JPART 20:i261–i279 INCENTIVES AND PUBLIC SERVICE PERFORMANCE: A SPECIAL ISSUE Does Public Service Performance Affect Top Management Turnover? George A. Boyne*, Oliver James† , Peter John‡, Nicolai Petrovsky§ *Cardiff University; University of Exeter; ‡University of Manchester; §University of Kentucky ABSTRACT Political and organizational theories suggest that the turnover of chief executives and other members of senior management teams are likely to be influenced by public service performance. We use a panel data set of 148 English local governments over 4 years to test this proposition. The empirical results show that performance has a negative effect on turnover, but that this effect is weaker for chief executives than for members of their senior management teams. In addition, top team turnover is higher in the years following a chief executive succession. The evidence suggests that chief executives can influence top team turnover by attributing responsibility for poor organizational performance to other senior managers. In this article, we explore whether public service performance makes a difference to the turnover of top managers in public organizations. This is an important question because most advanced democratic systems of government partially insulate top management (defined as chief executives and other senior managers) from politicians’ and other stakeholders’ judgments about performance, so that they can get on with managing services. Whereas fewer posts in the United States enjoy such protection than in most other advanced democracies, a large number of managerial posts in public organizations are protected against dismissal, which can only occur through procedures devised to guarantee due process. Given these protections, we should not expect a strong effect of performance on senior management change. On the other hand, senior managers are expected to be accountable for performance and to accept responsibility for results, ultimately by resigning or being dismissed from their positions, which implies there is such an effect. The latter view has become more prominent with New Public Management reforms emphasizing both managing for results and accountable management (Hood 1991; Pollitt and Bouckaert 2004). These developments raise the question of whether, and in what way, organizational performance affects senior management team turnover. Are top officials more likely to retain their posts when public service standards are deemed to be good, and to move on when standards are perceived as weak? Further, does the impact of performance on turnover differ between chief executives and other senior officials in public organizations? In Author order is alphabetical: all make an equal contribution to the article. Address correspondence to the author at boyne@cardiff.ac.uk. doi:10.1093/jopart/muq024 ª The Author 2010. Published by Oxford University Press on behalf of the Journal of Public Administration Research and Theory, Inc. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org i262 Journal of Public Administration Research and Theory particular, are chief executives able to pass on the blame for poor performance to members of their senior management team? Although a growing body of research in the last decade has supported the view that management matters for performance (Boyne et al. 2006), there are few studies of whether performance matters for managers. Theoretical arguments that organizational performance is associated with turnover derive mostly from the large literature on control of senior management in private firms. The sources of senior management turnover in public organizations are also of great practical and policy interest,1 yet there is little theoretical or empirical work on this topic (a rare exception is McCabe et al. 2008). We now propose hypotheses about the relationship between service results and turnover in the public sector, based largely on political principals’ incentives for removing bureaucratic agents who appear to have led an organization to weak performance and their motives for retaining agents who have overseen strong performance. We take account of the perceptions of performance of both government agencies and citizens and examine whether the link between performance and turnover differs between these measures. Next, we operationalize our theoretical variables and test our hypotheses against a 4-year panel data set on 148 English principal local governments, covering financial years 2002–03 to 2005–06. We estimate linear panel data models of the senior management turnover rate and logit models of the presence or absence of a chief executive succession. Finally, we draw a number of conclusions for research on leadership turnover. THEORIES OF ORGANIZATIONAL PERFORMANCE AND MANAGERIAL TURNOVER The large organization theory and management literatures on senior management turnover and executive successions in the private sector fall into three main strands (Pfeffer and Salancik 1977). One is concerned with the consequences of senior management turnover for large organizations. Another is concerned with the causes and consequences of ‘‘insider’’ or ‘‘outsider’’ succession. Much research fits in this strand, as does the limited work on the public sector, such as Boyne and Dahya (2002, 185–7), Hamidullah and Wilkins (2007), and Hill (2005). The final strand, of which this article is part, is concerned with the effects of organizational performance on senior management turnover. Many empirical studies of private corporations find that the tenure of senior managers is contingent on organizational performance, as measured by financial indicators, such as share prices and rate of return on investment (for summaries and reviews of these studies, see Brickley 2003; Furtado and Karan 1990; Kesner and Sebora 1994). The main theoretical interpretation of this finding is that board members use performance information to punish or reward senior managers. If performance falls below the standard achieved by competitors, or is less than has recently been achieved by the firm, then board members take steps to ease out or move aside the chief executive or the entire senior management team in an attempt to turn the tide. This is not an altruistic act to save the company: the reputation and remuneration of the board members themselves are strongly linked to the financial success of the companies they are associated with. 1 There is a particular interest in this topic in the United Kingdom; for instance, see ‘‘Tougher at the Top? Changes in the Labour Market for Single Tier and County Council Chief Executives—a Discussion Paper,’’ The Audit Commission, July 2008, and ‘‘Making a move’’ in the British local government professionals’ magazine The MJ, October 11, 2007. Boyne et al. Does Public Service Performance Affect Top Management Turnover? In this article, we draw on both organization theory and political science to examine the links between performance and the turnover of senior management of local public services provided by democratic governments. In the public sector, elected bodies such as legislatures, councils, and school boards can be seen as the rough analogue of the board of directors. There are clear incentives in the public sector for incumbents to care about public service performance. However, politicians and senior managers understand that it is not just performance but perceptions of their responsibility for performance that are at issue, leading to various blame games as individuals seek to avoid responsibility for poor outcomes (Hood 2002). In the contemporary public sector, it appears that politicians and senior management share the propensity for being blamed by stakeholders in policy fiascos (Hood et al. 2009; Weaver 1986) and for failure to meet policy targets (James 2004). Local political incumbents’ chances of re-election partly depends on the judgments of external stakeholders on the standard of public service provision. In moral hazard models of electoral responsibility, the repeated need for re-election and citizen monitoring gives incentives to politicians to exert effort (Ferejohn 1986). Research has shown that voters are concerned about public service performance at the local level, with low performance translating into vote share losses for the incumbent administration (James and John 2007). This research mirrors a more general tendency toward so-called valence issues, such as performance, where electors agree on the desirability of certain public services (Clarke et al. 2004; Stokes 1963). This tendency reinforces politicians’ concern with public service performance. The theories just discussed suggest that politicians should be concerned about their bureaucratic agents’ contribution to performance outcomes, and to apply rewards or sanctions to them, such as the decision to retain or replace the senior managers who run the organization. Even if performance is not directly the consequence of professional administrators’ action or inaction, moving them on may be a way for politicians to try and persuade voters that it was administrators rather than politicians who were responsible for disappointing results. Research on the performance of fifteen local governments in England found that a poor Comprehensive Performance Assessment (CPA) was a trigger for senior managers to depart in some, but not all, cases (Turner and Whiteman 2005). However, the extent to which politicians can change managers may be limited by the need to attract and retain competent managers. There are statutory protections against dismissal of managers in many public organizations. Nevertheless, even if outright dismissal might be difficult, senior managers can be induced to leave voluntarily through increases in oversight and micro-management by the elected body. For example, Brady and Helmich (1984, 30–1) describe how an American school board heavily involved itself in the day-to-day operations of the school district until the superintendent chose voluntary early retirement. Generally, we expect that politically motivated change mostly occurs by these and other means, including negotiated early retirements and early departures with severance packages. It is often impossible to disentangle such induced departures from voluntary departures, even in the private sector (Osborn et al. 1981; Wagner, Pfeffer, and O’Reilly 1984). The bottom line is that politicians have the motive and means to remove senior managers when organizational performance is perceived to be low. Thus, our first hypothesis is H1 The relationship between perceived organizational performance and the turnover of senior managers is negative. i263 i264 Journal of Public Administration Research and Theory Not all senior managers are exchangeable bureaucratic agents. Rather, the interests and actions of a chief executive and the remainder of the senior management team2 may diverge when organizational performance is perceived to be poor. The allocation of responsibility for performance between the chief executive and members of the senior management team may not be straightforward. The chief executive is part of the senior management of the organization, but he or she is also in charge of it and is responsible to politicians for overall performance. In this sense, he or she makes judgments about the rest of the management team. So, just as new political leaders may replace a chief executive who was appointed by their predecessors (Boyne et al. 2010), so a new chief executive may wish to change the membership of the senior management team. Studies of private organizations suggest that incoming chief executives seek to distance themselves from their predecessors and to adopt distinctive priorities and strategies (Bigley and Wiersema 2002; Simons 1994). In the public sector, radical strategic change, such as moving into new markets or closing down services, is not a likely course of action because of legal and political constraints (Boyne and Walker 2004). The default strategy is therefore to change the organization itself, which usually involves internal restructuring and transfer of responsibilities between departments. This may lead to the reconstitution of the senior management team, either through expansion, contraction, or reshuffling of portfolios between existing and new members. Even if no structural change takes place, a new chief executive is likely to want people in the senior management team who are sympathetic to the revised agenda for the organization. Managers appointed by the previous chief executive might be viewed as loyal to the old regime or lacking the skills to implement the new strategies. In this case, new senior managers who fit the chief executive’s preferences and policies are likely to be appointed. For these reasons, the impact of a chief executive is likely to increase turnover in subsequent periods, and our second hypothesis is H2 Chief executive succession is positively related to senior management team turnover. Even controlling for the effects of chief executive turnover, members of the senior management team other than the chief executive may be especially vulnerable if performance is perceived as poor. Just as politicians have incentives to shift blame for weak results onto appointed officials, so there may be an incentive for chief executives to remove other senior managers because being considered responsible for poor performance would be detrimental not only for their tenure in their present position but also for their chances on the market for similar positions. Chief executives’ prospects in the labor market depend to a large extent on the reputation of the organization they are working for. This holds even for chief executives in the later stages of their careers, who have an incentive to be associated with a high-reputation organization in order to obtain appointment to boards and other roles after their formal retirement (for a theoretical exposition, see Tadelis 2002). Boeker (1992) provides an illuminating study of chief executives attributing responsibility for low performance to other senior managers. Evaluating the effect of performance on rates of chief 2 A chief executive is the most senior appointed official in a public organization, who reports to the elected officials overseeing the organization. Examples include the city manager or appointed county director in US local government and the chief executive or head of paid service in UK local government. The rest of the senior management team is made up of the echelon of appointed officials who report directly to the chief executive. Examples are the department heads in US local government and the service directors in UK local government. This is the standard approach to defining the senior management team in prior studies (Carpenter, Geletkanycz, and Sanders 2004). Boyne et al. Does Public Service Performance Affect Top Management Turnover? executive succession and top management team turnover in 67 organizations over a 22-year period, he finds evidence in accordance with the idea that chief executives shift blame to their top management teams when performance is weak: ‘‘powerful chief executives buffer themselves from performance responsibility but ‘compensate’ by replacing top managers’’ (Boeker, 1992, 418). It should be noted that this might be scapegoating but it may also be an accurate allocation of responsibility. We are not aware of any prior research on this topic in the public sector, but blame shifting seems possible here just as in the private sector. Chief executives are more likely than other senior managers to spend time with their political masters and may develop close working (and perhaps social) relationships with them. The chief executive may take on the role of policy adviser and political confidante to the ruling politicians and thereby shape their perceptions of the organization in his or her favor. If performance problems emerge then chief executives may allocate responsibility or shift the blame to other senior managers, protecting their own reputation. In this case, chief executives may be largely insulated from the effects of declining or poor performance, whereas their colleagues in the senior management team are more vulnerable. Our third hypothesis is, therefore H3 The negative relationship between perceived performance and turnover is weaker for chief executives than for other members of the senior management team. MEASURES AND DATA We test our hypotheses against 4 years of panel data on all 148 English principal local governments (London boroughs, metropolitan boroughs, unitary authorities, and county councils). These entities are governed by popularly elected bodies with a Westminster-style cabinet system of political management. They are multipurpose governments that provide education, social services, regulatory services (such as land use planning and waste management), libraries, leisure services, housing, and welfare benefits.3 The advantages of studying these public organizations are that senior management appointments are subject to uniform statutory and regulatory constraints and that a wide range of consistently reported performance data are available on them, which tend not to available for US local governments, allowing us to extend existing research on the determinants of local government senior management turnover (for instance, Feiock and Stream 1998). We now introduce all variables. Summary statistics for them are provided in table 1. The summary statistics are listed for the common estimation sample that we use in all of our panel data models, covering the four financial years from 2002–03 through 2005–06.4 We are constrained to these 4 years because the indicators of performance used by British central government were first reported in December 2002. As we explain below, we also estimate models on a cross-section for the 2003–04 financial year. The estimation sample for these models is a proper subset of the common estimation sample described in table 1. This implies that the descriptives for all variables in the cross-sectional models are bounded by the extremes listed in table 1. 3 County councils do not provide the last two services listed (housing and welfare benefits). They are provided by a number of district councils within each county. 4 Each financial year runs from April 1 to March 31. i265 i266 Journal of Public Administration Research and Theory Table 1 Summary Statistics Mean Turnover rate Chief executive succession (dummy) Core service performance score Year-on-year change in core service performance score CPA grade: fair/two stars (dummy) CPA grade: good/three stars (dummy) CPA grade: excellent/four stars (dummy) CPA same as or worse than last year (dummy) Satisfaction with the local government (percent very/fairly satisfied) in 2003 Change in percent satisfaction with the local government from 2000 to 2003 Whole council election (dummy) Election by thirds (dummy) Difference between vote share of two biggest parties in the last election Change in political party control (dummy) Chief executive succession last year (dummy) Chief executive succession 2 years ago (dummy) Ability to improve score Logged population London borough (dummy) Metropolitan borough (dummy) 0.19 0.17 70 0.72 0.24 0.42 0.23 0.76 52 27.7 Standard Deviation Min Max 0.20 0.38 8.4 7.1 0.43 0.49 0.42 0.43 8.3 0 0 40 225 0 0 0 0 31 1 1 90 22 1 1 1 1 77 8.1 227 17 0.24 0.13 11 0.43 0.34 8.8 0.11 0.15 0.18 33 5.4 0.22 0.24 0.32 0.36 0.39 6.4 0.27 0.42 0.43 0 0 0.04 0 0 0 15 4.5 0 0 1 1 45 1 1 1 48 6.1 1 1 Note: These summary statistics are for the common estimation sample used by all models in Tables 2 and 4 (431 observations on 148 local governments). The estimation sample of all cross-sectional models in Tables 3 and 5 (141 observations) is a proper subset of this larger common estimation sample. Management Turnover The two dependent variables are (a) the turnover rate of the senior management team excluding the chief executive and (b) the presence or absence of a chief executive succession. We collected data on the composition of senior management teams from several sources—identifying the posts on the senior management team in every principal local government and year and the names of the post holders in every year. We measure the turnover rate as: Turnover rate 5 Number of new names on senior management team : Number of positions mentioned In this definition, the denominator is the number of positions on the senior management team in that year.5 5 As a robustness check, we also estimated our models using a slightly different indicator of the turnover rate, where in addition to new names on the senior management team, vacancies are also included in the numerator. This produces virtually the same results numerically and changes none of our conclusions. For the sake of clarity, we focus on the most cut and dried definition of the turnover rate. Boyne et al. Does Public Service Performance Affect Top Management Turnover? Our indicator captures exclusively those senior management successions where the new person was not already part of the senior management team. The indicator has a mean of about 0.2, which means that on average a fifth of the senior management team gets replaced in a given year. As our second dependent variable, we examine the presence or absence of a chief executive succession in a local government. We code a chief executive succession in the first year that a new person occupies the post of chief executive. In our estimation sample, there are 74 chief executive succession events. Out of 431 observations in the sample, ¨ this gives a naıve expected probability that a chief executive succession will occur in any given council in any given year of slightly less than 0.2. This figure squares with an average tenure of English local government chief executives of about 5 years that is mentioned in the practitioner press.6 In our panel data set, variation is made up of two components: variation within the local governments over time and variation between local governments. The former can be considered more useful variation for the purpose of causal inference since it does not depend on fixed differences between the local governments, which may be difficult to capture adequately with control variables. Previous case study work on local governments found enduring idiosyncratic features of local governments to influence the impact of the CPA process (Turner and Whiteman 2005). This problem bedevils purely cross-sectional analyses. It turns out that about 62% of the variation in the senior management turnover rate consists of changes within the local governments over time. Similarly, for the binary variable ‘‘presence or absence of a chief executive succession,’’ about 70% of the variation is within the local governments over time. All this is fairly encouraging for the robustness of our inferences against alternative explanations based on time-invariant omitted variables. Organizational Performance We operationalize public service performance in three different ways. The first two capture the British central government’s perceptions of local governments’ performance, as filtered through the Audit Commission, an independent agency that evaluates the service performance of local governments. The last captures local citizens’ perceptions of performance. The core service performance score, our first indicator, seeks to tap the objective elements of services (e.g., quantity and quality of output, speed of delivery). It is a weighted index covering annually collected quantitative output and outcome indicators on education, social care, environment, housing, benefits, and libraries and leisure as well as efficiency in the ‘‘management of resources’’ (Andrews et al. 2005, 641). The core service performance score ranges from 40 to 90, with a mean of 69.5. This measure is important to local politicians because deficiencies on its component indicators, particularly in education, are highlighted in league tables, can trigger visits by central government inspectors, and are likely to be publicized in local media. There is a direct incentive to monitor managerial success in delivering high scores on the indicators going into the core service performance score. The CPA, our second indicator, includes not only the core service performance score but also the Audit Commission’s judgment of local government’s ‘‘ability to improve,’’ 6 Compare ‘‘Making a move’’ in The MJ, October 11, 2007. i267 i268 Journal of Public Administration Research and Theory which reflects Audit Commission inspectors’ views of the appropriateness of the management arrangements and plans that are in place. The CPA was a performance evaluation and incentives program that central government introduced in 2002 and that was conducted annually until 2008. Most interestingly, the CPA had real teeth: Local governments classified as performing only in the lowest category have seen a temporary tightening of central government control (Lowndes 2003). Moreover, the CPA classification had special prominence for local politicians and officers because their local government was publicly evaluated in the national and local media using a five-point star rating system, receiving between zero (for ‘‘poor’’ performance) and four stars (for ‘‘excellent’’ performance). We create dummy variables for all but the two lowest categories, poor/no star, and weak/one star, which form our base group.7 It contains about 11% of observations in our estimation sample. About 24% show fair performance (two stars), 42% show good performance (three stars), and 23% show excellent performance (four stars). Comparing the last election before the introduction of the CPA and the election immediately following it, James and John (2007) show that performing badly on the CPA had adverse consequences for political incumbents’ electoral support, providing them with a strong incentive to retain managers who can avoid these bad ratings. There had been concerns that the CPA was not a valid measure of overall local government performance in the sense that it partially rewarded local governments that serve less demanding areas and punished those serv´ ing areas of high deprivation. Both Andrews et al. (2005) and Gutierrez-Romero, Haubrich, and McLean (2008) find that CPA results are partially influenced by constraints that local governments have no immediate control over. However, these influences do not invalidate the measures, and CPA reflected central government perceptions of performance. Local governments that stood out as by far the worst performers at the introduction of the CPA made very notable improvements. More broadly, although local governments in areas of high deprivation face a more challenging task, part of their statutory purpose is to provide local services and therefore it is reasonable to hold them accountable for this activity using performance measures. Finally, local citizens’ perceptions of performance are captured by the estimated percentage of the population stating that they are satisfied overall with the services provided by their local government. For each local government, this estimate is based on a large random sample of the adult population residing within the local government’s jurisdiction and is carried out according to central government quality standards. These ‘‘Best Value’’ satisfaction surveys take place every 3 years. We use the value for the 2003–04 financial year. On average, only 52% of the population in a local government jurisdiction is content with the way the organization conducts its business. We examine performance both absolutely and relative to the past. Even if performance remains above average, a dip from recent levels is likely to lead to pressures on senior managers (Farrell and Whidbee, 2003; Puffer and Weintrop, 1991). We therefore include indicators of performance change.8 For the core service performance score, we include the difference in performance between this year and last year. For the CPA, we include 7 We group these two categories together because there are only 10 observations with a zero stars rating. Whether or not these two categories are grouped together does not affect our findings; however, the exposition is clearer when they are grouped together, as this crisply brings out the contrast between low performance and better-than-low performance. 8 Although our performance data are for the same financial year as the dependent variables, they nevertheless are temporally prior because they refer to performance in the preceding financial year. Boyne et al. Does Public Service Performance Affect Top Management Turnover? a dummy variable that takes on the value of one if the CPA rating declined by one or more categories relative to last year or remained unchanged. Otherwise, this variable takes on the value of zero. For citizens’ perceptions of performance, we include the percentage point difference between the population satisfied in 2000 and the population satisfied in 2003. On average, the percentage of the population who are content with their local government has shrunk, apparently because of the increased performance expectations the CPA created (James 2009). Chief Executive Succession as an Explanatory Variable In order to test hypothesis 2, that chief executive successions tend to lead to replacements of other senior managers, we include 1- and 2-year lags of chief executive succession as explanatory variables in the models of top team turnover. Control Variables Our theoretical argument presents the elected council as the local government analogue of the board of directors in a private corporation. This implies that there are a set of political factors that may also influence turnover. First, an election is likely to lead to some new faces on the local council, even if the incumbent majority party retains control. This churn in the ruling group may disrupt relationships between politicians and managers. To capture precisely any election-induced differences in turnover, we create dummy variables for elections where the whole council is elected and for elections where there is an election by thirds. The latter implies that one council seat from each three-seat ward is up for election. This peculiar type of election takes place exclusively in metropolitan boroughs and a number of unitary authorities. The absence of an election forms the base group of observations. Second, the expected competitiveness of the political environment in a local government could also heighten turnover. A more competitive environment implies closer monitoring of management by the ruling party. Our indicator of perceived competitiveness of the next election is the percentage point difference between the vote share of the party with the largest number of votes and the party with the second-largest number of votes at the last election. Finally, if a change in political party control takes place, a new set of political leaders may regard the incumbent officers as too closely associated with the previous regime, and unlikely to support new policies, or they may value the symbolic effects of appointing a new chief executive or entire senior management team. McCabe et al. (2008) show that turnover of council members in US municipalities is associated with higher managerial turnover. We consequently include a dummy for a change in political control of the council, including changes to or from a situation where no political party holds an overall majority. Perceived management capacity could be linked to senior management turnover. Where it is relatively low, councils will be more likely to replace senior managers. As a proxy for senior management capacity, we include the CPA ‘‘ability to improve’’ score for 2002–03. It captures the domains of financial management, information technology management, capital management, and managing for results and thus most of the elements of management capacity identified by Ingraham, Joyce, and Donahue (2003). This measure i269 i270 Journal of Public Administration Research and Theory is included only in our cross-sectional citizen satisfaction models (tables 3 and 5) because it was only assessed comprehensively for all councils in 2002. Larger organizations may have a higher likelihood of leadership turnover because a bigger pool of potential replacements for the top leadership posts is readily available (Fredrickson, Hambrick, and Baumrin 1988, 258). As a proxy for the size of the organization, we include the size of the population living within the jurisdiction, drawing on the 2001 census. We use the natural log of this variable because of the very large values it takes on for a number of local governments. Finally, the location of the local government may be important. Levels of turnover may be higher in London and metropolitan authorities because there are many other neighboring local governments that could be an alternative source of employment for senior managers without requiring a household move, thus easing exit and replacement. We therefore include dummies for London and metropolitan boroughs as these are the major population centers of the country and are marked by a greater choice of occupational alternatives that are within commuting distance for local government senior managers. METHODS Our types of dependent variable differ: the senior management turnover rate in a given year is continuous, whereas the presence or absence of a chief executive succession in a given year is dichotomous. We use linear random effects models to estimate the specifications explaining the senior management turnover rate and a logit estimator for the specifications explaining chief executive succession. In both cases, we resort to Huber-White standard errors that correct for clustering on each of the 148 local governments. We have up to four observations on each of the 148 different local governments (from financial year 2002–03 up to and including financial year 2005–06), so we can exploit one of the major advantages of panel data: testing for the presence of omitted time-invariant factors that may bias our results. For this purpose, we use Hausman tests that statistically compare the results from a model that controls for all time-invariant differences between local governments (a fixed effects model) to the results from a model that assumes that there are no omitted time-invariant differences that affect the results (a random effects model). If the results from the two models do not differ statistically, it is defensible to use the random-effects model. It has two advantages. First, it allows the researcher to make statements about time-invariant explanatory variables that are swept out of fixed effects models. Second, it is more efficient—it can better discern existing but noisy relationships. The Hausman tests indicate that all of our models produce statistically the same results whether they are estimated by fixed or random effects. Finally, for all linear panel data models we carry out the test for serial correlation suggested by Wooldridge (2002) and implemented by Drukker (2003). Fortunately, this test indicates that none is present. FINDINGS Our findings are reported in tables 2–5. We discuss them in the order of our hypotheses. Boyne et al. Does Public Service Performance Affect Top Management Turnover? Table 2 Panel Linear Regression Models (Random Effects) Explaining the Senior Management Turnover Rate Model 2.1 Core service performance score Year-on-year change in service performance score CPA grade: fair/two stars (dummy) CPA grade: good/three stars (dummy) CPA grade: excellent/four stars (dummy) CPA same as or worse than last year (dummy) Whole council election (dummy) Election by thirds (dummy) Difference between vote share of two biggest parties in the last election Change in political party control (dummy) Chief executive succession last year (dummy) Chief executive succession 2 years ago (dummy) Logged population London borough (dummy) Metropolitan borough (dummy) Year dummy for 2004 Year dummy for 2005 Constant Wald chi-square test of H0: ‘‘The model explains nothing’’ Model 2.2 20.003 (3.04)** 0.000 (0.21) 0.036 (1.33) 0.020 (0.53) 20.001 (1.23) 20.142 20.081 20.141 20.015 0.038 0.017 20.001 20.004 (0.13) 0.075 (2.59)** 0.062 (2.11)* 0.019 (0.43) 0.059 (2.42)* 20.000 (0.01) 20.051 (2.18)* 0.027 (0.97) 0.282 (1.09) p , .0001** 20.012 (0.37) 0.074 (2.57)* 0.062 (2.06)* 0.027 (0.62) 0.052 (2.17)* 20.003 (0.12) 20.058 (2.62)** 0.036 (1.35) 0.136 (0.56) p , .0001** (4.30)** (2.55)* (4.74)** (0.67) (1.41) (0.50) (1.11) Note: Both models estimated over the same 431 observations on 148 local governments. The absolute values of Huber-White t-statistics (adjusted for clustering on local government) are provided in parentheses. *, indicates a p value of .05 or less; **, indicates a p value of .01 or less. Hypothesis 1 suggests that perceived organizational performance is negatively related to senior management turnover. Overall, our results for the central government assessments of performance (the core service performance score and the CPA) bear this out.9 As the first model in table 2 shows, there is a relationship between performance (at least the central government’s perceptions of it) and the senior management turnover rate. The mean senior management team turnover rate is about 0.2% or 20% (a fifth of the senior management team leaves each year on average) and the median size of the senior management team excluding the chief executive is six. As the first model in table 2 shows, a 10-point decrease in the core service performance score increases the senior management team turnover rate by three percentage points. All else equal, this would only increase the mean senior management team turnover rate to 23%, which would translate into a replacement of an additional manager only in teams greater than 15 (using the convention that a probability of replacement greater than 1/2 predicts the occurrence of a replacement). Consequently, the effect of the core service performance score on the senior management 9 Although we find evidence of a linear relationship, we also examined whether the relationship might be non-linear. It is conceivable that turnover might be higher at the top and bottom ends of performance than where performance is mediocre. We tested this by including a squared term of performance in all models using the service performance score and overall citizen satisfaction. In none of these models could the null hypothesis of no non-linear relationship be rejected. i271 i272 Journal of Public Administration Research and Theory Table 3 Cross-Sectional Linear Regression Model Explaining the Senior Management Turnover Rate Based on Local Citizens’ Satisfaction Satisfaction with the local government (percent very/fairly satisfied) in 2003 Change in per cent satisfaction with the local government from 2000 to 2003 Whole council election (dummy) Election by thirds (dummy) Difference between vote share of two biggest parties in the last election Change in political party control (dummy) Ability to improve score Chief executive succession last year (dummy) Chief executive succession 2 years ago (dummy) Logged population London borough (dummy) Metropolitan borough (dummy) Constant General F test of H0: ‘‘The model explains nothing’’ R2 0.002 (1.12) 20.005 (2.03)* 0.015 (0.20) 0.027 (0.33) 20.003 (1.77) 20.008 (0.14) 20.006 (2.14)* 0.092 (1.76) 0.051 (1.24) 0.048 (0.48) 0.067 (1.20) 20.043 (0.59) 20.028 (0.05) p 5 .003** 0.15 Note: Estimated over a cross-section of 141 local governments. The absolute values of Huber-White t-statistics are provided in parentheses. *, indicates a p value of .05 or less; **, indicates a p value of .01 or less. team turnover rate is more statistically than substantively significant, except for half a dozen local governments with the very largest senior management teams. On the other hand, the relationship between the CPA and the senior management team turnover rate (in the second model in table 2) is not only statistically but also substantively significant. Any of the three highest categories of the CPA predict a lower rate of turnover than the base group (low performance). On average, compared to the base group, councilyears with a CPA of two stars have a 14 percentage point lower rate of turnover; councilyears with a CPA of three stars have an 8 percentage point lower rate of turnover; and council-years with a CPA of four stars have a 14 percentage point lower rate of turnover.10 The interpretation of these findings is that on average, one additional senior manager will depart if a local government only achieves one star or zero stars on the CPA as opposed to two stars. Note that on average, there is no relationship between stagnation or worsening of the CPA and senior management team turnover. This implies that increases in the CPA do not lead to decreases in senior management team turnover and underlines that the relationship between the CPA and senior management turnover that we observe is driven by the contrast between poor performance on the one hand and at least mediocre performance on the other hand. The results in table 3 suggest that there is a less clear impact of the level of citizen satisfaction on turnover. Changes in local citizens’ satisfaction with their local government’s service provision show a negative relationship with the senior management turnover rate: local governments that saw a 10 percentage point reduction in citizen satisfaction (not an uncommon occurrence, as table 1 shows) on average had a five percentage point higher senior management turnover rate. Thus senior management team turnover appears to be 10 These reductions in turnover compared to the base group are not statistically identical: a joint test of the restriction that the coefficients on the three higher CPA category dummies are equal is rejected (p 5 .02). This is because the reduction in turnover is somewhat smaller for councils with three stars than for those with two or four stars. The reduction in turnover is identical for the latter: the hypothesis that the coefficients on the dummies for two stars and four stars are equal cannot be rejected (p 5 .99). Boyne et al. Does Public Service Performance Affect Top Management Turnover? Table 4 Panel Logit Models Explaining the Presence or Absence of a Chief Executive Succession Model 4.1 Marginal Effects Logit (Baseline Predicted Coefficients Probability 5 .157) Core service performance score Year-on-year change in service performance score CPA grade: fair/two stars (dummy) CPA grade: good/three stars (dummy) CPA grade: excellent/ four stars (dummy) CPA same as or worse than last year (dummy) Whole council election (dummy) Election by thirds (dummy) Difference between vote share of two biggest parties in the last election Change in political party control (dummy) Logged population London borough (dummy) Metropolitan borough (dummy) Year dummy for 2004 Year dummy for 2005 Constant Wald chi-square test of H0: ‘‘The model explains nothing’’ 20.041 (2.34)* 0.017 (0.77) Model 4.2 Marginal Effects Logits (Baseline Predicted Coefficients Probability 5 .150) 20.005 (2.37) 0.002 (0.77) 20.418 (1.20) 20.049 (1.31) 21.061 (2.87)** 20.129 (3.05) 21.070 (2.56)* 20.113 (3.18) 20.833 (2.92)** 20.123 (2.66) 0.157 (0.42) 0.021 (0.41) 0.228 (0.60) 0.030 (0.58) 0.128 (0.24) 0.017 (0.23) 0.171 (0.31) 0.023 (0.30) 0.022 (1.70) 0.003 (1.74) 0.026 (1.93) 0.003 (1.99) –0.106 (0.22) 20.014 (0.23) 20.259 (0.54) 20.031 (0.58) 0.058 (0.14) –0.569 (1.58) 0.008 (0.14) 20.068 (1.80) 0.286 (0.66) 20.703 (1.99)* 0.036 (0.66) 20.078 (2.34) 20.547 (1.46) 20.066 (1.66) 20.778 (2.12)* 20.086 (2.59) 20.514 (1.28) 0.446 (1.21) 0.808 (0.30) p 5 .04* 20.064 (1.35) 0.062 (1.17) 20.708 (1.67) 0.484 (1.34) 21.775 (0.70) p 5 .003** 20.084 (1.82) 0.065 (1.28) Note: Both models estimated over the same 431 observations on 148 local governments. For each model, the results are presented in two columns. The left-hand side column shows the raw coefficient estimates while the right-hand side column shows the marginal effects. Absolute values of Huber-White z-statistics (adjusted for clustering on local government) are provided in parentheses. *, indicates a p value of .05 or less; **, indicates a p value of .01 or less. i273 i274 Journal of Public Administration Research and Theory Table 5 Cross-Sectional Logit Model Explaining the Presence or Absence of a Chief Executive Succession Marginal Effects Baseline Predicted probability 5.158 Logit Coefficients Satisfaction with the local government (percent very/fairly satisfied) in 2003 Change in percent satisfaction with the local government from 2000 to 2003 Whole council election (dummy) Election by thirds (dummy) Difference between vote share of two biggest parties in the last election Change in political party control (dummy) Ability to improve score Logged population London borough (dummy) Metropolitan borough (dummy) Constant Wald chi-square test of H0: ‘‘The model explains nothing’’ 0.058 (1.58) 0.008 (1.76) 20.079 (1.65) 20.011 (1.81) 20.125 (0.13) 0.525 (0.58) 20.033 (1.31) 20.016 (0.13) 0.074 (0.54) 20.004 (1.33) 20.345 0.001 20.520 20.020 21.609 21.807 20.042 0.0002 20.069 20.003 20.164 (0.45) (0.03) (0.46) (0.02) (1.61) (0.28) (0.49) (0.03) (0.46) (0.02) (2.12) p 5 .49 Note: Estimated over a cross-section of 141 local governments. The results are presented in two columns. The left-hand side column shows the raw coefficient estimates while the right-hand side column shows the marginal effects. The absolute values of Huber-White zstatistics are provided in parentheses. *, indicates a p value of .05 or less; **, indicates a p value of .01 or less. more closely connected to central than local perceptions of service performance, which may reflect the high level of centralization in England.11 Hypothesis 2 suggests that chief executive successions will be followed by increases in senior management team turnover. There is quite a bit of evidence for this in our data: in table 2, a chief executive succession last year is related to a seven percentage point higher turnover rate this year, and a chief executive succession 2 years ago is related to a six percentage point higher senior management turnover rate.12 In the cross-sectional model using citizen satisfaction as the indicator of performance (reported in table 3), a chief executive succession in financial year 2002–03 is related to a nine percentage point higher turnover rate in 2003–04, though this finding is only weakly statistically significant (p 5 .08). Thus new chief executives appear to alter the composition of their senior management team. Such top-team shaping has been recognized as an important effect of new chief executives in the private sector (Klimoski and Kiechel Koles, 2001). In English local government, it tends to occur through reorganization of service departments, in part to circumvent job protections. Senior managers would not normally be easily dismissible by 11 It is also conceivable that the weakness of these findings is explained by the lack of clarity of responsibility in the rural counties, where some of the services are provided not by the county council but by a number of district councils. We tested this possibility by re-estimating the models in Tables 3 and 5 excluding the counties. The results in Table 3 are virtually unchanged, and in Table 5 the standard error on satisfaction slightly shrinks. The substantive conclusions from both tables remain unchanged. 12 We examined whether the relationship between lagged chief executive succession and senior management turnover is moderated by performance but there is no evidence of this in any model. Nor does chief executive succession, be it in the same year as the senior management turnover, in the year before, or in either year moderate the relationship between performance and the senior management turnover rate. Boyne et al. Does Public Service Performance Affect Top Management Turnover? a chief executive. Yet when their remits no longer exist because the chief executive has created new departments by merging or splitting existing ones, the posts of director of the new departments can be advertised and the existing director will just be one applicant among others. In field research conducted as part of the grant that this article stems from, we found a recently arrived chief executive to be doing exactly this. They were up-front about it and stated that they would prefer not to re-hire some of the existing service directors. We now proceed to test hypothesis 3. Recall our finding that the CPA has a substantively important effect on the senior management turnover rate—on average, one additional senior manager will depart if a local government only achieves one star on the CPA as opposed to two stars. Is there a similar finding for chief executives? At first glance, there is a statistically significant effect of the core service performance score and the highest categories of the CPA on the likelihood of a chief executive succession, but no significant effect of citizen satisfaction—the latter model performs no better than a constant-only model, as the overall Wald test at the bottom of table 5 shows. For citizen satisfaction, then, hypothesis 3 is corroborated. The two indicators of central government perceptions of performance are statistically related to both the senior management turnover rate and the likelihood of a chief executive succession (though only the CPA has a substantive effect on the turnover rate). For hypothesis 3 to be corroborated here, neither the core service performance score nor the CPA must have any substantive effect on the probability of a chief executive succession. To evaluate whether this is the case, we obtained predicted probabilities of a chief executive succession, holding all variables at their means. Now, the predicted probability of a chief executive succession is about 0.15 in both models in table 4. A substantively significant effect of either indicator of performance would move this probability above 0.5, producing a greater than even chance of a chief executive succession. Model 4.1 shows that the impact of a 10-percentage point decrease of the core service performance score on the probability of a chief executive succession is only 0.05; in other words, the probability of a chief executive succession increases by five percentage points but is still only 0.2% or 20%. This lack of a substantively significant effect of the core service performance score on the probability of a chief executive succession parallels the finding for the senior management turnover rate. Thus, the extent of support for hypothesis 3 depends on the results for the CPA. As model 4.2 (the second model in table 4) shows, the hypothesis passes the litmus test: whereas three or four stars on the CPA as opposed to zero or one star predict a lower probability of a chief executive succession statistically,13 substantively the probability of a chief executive succession in case of a zero or one star rating as opposed to a three or four star rating increases only to about 0.25 or 25 percent. Thus, it is still three times more likely for the chief executive to stay than to go in this circumstance, while one of the senior managers is expected to depart.14 In conclusion, hypothesis 3, which predicts that any negative relationship between performance and turnover would be weaker for chief executives than for other members of the senior management team, is corroborated in our study. This is unfortunate news for The two categories are statistically identical. Note that the predicted probability of a chief executive succession decreases by 0.12 in council-years that maintained their CPA rating or got worse. Although one interpretation of this finding is that it may be more difficult to attract new chief executives to such local governments, we would not wish to attach too much weight to a single unexpected relationship. 13 14 i275 i276 Journal of Public Administration Research and Theory those who think that ‘‘a fish rots from the head,’’ as the Russian proverb goes. From our findings, it appears that when it smells rotten, a gill is removed first, not the head. Few of our control variables show any statistical relationship with the senior management team turnover rate, and none do so in our chief executive succession models.15 Changes in political control are overall statistically unrelated to both dependent variables. Partly this is because these effects are weak overall, and partly this has to do with the nature of the data set used for this article, which covers both years with elections and years without elections. When a test for the effect of a change in political control on the senior management team turnover rate is conducted for only those observations when an election was held, a relationship does emerge (Boyne et al. 2010). Management capacity, as measured by the ability to improve score, shows a negative relationship with the senior management turnover rate in the cross-sectional model in table 3, where a one point greater ability to improve is related to a sixth of a percentage point lower rate of senior management turnover. The tenure of senior officials is, therefore, more secure in local governments that are judged by external inspectors as having appropriate structures and processes, regardless of citizen evaluations of performance. The Audit Commission inspectors appear to bestow legitimacy on senior management teams, protecting them from the repercussions of performance problems. Finally, there is some evidence that London boroughs tend to have a higher rate of turnover than local governments in the more rural parts of the country. The models in table 2 suggest that London boroughs tend to have a five percentage points higher rate of turnover than shire counties and unitary authorities. This may partly reflect the extra pressures on managers in the capital, or it may stem from a more plentiful supply of potential senior managers in this metropolis covered by 32 separate jurisdictions and a host of other public sector employers. CONCLUSIONS We have provided evidence for a connection between public service performance and the rate of turnover of appointed senior officials. Central government’s perceptions of performance, and to some extent also local citizens’ satisfaction with services, predict the rate of senior management turnover in English local governments. As we expected, lower perceptions of performance are associated with a higher rate of senior management turnover. On the other hand, there appears to be a weaker relationship between organizational 15 One possible concern about our findings is that there might be other reasons for senior managers to move on than the ones we test or control for here. In particular, senior management turnover might be higher in a local government when one or more senior managers: (a) hit retirement age or (b) move to another employer simply because the conditions there are better. The first concern might exaggerate our findings if local governments that experience low performance also tend to have a higher rate of senior managers reaching retirement age than other local governments. Unfortunately for our research, age is private information and, as such, protected under the UK’s Data Protection Act 1998. This information cannot be reliably obtained for all senior management teams while complying with the law. Yet it is very unlikely that age works to exaggerate our findings. If the senior management age structure varies between authorities but remains constant within authorities over the four years of our study, we know that, based on the Hausman test discussed above, this variable does not affect our results because they are statistically identical to those from a fixed effects model, which completely eliminates all time-invariant differences between authorities. If the age structure within local authorities varies over time, it is most likely to do so in a way that is not directly correlated with organizational performance. Therefore, any bias induced by such correlation would make it less likely that we still find a relationship between a change in service performance and the senior management turnover rate. The same reasoning applies to the possibility that senior managers move to another local authority simply because the conditions there are better, without any pressure from their old employer. Boyne et al. Does Public Service Performance Affect Top Management Turnover? performance and chief executive succession. This is consistent with arguments that chief executives deflect blame or attribute responsibility for poor results to their senior management teams who, like their private sector counterparts, are vulnerable to scapegoating. Our results show that new chief executives also have a direct effect on top team composition: the rate of senior management turnover tends to be higher in the wake of their arrival at the top of the organization. These results suggest that the tenure of senior managers in the public sector is partly contingent on organizational performance. Senior managers are more likely to remain in post when performance is perceived to be high, and more likely to depart when performance is perceived to be poor. Public managers are not insulated from performance, but are at least loosely accountable, either through forced or voluntary resignation, when public services are deemed to be unsatisfactory. The evidence implies that senior public officials have an incentive to achieve good public service results because this will make it easier for them to retain their current positions and make them more attractive to alternative employers. Similarly, they have an incentive to avoid poor results. This incentive is, however, sharpest not for the very top manager in our sample of organizations but for the other members of the senior management team. Our finding of a difference between the performance effect for chief executives and other senior managers also makes it unlikely that an alternative explanation drives these results, namely, that better-performing organizations are also better at retaining high-quality managers. All three of our performance indicators (the core service performance score, the Comprehensive Performance Assessment, and the citizen satisfaction percentage) are summary measures, covering perceptions of organizational performance as a whole. Consequently, they guard against the threat that our finding would be specific to only one or a few very dimensions of performance. Nevertheless, the relationship between performance and turnover clearly needs to be investigated in a variety of contexts before general conclusions can be drawn. It seems plausible that the strength of this relationship may depend on factors that vary across national and institutional settings, such as the clarity of performance information, the nature of employment contracts, the constitutional arrangements between elected politicians and appointed officials, and the impact of public service standards on electoral support for ruling parties. Another contingency might be the overall economic situation. We speculate that tight budgets weaken the relationships observed in our study, which covers a period of growing public-sector budgets. Austerity is likely to affect the ability of public organizations to both let go of current senior managers, which tends to require costly severance packages, and to hire new ones, which tends to require both substantial recruitment costs, and, more often than not, a substantial pay increase relative to the previous post holder. FUNDING We thank the Economic and Social Research Council Public Services Programme (RES166-25-0026). REFERENCES Andrews, Rhys, George A. Boyne, Jennifer Law, and Richard M. Walker. 2005. External constraints on local service standards: The case of comprehensive performance assessment in English local government. Public Administration 83:639–56. i277 i278 Journal of Public Administration Research and Theory Bigley, Gregory, and Margarethe Wiersema. 2002. New CEOs and corporate strategic refocusing: How experience as heir apparent influences the use of power. Administrative Science Quarterly 47:707–27. Boeker, Warren. 1992. Power and managerial dismissal: Scapegoating at the top. Administrative Science Quarterly 37:400–21. Boyne, George A., and Jay Dahya. 2002. Executive succession and the performance of public organizations. Public Administration 80:179–200. Boyne, George A., Oliver James, Peter John, and Nicolai Petrovsky. 2010. Does political change affect senior management turnover? An empirical analysis of top-tier local authorities in England. Public Administration 88:136–153. Boyne, George A., Kenneth J. Meier, Laurence J. O’Toole, Jr. and Richard M. Walker (eds). 2006. Public service performance: Perspectives on measurement and management. Cambridge: Cambridge Univ. Press. Boyne, George A., and Richard M. Walker. 2004. Strategy content and public service organizations. Journal of Public Administration Research and Theory 14:231–52. Brady, Gene F., and Donald L. Helmich. 1984. Executive succession: Toward excellence in corporate leadership. Englewood Cliffs, NJ: Prentice-Hall. Brickley, James A. 2003. Empirical research on CEO turnover and firm-performance: A discussion. Journal of Accounting and Economics 36:227–33. Carpenter, Mason, Marta Geletkanycz, and Gerard Sanders. 2004. Upper echelons research revisited: Antecedents, elements, and consequences of top management team composition. Journal of Management 30:749–78. Clarke, Harold D., David Sanders, Marianne C. Stewart, and Paul Whiteley. 2004. Political choice in Britain. Oxford: Oxford Univ. Press. Drukker, David M. 2003. Testing for serial correlation in linear panel-data models. Stata Journal 3:168–77. Farrell, Kathleen, and David Whidbee. 2003. Impact of firm performance expectations on CEO turnover and replacement decisions. Journal of Accounting and Economics 36:165–96. Feiock, Richard C., and Christopher Stream. 1998. Explaining the tenure of local government managers. Journal of Public Administration Research and Theory 8:117–30. Ferejohn, John A. 1986. Incumbent performance and electoral control. Public Choice 50:5–25. Fredrickson, James W., Donald C. Hambrick, and Sara Baumrin. 1988. A model of CEO dismissal. Academy of Management Review 13:255–70. Furtado, Eugene, and Vijay Karan. 1990. Causes, consequences, and shareholder wealth effects of top management turnover: A review of the empirical evidence. Financial Management 6:60–75. ´ Gutierrez-Romero, Roxana, Dirk Haubrich, and Iain McLean. 2008. The limits of performance assessments of public bodies: External constraints in English local government. Environment and planning C. Government and Policy 26:767–87. Hamidullah, Madinah, and Vicky M. Wilkins. 2007. Looking for their Dick Vermeil: How prior performance influences the selection decision. Presented at the Public Management Research Conference, Tucson, Arizona. Hill, Gregory C. 2005. The effects of managerial succession on organizational performance. Journal of Public Administration Research and Theory 15:585–97. Hood, Christopher. 1991. A public management for all seasons? Public Administration 69:3–19. ———. 2002. The risk game and the blame game. Government and Opposition 37:15–37. Hood, Christopher, Will Jennings, Ruth Dixon, Brian Hogwood, and Craig Beeston. 2009. Testing times: exploring staged responses and the impact of blame management strategies in two exam fiasco cases. European Journal of Political Research 48:695–722. Ingraham, Patricia W., Philip G. Joyce, and Amy Kneedler Donahue. 2003. Government performance: Why management matters. Baltimore, MD: The Johns Hopkins Univ. Press. James, Oliver. 2004. The UK core executive’s use of public service agreements as a tool of governance. Public Administration 82:397–418. Boyne et al. Does Public Service Performance Affect Top Management Turnover? ———. 2009. Evaluating the expectations disconfirmation and expectations anchoring approaches to citizen satisfaction with local public services. Journal of Public Administration Research and Theory 19:107–23. James, Oliver, and Peter John. 2007. Public management at the ballot box: Performance information and electoral support for incumbent English local governments. Journal of Public Administration Research and Theory 17:567–80. Kesner, Idalene, and Terrence Sebora. 1994. Executive succession: Past, present and future. Journal of Management 20:327–72. Klimoski, Richard J., and K. Lee Kiechel Koles. 2001. The chief executive officer and top management team interface. In The nature of organizational leadership: Understanding the performance imperatives confronting today’s leaders, eds. Stephen J. Zaccaro and Richard J. Klimoski. Hoboken, NJ: Wiley. Lowndes, Vivian. 2003. Between rhetoric and reality: Does the 2001 White Paper reverse the centralising trend in Britain? In Regulating local authorities, eds. P. Carmichael and A. Midwinter. London: Frank Cass. McCabe, Barbara, Richard Feiock, James Clingermayer, and Christopher Stream. 2008. Turnover among city managers: The role of political and economic change. Public Administration Review 68:380–6. Osborn, Richard, Lawrence Jauch, Thomas Martin, and William Glueck. 1981. The event of CEO succession: Performance and environmental conditions. Academy of Management Journal 24:183–91. Pfeffer, Jeffrey, and Gerald R. Salancik. 1977. Organizational context and the characteristics and tenure of hospital administrators. Academy of Management Journal 20:74–88. Pollitt, Christopher, and Geert Bouckaert. 2004. Public management reform. Oxford: Oxford Univ. Press. Puffer, Sheila, and Joseph Weintrop. 1991. Corporate performance and CEO turnover: The role of performance expectations. Administrative Science Quarterly 36:1–19. Simons, Robert. 1994. How new top managers use control systems as levers of strategic renewal. Strategic Management Journal 15:169–89. Stokes, Donald E. 1963. Spatial models of party competition. American Political Science Review 57:368–77. Tadelis, Steven. 2002. The market for reputations as an incentive mechanism. Journal of Political Economy 110:854–82. Turner, David, and Philip Whiteman. 2005. Learning from the experience of recovery: The turnaround of poorly performing local authorities. Local Government Studies 31:627–54. Wagner, Gary, Jeffrey Pfeffer, and Charles O’Reilly. 1984. Organizational demography and turnover in top management teams. Administrative Science Quarterly 29:74–92. Weaver, R. Kent. 1986. The politics of blame avoidance. Journal of Public Policy 6:371–98. Wooldridge, Jeffrey M. 2002. Econometric analysis of cross section and panel data. Cambridge, MA: MIT. i279