How Quickly Do Ceos Become Obsolete
How Quickly Do Ceos Become Obsolete
How Quickly Do Ceos Become Obsolete
Strat. Mgmt. J., 27: 447460 (2006) Published online 27 February 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.524
HOW QUICKLY DO CEOs BECOME OBSOLETE? INDUSTRY DYNAMISM, CEO TENURE, AND COMPANY PERFORMANCE
ANDREW D. HENDERSON,1 * DANNY MILLER2 and DONALD C. HAMBRICK3
McCombs School of Business, University of Texas at Austin, Austin, Texas, U.S.A. Ecole des Hautes Etudes Commerciales, Montreal, Quebec, and University of Alberta, Edmonton, Alberta, Canada 3 Smeal College of Business Administration, Pennsylvania State University, University Park, Pennsylvania, U.S.A.
2 1
Scholars have characterized CEO tenures as life cycles in which executives learn rapidly during their initial time in ofce, but then grow stale as they lose touch with the external environment. We argue, however, that the opportunities for adaptive learning are limited because (1) a CEO assumes ofce with a relatively xed paradigm that changes little thereafter; (2) inertia limits the speed at which an organization can align itself with a new CEOs paradigm; and (3) for any within-paradigm learning to occur, the external environment must be stable enough so that the causeeffect relationships that CEOs glean today remain relevant tomorrow. In a longitudinal study of 98 CEOs in the relatively stable branded foods industry and 228 CEOs in the highly dynamic computer industry, we found results that strongly supported our hypotheses. In the stable food industry, rm-level performance improved steadily with tenure, with downturns occurring only among the few CEOs who served more than 1015 years. In contrast, in the dynamic computer industry, CEOs were at their best when they started their jobs, and rm performance declined steadily across their tenures, presumably as their paradigms grew obsolete more quickly than they could learn. Copyright 2006 John Wiley & Sons, Ltd.
In recent years, a signicant body of research has focused on the ways that top executives inuence strategic choices and organizational performance. Based on the premise that organizations are susceptible to the actions of their uppermost managers, but that managers are only boundedly rational, researchers have found that organizations
Keywords: CEO tenure; learning; obsolescence; CEO paradigms; strategic decision making
Correspondence to: Andrew D. Henderson, McCombs School of Business, CBA 4.202, B6300, University of Texas at Austin, Austin, TX 78712-0210, U.S.A. E-mail: andy.henderson@mccombs.utexas.edu
become reections of their top executives (Cyert and March, 1963; Hambrick and Mason, 1984). Studies have found, for example, that CEO personalities and functional backgrounds predict strategic actions and orientations (Gupta and Govindarajan, 1984; Miller, Kets de Vries, and Toulouse, 1982). Most research linking executive characteristics to organizational outcomes, however, has been cross-sectional in design and static in its logic. Apart from studies of CEO succession, it has been rare to dynamically model executive orientations and rm performance, with time playing a central theoretical role (e.g., Gabarro, 1987; Miller, 1991).
448
Yet from literatures on organizational evolution (Nelson and Winter, 1982), human factors in decision making (Rubin and Brockner, 1975; Staw and Ross, 1987), and executive cognition (Hambrick, Geletkanycz, and Fredrickson, 1993), it appears likely that senior executives, including CEOs, do not think, behave, or perform uniformly over their tenures. A few papers have examined how the impact of CEOs varies over their time in ofce. Hambrick and Fukutomi (1991) proposed that new CEOs begin with a knowledge decit but steadily learn about their jobs, organizations, and environments. After some point, though, CEOs are thought to become insular and overly wedded to their early formulas, resulting in an inverted U-shaped relationship between tenure and rm performance. In line with this model, Miller and Shamsie (2001), in a longitudinal study of the lm industry, found that rm performance increased for the rst 810 years of CEOs tenures and then began to fall. Pointing to the same pattern, Miller (1991) found that the environmentorganization alignments prescribed by contingency theory (e.g., Lawrence and Lorsch, 1967) were indeed observed for companies whose CEOs had been in ofce for a brief or moderate period; but those alignments were not observed when CEOs had been in ofce for 10 years or more, resulting in lower company performance. Miller concluded that CEOs are alert to the environment during their early years in ofce, but then lose touch and become stale in the saddle. These three works are strikingly consistent in their portrayals of CEO tenures, as each argues that CEOs pass through two phases during their time in ofce. The rst is an initial period of adaptive improvement, in which CEOs learn by doing and gradually implement a strategic orientation that ts the rm to its environment. Eventually, though, a second tendency takes hold that causes performance to decline. CEOs become overly committed to their earlier formulas, and their organizations become so tightly aligned with the status quo that change becomes difcult to consider and even harder to execute. The result is an inverted-U relationship between CEO tenure and rm performance. Such a relationship may generally occur, but an important contingency has yet to be considered. If we assume that CEOs often arrive at their jobs with paradigms that are well-suited to current conditions, yet also relatively xed, then the
Copyright 2006 John Wiley & Sons, Ltd.
449
paradigm and the external environment, will diminish rm performance at an increasing rate. Longtenured CEOs become increasingly committed to their earlier policies (Staw and Ross 1987). And through time, CEOs increasingly surround themselves with like-minded executives who reinforce the CEOs entrenched point of view (Hambrick, 1995). While new CEOs are highly attuned to the external environment (having been selected specically for their t with current conditions), longer-tenured executives are increasingly isolated from it (Hambrick and Fukutomi, 1991). Over time, mismatches between the CEOs paradigm and the environment tend to mount at an increasing rate due to a compounding process: with each passing year, environmental conditions move further away from those that the CEO was initially equipped to face; at the same time, the CEO becomes increasingly inattentive to those developments.
HYPOTHESES
If the CEO tenure clock paces two processes, one adaptive and involving the implementation of a CEOs initially suitable paradigm, the second maladaptive and involving the growing mismatch between a CEOs paradigm and the environment, what are the net effects on rm performance? The answer, we propose, is contingent on the dynamism of the external environment. Figure 1(a, b) depictsfor stable and dynamic settings, respectivelyhow the benets of internal alignment with a CEOs paradigm and the penalties of its external mist change over the course of a CEOs tenure, and what the resultant effects on rm performance are likely to be. A stable industry is one in which customer preferences, technologies, and competitive dynamics change little. There, as Figure 1(a) indicates, the benets of internal t and learning accrue steadily over a relatively long period, and the penalties for external mismatches accumulate slowly, becoming substantial only for CEOs who remain in ofce for extraordinary lengths of time. In stable settings, boards can select CEOs whose paradigms provide reasonable matches to both present and future conditions. Such foresight is possible because historical conditions are strongly correlated with future ones, and boards can observe alternative CEO candidates and project with some accuracy how
Strat. Mgmt. J., 27: 447460 (2006)
450
Overall Performance
(a)
CEO Tenure
Overall Performance
Figure 1.
Predicted effects of CEO tenure on rm performance. (a) Effects in stable environments. (b) Effects in dynamic environments
their paradigms will serve in coming years. In stable settings, the potential for steady improvement is considerable, since knowledge acquired today will, to a fair degree, also apply tomorrow (March, 1991; Weick, 1991). Therefore, even if CEOs paradigms evolve very little, they can gradually discern how to better implement their ideas, resulting in steady improvements over long periods. Later improvements will be smaller than earlier ones, as the limitations of a given paradigm are reached, but the overall trend is favorable. Figure 1(a) further indicates that, even in stable settings, there is at least some penalty to be paid for the growing mismatch between a CEOs paradigm and the environment, which arises from the accumulated effects of environmental drift, coupled with the tendency for long-tenured executives to lose touch with their external settings. While those hazards are relatively small and mount very gradually in stable settings, we expect they will become noticeable among the longest serving CEOs, resulting in an inverted U-shaped relationship. Most CEOs in stable industries will therefore leave ofce while their companys performance is still on the rise, but the few who stay in ofce the
Copyright 2006 John Wiley & Sons, Ltd.
longest will eventually exhibit declining performance. In comparison, Figure 1(b) depicts the scenario typical of dynamic industries in which the longrange benets of internal t are limited, and the penalties imposed by external mismatches mount quickly. In these settings, the potential for early improvement is sizeable because the predecessors paradigm is likely to have grown severely obsolete, giving the board an opportunity to appoint a successor whose worldview is much better aligned with current environmental conditions. Moreover, dynamic industries are characterized by an ongoing barrage of external jolts that disrupt the status quo and allow top executives to make major strategic alterations (Hambrick and Finkelstein, 1987; Tushman and Romanelli, 1985). As a result, new CEOs in dynamic environments can align their rms with their paradigms fairly quickly, which Figure 1(b) reects in the rapid initial rise of its internal t curve. While near-term benets can be grasped quickly in dynamic settings, two factors limit longterm improvements. First, accurate long-range prediction is all but impossible in dynamic settings,
Strat. Mgmt. J., 27: 447460 (2006)
451
Hypothesis 2: Peak rm-level performance will occur earlier in CEOs tenures in a dynamic industry than in a stable one.
METHODOLOGY
We drew samples from the computer and branded foods industries, selected for their widely contrasting degrees of dynamism. Prior research suggests that the factors affecting environmental dynamism include degrees of innovation, technological change, supply instability, competitive rivalry, and market growth (Aldrich, 1979; Dess and Beard, 1984). Over the study period, 195594, these two industries differed dramatically on all those dimensions. The computer industry was characterized by rapid technological change, much innovation, volatile growth, and unstable demand (Brown and Eisenhardt, 1997). In contrast, the food industry was much more stable, with slower and more consistent growth, and little technological or competitive change (Geletkanycz and Hambrick, 1997). For the computer industry, we examined all companies with a primary SIC code of 3570, 3571, or 3572 that were listed in COMPUSTAT in any year from 1955 through 1994. We then searched major public sources (Wall Street Journal Index, NEXIS, Standard and Poors Register of Directors and Executives) and company lings to identify the year that each CEO began and ended his or her tenure. Tenures that continued beyond 1994 were treated as right-censored. We considered the companys highest-ranking executive (typically the chairman) to be the CEO, unless another executive expressly held the CEO title. In all, the computer sample included 228 CEO tenures, ranging from 1 to 36 years. The median tenure was 4 years, and 15 percent were longer than 10 years. The total number of CEO-years was 1397. For the food industry, we examined companies with a primary two-digit SIC code of 20 (excluding those that were involved principally in agricultural commodities). The sample consisted of rms that were listed on the New York or American Stock Exchange in any year from 1955 to 1994. We used the same public sources noted above to identify the dates that CEOs started and completed their tenures. Again, tenures that continued beyond 1994 were treated as right-censored. The food industry sample included 98 CEO tenures, ranging
Strat. Mgmt. J., 27: 447460 (2006)
Hypothesis 1: There will be an inverted Ushaped relationship between CEO tenure and company performance. Performance will initially increase, but then later decrease.
Copyright 2006 John Wiley & Sons, Ltd.
452
from 1 to 36 years. The median was 6 years, and one-fourth were greater than 10 years. In all, there were 847 CEO-years. Measures Dependent variable To assess rm performance, we examined three annual measures of protability: (1) return on sales = net incomet /salest ; (2) return on assets = net incomet /net assetst ; and (3) return on invested capital = net incomet /(shareholders equityt + debtt ). All nancial data were taken from COMPUSTAT. Within each industry, those measures were highly correlated, so we used weights from factor analyses to extract a unied construct. In the computer sector, that analysis yielded a onefactor weighted average of protability with an eigenvalue of 2.28 that explained 76.1 percent of the total variance in return of sales, assets, and invested capital. The food industry analysis also yielded a one-factor solution. Its eigenvalue was 2.52, and it explained 84.1 percent of total variance. Independent variables CEO tenure was measured by counting the years a chief executive had been in ofce. We also calculated the square of tenure and included it where it was signicant. Unless otherwise noted, all predictors were updated annually and lagged by 1 year. Organizational controls To control for economies of scale, we measured rm size by taking the natural log of the number of employees. Because performance may exhibit path dependencies due to organizational learning, inertia, or the development of rm-specic capabilities (e.g., Barney, 1991; Teece, Pisano, and Shuen, 1997), we controlled for prior performance, as measured by prof itabilityt1 . Similarly, we controlled for rm age, measured in years. Its square was not signicant in any analyses. Changes in strategy may affect rm performance, so we assessed year-to-year adjustments in four key resource allocations: (1) expenditures on property, plant, and equipment (PP&E); (2) R&D expenditures; (3) advertising expenditures;
Copyright 2006 John Wiley & Sons, Ltd.
453
otherwise (cf. Haveman, 1992). Unlike our other variables, this dummy did not vary with time. Modeling and estimation We analyzed annual protability using generalized estimating equations (GEEs), an extension of the family of generalized linear models, which are designed to handle longitudinal data (Liang and Zeger, 1986; Lipsitz et al., 1994). GEEs derive maximum likelihood estimates from models with four parts: (1) a linear component, i = xi ; (2) a link function, g, which may be nonlinear, that describes how the means of observed values are related to the linear component: g(i ) = i = xi ; (3) a specication of the distribution from which i is drawn; and (4) a specication of the correlation structure among time-series outcomes. All models had a rst-order autoregressive correlation structure, corrected for under- and overdispersion, and provided robust variance estimates (White, 1980) to account for heteroscedasticity and unobserved differences across rms and CEOs. We assessed the overall goodness of t of each model by calculating minus two times its log likelihood score. Differences in that quantity across nested models have a chi-squared distribution.2 We estimated separate models for the food and computer industries rather than pooling them and using interactions to test cross-industry differences. Although pooling increases statistical power, it is problematic when the variance of the disturbance terms, 2 , differs substantially across groups (Greene, 1993: 236). That was clearly the case here since error variances were over 10 times larger in the dynamic computer industry than in the stable food sector.
2 To avoid confounding the effects of CEO tenure with other processes, we did not use xed-effects models. Most CEOs had a short time-series of data, so there were not enough observations to use CEO-based dummy variables to implement xed effects. Alternatively, we could have mean-deviated each CEOs observations (Greene, 1993), but that would have confounded the effects of CEO tenure, CEO age, and rm age. To illustrate, suppose a CEO was in ofce for 3 years, beginning at age 49 in an 84-year-old rm. When the within-CEO mean of each of those variables (2, 50, and 85 respectively) is subtracted from its corresponding observation, all three variables have a time-series of 1, 0, 1. As a result, they could not be modeled together. GEE models allowed us to include all three variables while controlling for unobserved differences across CEOs via the autocorrelation correction and robust variance estimators, which together, provide very conservative results (Liang and Zeger, 1986; Lipsitz et al., 1994).
Here, i represents the focal rm, t is the current year, t 1 is the prior year, and j indexes across all rms in the industry. Since contingency theory suggests that performance is enhanced when rms change their operations to match changes in the environment (Lawrence and Lorsch, 1967), we also assessed the interaction of strategic change and industry change mass and included it where it was signicant. To account for other within-industry differences across time, we coded four period-specic dummies that identied the 1950s, 1960s, 1970s, and 1980s. The 1990s were the omitted category. Results were unchanged using dummies that identied 5-year rather than 10-year periods. CEO controls As CEOs accumulate tenure, they grow older, which may affect their abilities (Walsh, 1995), so we controlled for CEO age, measured in years. Its square was not signicant. Left-censoring occurred among CEOs who (a) were in ofce prior to 1955, the start of the sample window, or (b) were already in ofce when their rm went public and rst entered COMPUSTAT. To control for that, the leftcensored CEO dummy variable was coded 1 across the entire time-series of those CEOs who were not observed from their rst year in ofce, and 0
Copyright 2006 John Wiley & Sons, Ltd.
454
Correcting for sample selection bias If boards tend to weed out incompetent CEOs, then inferior managers would have relatively short stays in ofce and capable CEOs would have long tenures. To correct for that problem, we used the technique described by Lee (1983). This approach, which is more robust than the two-step procedure described by Heckman (1979), rst uses predictors measured at t 1 to estimate the likelihood that a CEO will leave ofce in year t. We did that using an accelerated failure time (AFT) model with an exponential distribution. Next, we calculated: i,t = [ (F (i, t))] 1 F (i, t)
1
where i indexes CEOs, t indexes time, (x) is the standard normal density function, 1 (x) is the functional inverse of the standard normal distribution, and F (i, t) is the cumulative hazard function obtained from the AFT model. Once calculated, i,t was controlled in all analyses.
RESULTS
Tables 1 and 2 provide descriptive statistics for the branded foods and computer industries. We assessed multicollinearity using condition indices
0.05
0.04
Influence on Profitability
0.03
0.02
0.01
0.00
-0.03 0 5 10 15 20 25 30 35
Figure 2. Effects of CEO tenure on protability in the food industry at three levels of strategic change
Copyright 2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 447460 (2006)
Table 1. 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Variable
Mean
S.D.
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 0.46 0.16 0.43 0.66 0.26 0.00 0.29 0.39 0.16 0.67 0.30 0.21 0.04 0.49 0.18
Protability 0.08 0.04 CEO tenure 7.82 6.59 0.11 Firm size 2.48 1.36 0.01 0.19 Firm age 71.02 25.98 0.30 0.05 0.04 Strategic change 2.73 2.08 0.07 0.06 0.71 0.05 R&D missing 0.49 0.50 0.06 0.01 0.10 0.29 0.15 Advertising missing 0.30 0.46 0.10 0.30 0.26 0.34 0.29 0.48 Industry protability 0.08 0.02 0.18 0.05 0.02 0.02 0.07 0.22 0.24 Industry density 160.52 20.55 0.12 0.06 0.03 0.01 0.08 0.25 0.30 0.58 Industry change mass 10.49 2.39 0.13 0.20 0.06 0.20 0.20 0.36 0.56 0.58 The 1950s 0.01 0.04 0.03 0.03 0.01 0.02 0.05 0.04 0.06 0.03 The 1960s 0.15 0.36 0.08 0.21 0.11 0.22 0.25 0.39 0.64 0.58 The 1970s 0.37 0.48 0.20 0.19 0.07 0.22 0.06 0.11 0.06 0.66 The 1980s 0.29 0.46 0.09 0.13 0.09 0.18 0.21 0.11 0.37 0.15 CEO age 56.82 9.39 0.15 0.58 0.26 0.12 0.21 0.08 0.06 0.08 Left-censored CEO 0.04 0.21 0.04 0.01 0.03 0.19 0.16 0.18 0.34 0.27 Lambda 0.14 0.16 0.17 0.18 0.27 0.19 0.24 0.37 0.16 0.14
0.02 0.03 0.32 0.02 0.27 0.50 0.03 0.02 0.08 0.06 0.16 0.43 0.12 0.14 0.26 0.03 0.23 0.09 0.17 0.38 0.03
N = 847 CEO-years
455
456
Table 2. 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Variable
Mean
S.D.
Protability CEO tenure Firm size Firm age Strategic change R&D missing Advertising missing Industry protability Industry density Industry change mass The 1950s The 1960s The 1970s The 1980s CEO age Left-censored CEO Lambda
0.08 0.51 6.59 5.85 0.02 1.50 1.60 0.27 0.14 19.89 20.73 0.10 0.03 0.66 2.78 3.02 0.26 0.11 0.79 0.56 0.05 0.21 0.00 0.09 0.00 0.04 0.06 0.50 0.50 0.01 0.01 0.05 0.05 0.01 0.09 0.11 0.14 0.17 0.04 0.06 0.01 0.05 0.07 0.02 68.24 22.15 0.19 0.04 0.11 0.01 0.02 0.14 0.12 0.75 13.14 2.42 0.19 0.03 0.17 0.04 0.03 0.17 0.14 0.74 0.96 0.01 0.05 0.03 0.02 0.05 0.06 0.00 0.11 0.05 0.12 0.16 0.23 0.04 0.19 0.09 0.02 0.14 0.04 0.04 0.13 0.20 0.36 0.52 0.64 0.01 0.18 0.38 0.15 0.08 0.01 0.04 0.05 0.05 0.03 0.56 0.65 0.52 0.02 0.09 0.44 0.50 0.02 0.01 0.06 0.02 0.01 0.05 0.14 0.08 0.32 0.26 0.05 0.17 0.42 51.09 8.34 0.03 0.23 0.39 0.42 0.32 0.08 0.01 0.00 0.01 0.02 0.02 0.03 0.02 0.00 0.26 0.44 0.08 0.48 0.05 0.21 0.00 0.04 0.04 0.05 0.05 0.04 0.01 0.01 0.07 0.04 0.28 0.28 0.11 0.20 0.16 0.02 0.16 0.02 0.11 0.01 0.43 0.46 0.46 0.10 0.26 0.29 0.01 0.54 0.27
N = 1397 CEO-years
457
Predictor variables
0.001 (0.015) 0.181 (0.010) 0.002 (0.001) 0.003 (0.007) 0.044 (0.027) 0.071 (0.071) 0.029 (0.013) 0.000 (0.001) 0.032 (0.009) 0.063 (0.142) 0.099 (0.118) 0.113 (0.053) 0.167 (0.049) 0.001 (0.002) 0.065 (0.058) 0.154 (0.089) 2346.724 n.a.
0.020 (0.006) 0.001 (0.000) 0.002 (0.001) 0.008 (0.014) 0.180 (0.010) 0.002 (0.001) 0.012 (0.010) 0.042 (0.026) 0.060 (0.064) 0.032 (0.013) 0.001 (0.001) 0.034 (0.009) 0.003 (0.138) 0.120 (0.121) 0.134 (0.060) 0.160 (0.053) 0.003 (0.002) 0.036 (0.052) 0.189 (0.098) 2335.320 11.404
0.169 (0.062) 0.622 (0.054) 0.004 (0.003) 0.050 (0.026) 0.022 (0.316) 0.095 (0.117) 0.447 (0.167) 0.041 (0.069) 1.891 (0.838) 1.595 (0.708) 1.525 (0.381) 0.689 (0.190) 0.002 (0.010) 0.131 (0.156) 1.974 (0.902) 4359.546 n.a.
0.254 (0.067) 0.607 (0.056) 0.005 (0.003) 0.002 (0.027) 0.002 (0.298) 0.114 (0.122) 0.432 (0.163) 0.053 (0.082) 0.029 (0.009) 1.733 (0.832) 1.595 (0.669) 1.453 (0.377) 0.692 (0.189) 0.006 (0.010) 0.379 (0.189) 1.564 (0.957) 4349.036 10.510
p < 0.001; two-tailed tests. N = 847 CEO-years for the food industry; N = 1397 CEO-years for the
used Model 2 to calculate the partial derivative, protability/ tenure. That quantity changed from positive to negative for each observed value of strategic change, so Hypothesis 1 was supported. Models 3 and 4 of Table 3 analyze protability in the dynamic computer industry. Model 3 contains the controls, and Model 4 adds CEO tenure and the signicant interaction of industry change
Copyright 2006 John Wiley & Sons, Ltd.
mass and strategic change. (Here, neither tenure2 nor the interaction of tenure and strategic change was signicant.) As Model 4 shows, CEO tenure had a negative and signicant effect on protability. We had envisioned a brief period of improvement for computer CEOs, followed by steep deterioration, but, instead, protability declined immediately and steadily throughout a CEOs tenure.
Strat. Mgmt. J., 27: 447460 (2006)
458
Computer CEOs may have improved their rms performance within their rst year in ofce before declining, but our annual data precluded our detecting that. Results, then, for the computer industry did not support the predicted inverted U-shaped relationship between tenure and company performance. Instead, those executives were at their best at the outset of their tenures and then steadily worsened. Hypothesis 2 predicted that performance would peak earlier in a dynamic industry than a stable one, and our results support that expectation. Peak performance occurred in year 1 of the tenures of computer industry CEOs. In comparison, as Figure 2 indicates, performance peaked much later among chief executives in the stable food industry. Indeed, the peak occurred at about 11 years of tenure among food executives making large strategic changes, at about 17 years for those making medium-size changes, and at about 23 years for those making small changes. Notably, about three-fourths of food industry CEOs exited before 11 years, so very few stayed in ofce long enough to see their performance drop; in contrast, according to our estimates, virtually all CEOs in the dynamic computer industry served long enough to see their early successes followed by performance deterioration.
DISCUSSION
Prior research suggests that CEOs adaptively learn for a decade or more of their tenures before rm performance suffers (Miller, 1990; Miller and Shamsie, 2001). In comparison, we nd that the potential for any performance improvements is contingent on the dynamism of the external environment. Among rms in the stable branded foods industry, and consistent with current theory, rm performance increased for at least ten years of a CEOs tenure before declining. Yet in the turbulent computer industry, we could nd no evidence of time-related performance improvements. There, CEOs were at their best during their rst year in ofce and worsened steadily thereafter. The early onset of CEO obsolescence in dynamic settings contrasts sharply with prior research, which has emphasized adaptive learning over extended periods. An immediate and steady decline is consistent, though, with our premise that CEOs assume ofce with relatively xed paradigms.
Copyright 2006 John Wiley & Sons, Ltd.
459
ACKNOWLEDGEMENTS
We are indebted to Syd Finkelstein, Jim Fredrickson, Heather Haveman, Paul Ingram, Associate Editor Ed Zajac, and two anonymous reviewers for their valuable suggestions on earlier drafts of this paper. We thank Peter Aranda, Greg Henley, and Eric Jackson for their research assistance. Generous nancial support was provided to the rst author by the Kelleher Center for Entrepreneurship, Growth, and Renewal and to the second author by the Social Sciences and Humanities Research Council of Canada.
REFERENCES
Aldrich HE. 1979. Organizations and Environments. Prentice-Hall: Englewood Cliffs, NJ. Argote L, Beckman SL, Epple D. 1990. The persistence and transfer of learning in industrial settings. Management Science 36: 140154. Argote L, Epple D. 1990. Learning curves in manufacturing. Science 247: 920924. Barnett WP, Amburgey TL. 1990. Do larger organizations generate stronger competition? In Organizational Evolution: New directions, Singh JV (ed). Sage: Newbury Park, CA; 78102. Barney J. 1991. Firm resources and sustained competitive advantage. Journal of Management 17: 99120. Baum JAC, Mezias SJ. 1992. Localized competition and organizational failure in the Manhattan hotel industry. Administrative Science Quarterly 37: 580604. Belsley DA. 1991. Conditioning Diagnostics. WileyInterscience: New York. Brown SL, Eisenhardt KM. 1997. The art of continuous change: link complexity theory and time-paced evolution in relentlessly shifting organizations. Administrative Science Quarterly 42: 134. Cyert R, March J. 1963. A Behavioral Theory of the Firm. Prentice Hall: Englewood Cliffs, NJ. Dess GG, Beard DW. 1984. Dimensions of organizational task environments. Administrative Science Quarterly 29: 5273. Finkelstein S, Hambrick DC. 1996. Strategic Leadership: Top Executives and their Effects on Organizations. West Publishing: St Paul, MN. Gabarro J. 1987. The Dynamics of Taking Charge. Harvard Business School Press: Boston, MA.
Copyright 2006 John Wiley & Sons, Ltd.
Geletkanycz MA, Hambrick DC. 1997. The external ties of top executives: implications for strategic choice and performance. Administrative Science Quarterly 42: 646681. Greene W. 1993. Econometric Analysis. Macmillan: New York. Gupta A, Govindarajan V. 1984. Business unit strategy, managerial characteristics, and business unit effectiveness at strategy implementation. Academy of Management Journal 27: 2541. Hambrick DC. 1995. Fragmentation and the other problems CEOs have with their top management teams. California Management Review 37: 110127. Hambrick DC, Finkelstein S. 1987. Managerial discretion. In Research in Organizational Behavior , Staw BM, Cummings LL (eds). JAI Press: Greenwich, CT; 9: 369406. Hambrick DC, Fukutomi G. 1991. The seasons of a CEOs tenure. Academy of Management Review 16: 719742. Hambrick DC, Geletkanycz M, Fredrickson J. 1993. Top executive commitment to the status quo: some tests of its determinants. Strategic Management Journal 14(6): 401418. Hambrick DC, Mason P. 1984. Upper echelons: the organization as a reection of its top managers. Academy of Management Review 9: 193206. Hannan MT, Carroll G. 1992. Dynamics of Organizational Populations. Oxford: New York. Hannan MT, Freeman J. 1984. Structural inertia and organizational change. American Sociological Review 49: 149164. Haveman HA. 1992. Between a rock and a hard place: organizational change and performance under conditions of fundamental environmental transformation. Administrative Science Quarterly 37: 4875. Heckman JJ. 1979. Sample selection bias as a specication error. Econometrica 47: 153161. Kao J. 1985. American Express Travel Related Services Company (9-485-174). Harvard Business School: Boston, MA. Kotter JP. 1982. The General Managers. Free Press: New York. Lawrence P, Lorsch J. 1967. Organization and Environment. Harvard Business School Press: Boston, MA. Lee L. 1983. Generalized econometric models with selectivity. Econometrica 51: 507512. Levinthal D, March JG. 1981. A model of adaptive organizational search. Journal of Economic Behavior and Organization 2: 307333. Levinthal D, March J. 1993. The myopia of learning. Strategic Management Journal Winter Special Issue 14: 95112. Liang K, Zeger SL. 1986. Longitudinal data analysis using generalized linear models. Biometrika 73: 1322. Lipsitz SR, Fitzmaurice GM, Orav EJ, Laird NM. 1994. Performance of generalized estimating equations in practical situations. Biometrics 50: 270278. Lounamaa P, March J. 1987. Adaptive coordination of a learning team. Management Science 33: 107123.
Strat. Mgmt. J., 27: 447460 (2006)
460
March JG. 1991. Exploration and exploitation in organizational learning. Organization Science 2: 7187. Miller D. 1990. The Icarus Paradox . Harper Business: New York. Miller D. 1991. Stale in the saddle: CEO tenure and the match between organization and environment. Management Science 37: 3452. Miller D. 1993. The architecture of simplicity. Academy of Management Review 18: 116138. Miller D, Dr ge C. 1986. Psychological and traditional o determinants of structure. Administrative Science Quarterly 31: 539560. Miller D, Kets de Vries M, Toulouse JM. 1982. Top executive locus of control and its relationship to strategy making, structure and environment. Academy of Management Journal 24: 221235. Miller D, Shamsie J. 2001. Learning across the life cycle: experimentation and performance among the Hollywood studio heads. Strategic Management Journal 22(8): 725745. Nelson R, Winter S. 1982. An Evolutionary Theory of Economic Change. Belknap: Cambridge, MA. Rubin JZ, Brockner J. 1975. Factors affecting entrapment in waiting situations: the Rosencrantz and Guildenstern effect. Journal of Personality and Social Psychology 31: 10541063. Siggelkow N. 2002. Evolution toward t. Administrative Science Quarterly 47: 125159.