advance understanding of the underlying mechanisms of CEO turnovers. The Red Gueen theory provide a profound meaning of internal competition of top management teams. This research examines those main aspects by introducing two main new concepts: family pattern and professional pattern. Our findings show that family pattern increases the likelihood of CEO dismissal while professional pattern decreases the hazard rates. China is a special contry that has a long time of feudal society. The culture exerts influences on the management in which the internal competition plays an important role. The resulst shown that in Chinese management context, CEOs origin as an insider increases the dismissal hazard rate which is contradictory to what other scholars had found in their research. Those scholars suggest that CEO origin as an outsider increases the likelihood of CEO dismissal followed by inside succession in investigaing 387 large U.S. corporations (Wei Shen, 11 / 16 2002). The inconsistent foundings indicate that the signidicant differences between Chinese management and the United States. As we know that China has a long history of feudalism in which citizens in the large area battled and rivaled with each other. The history is a remark of internal competition of Chinese. Our findings provide evidences for many literatures in history and nolvelty tell the same story that underlining the fierce internal competition reflected by the unique Chinese culture. The dual pathes of careers are another interesting finding in the research, which is consistent with previous study. By applying the liability of newness and adolescence directly into CEO career study, our paper also contribute to the management literatures. Our results surport the hypotheses about the liabilities. The two control variables provide additional clues about the turnover dynamics at the tops of firms. Limitations and Directions for Future Research Although this study advances understanding of the dynymic turnovers of CEOs, but it has limitations. First, unlike many other researches taking a power perspective by examing the influences of variables such as the Non-CEO executive ownership or the proportion of non-CEO inside directors, our research focuses on the dynymic process of turnovers. We are more interested in the dynamic process itself. But omitting some predictable varialbes with the characteristics of firms and top management teams might make our research yeild limitations. Performance is one of the more important variable and team heterogeinity might be the team-level variables that should be involved in our future advanced research. Second, we have the hypotheses both about the liability of newness and adolescences and our results surport all the propostions. But our findings just indicate the liablitites do exist in the dual pathes of CEO careers. The underlying mechanisms haven?t been carefully examined. At the same time, there might be some interactions between the liabities and dynamic turnover processes. We find that the „honeymoon? frequently mentioned in the business press does exit in the dual paths of CEOs? careers. That orients the future research in management fields. For instance, there must be some linkages between the honeymoon and capability of CEOs. A relatively longer honeymoon might result and indicate higher capability of a CEO. This research provide a road but it is limited and should be more explored by future studies. REFERENCES [1]Barron, D. N., E. West and M. T. Hannan (1994). A time to grow and a time to die: Growth and mortality time to grow and a time to die: Growth and mortality of credit unions in New York, 1914-1990, American Journal of Sociology, 100, pp. 381-421. [2]Ch?ao-Ting Chi. (1934: 9). The Economic Basis of Unity and Division in Chinese History [3]Fama, E., Michael C. Jensen. (1983).Separation of ownership and control. Journal of Law and Economics, 26, 301-325. [4]Fan Ronghui and Lan Hailin . Determinants of CEO Turnover-performance Sensitivity. [5]Fredrickson, J. W., Hambrick, D. C., & Baumrin, S. 1988. A model of CEO dismissal. Academy of Management Review, 13: 255-270. [6]G. Brunello, C. Graniano, and B. M. Parigi. CEO Turnover in Insider-Dominated Boards: The Italian Case. Journal of Banking and Fiance, 2003, 27(6): 1027-1051. [7]Gabarro, J. J. 1987. The dynamics of taking charge. Boston: Harvard Business School Press. [8]Hambrick, D.C., & D?Aveni, R. A. 1992. Top team deteroraton as part of the downward spiral of large corporate bankruptcies. Management Science, 38: 1445-1466. [9]Hannan, M.T. and J. Freeman (1988). The ecology of organizational mortality; American labor unions, 1936-1985, American Journal of Sociology, 94, pp. 25-52. [10]J. W. Fredrickson, D. C. Hambrick, and S. Baumrin. A Model of CEO Dismissal. Academy of Management Review, 1988, 13(2): 255-270. [11]Jeremy Hall and Philip Rosson (2006). The Impact of Technological Turbulence on Entrepreneurial Behavior, Social Norms and Ethics: Three Internet-based Cases. Journal of Buniness Ethics, 64: 231-248. [12]John Freeman; Glenn R. Carroll; Michael T. Hannan (1983). The Liability of Newness: Age Dependence in Organizational Death Rates. American Sociological Review, Vol. 48, No. 5 (Oct., 1983), 692-710. [13]Jiangguo Yuan; Chunsheng Yuan; Xianlan Deng and Chunsheng Yuan. (2008). The Effects of Manager 12 / 16 Compensation and Market Competition on Financial Fraud in Public Companies: An Empirical Study in China. International Journal of Management, Vol. 25. No. 2. [14]March, J. G. and H. A. Simon (1958). Organizations. John Wiley, New York. [15]K. J. Murphy, and J. Zimmerman, Financial Performance Surrounding CEO Turnover (1993), Journal of Accounting and Economics, 16(1-3): 273-315. [16]Josef Bruderl and Rudolf Schussler (1990). Organizational Mortality: The Liabilities of Newness and Adolescence. Administrative Science Quarterly, Vol, 35, No, 3 (Sep., 1990), 530-547. [17]Shen, W., & Cannella, A. A. Jr. 2002. Power dynamics with top management and their impacts on CEO dismissal followed by inside succession. Academy of Management Journal, 45: 1195-1206. [18]Shen W. 2003. The Dynamics of the CEO-Board Relationship: An Evolutionary Perspecive. The Academy of Management Review, Vol. 28, No. 3 (Jul., 2003), pp. 466-476. [19]Stata. 1999. Stata reference manual, release 6, vol. 3. College Station, TX: Stata. [20]Tuma, N.B., & Hanan, M. T. 1984. Social dynamics. Orlando, FL: Academic Press [21]Tuma, N.B. and M. T. Hannan (1984). Social Dynamics. Academic Press, New York. [22]William P. Barneit, & Morten T. Hansen. 1996. The Red Queen in Organizational Evolution. Strategic Management Journal, Vol. 17. 139-157. [23]Van Valen, L. (1973). „A new evolutionary law?, Evolutionary Theory, 1, pp. 1-30. [24]Yamaguchi, K. 1991. Event History Analysis. Newbury Park, CA: Sage. [25]Shao-chuan Leng. Changes in China: Party, State and Society. Edited by The modernizaton of China.p9. 13 / 16 Table 1. Variable Means, Standard Deviations, and Correlation Coefficients Variable Mean s.d. 1 2 3 4 5 6 7 1.Insider CEO 0.43 0.50 2. Follower 0.41 0.49 -.09 3. Early years of tenure 0.90 0.30 .01 .02 4. Age 47.77 7.50 .02 -.02 -.06 5.Age2 2338.73 727.08 -.02 .02 .06 -.10 6. Duality 0.10 0.30 .08 -.06 .03 .01 -.01 7. Agent 0.96 0.19 -.04 -.02 -.01 .03 -.03 .11 8.Concorrent 0.17 0.38 -.06 .04 .01 .01 -.01 -.18 -.05 Pooled data, with n = 498 are shown. Correlation coefficients greater than .04 or less than 0.04 are significant at p<.05 14 / 16 αTable 2. Results of Patterns in Different Stages Early stage (Tenune?5) Later stage (5