Prime ministers (PMs) play a key role in making parliamentary democracy work. In a principal–agent perspective, which conceives parliamentary democracy as a two-directional chain of delegation and accountability, the PM as chief executive ‘connects the elected representatives of the people and the administrators of the state’ (Strøm Reference Strøm2000: 270). More concretely, the PM is the agent to whom the parliamentary majority delegates the running of state affairs, which includes a variety of different tasks such as shaping government policy, managing crises and securing national interests abroad (Strangio et al. Reference Strangio, ‘t Hart and Walter2013; ‘t Hart and Schelfhout Reference ‘t Hart and Schelfhout2016). To carry out these tasks, the PM relies on cabinet ministers as her agents for developing policies and directing the bureaucracy in the government departments. Ministers also provide advice and support for the PM in her relations with other political actors, such as coalition parties and parliament. Hence, the performance of PMs crucially depends on whether the ministers fulfil their responsibilities in an effective and accountable manner.
Against this background, cabinet reshuffles – defined as personnel changes in the ministerial composition of a government – may significantly affect prime-ministerial performance. Since new ministers need considerable time to guide and control the civil servants effectively in their departments (Rose Reference Rose1971: 407), ministerial turnover tends to undermine the stability and functional capacity of the cabinet that a PM needs to fulfil her tasks properly.Footnote 1 Of course, PMs may also change incompetent or self-interested ministers in order to regain political control over their government and thus improve their performance. Such reshuffles are well known from Westminster democracies where PMs occasionally use their extraordinary political power to fire and hire cabinet members during the legislative period (Indriðason and Kam Reference Indriðason and Kam2008). However, the effects of ministerial turnover on prime-ministerial performance have not been systematically explored so far.
Our article investigates how cabinet reshuffles influence the performance of PMs in Central and Eastern Europe (CEE). It contributes to the extant literature in several respects. First, we argue that ministerial turnover has an overall negative effect on whether the PM succeeds in accomplishing her various office tasks and confirm this assumption by analysing a novel data set of 131 cabinets in 11 CEE democracies between 1990 and 2018. This finding hence adds to the small set of research engaging with the consequences of reshuffles, notably on government popularity (Dewan and Dowding Reference Dewan and Dowding2005; Miwa Reference Miwa2018) or policy outcomes (Huber Reference Huber1998), and goes beyond the bulk of relevant studies that deal with the determinants of cabinet reshuffles (Berlinski et al. Reference Berlinski, Dewan and Dowding2010; Fischer et al. Reference Fischer, Dowding and Dumont2012; Hansen et al. Reference Hansen2013).
Second, we offer a nuanced perspective on how cabinet reshuffles affect the success of the PM in office in multiparty governments. According to our findings, changes in ministers who belong to the PM's coalition partners appear to induce the highest cost for her performance, while changes in ministers from her party or those in charge of core portfolios are less consequential. These results add to our understanding of the intra-cabinet determinants of prime-ministerial performance and enrich the extant literature that focused on the effects of PMs' careers (Grotz et al. Reference Grotz2021) as well as their relationships with their principals – that is, the parliamentary majority and their own parties (Samuels and Shugart Reference Samuels and Shugart2010; Strangio et al. Reference Strangio, ‘t Hart and Walter2013).
Third, our article sheds new light on the differentiated impact of government stability in the post-communist democracies of CEE where cabinets have been relatively short-lived compared to Western democracies (Grotz and Weber Reference Grotz and Weber2012; Somer-Topcu and Williams Reference Somer-Topcu and Williams2008). While previous research found that prime-ministerial performance does not strongly correlate with cabinet duration in the CEE context (Grotz et al. Reference Grotz2021), our analysis shows that it is significantly affected by the frequency of ministerial turnover. Hence, the findings contribute to the debate on which forms of government instability are more detrimental for the functioning of political executives (Huber Reference Huber1998; Huber and Martinez-Gallardo Reference Huber and Martinez-Gallardo2008).
The impact of cabinet reshuffles on prime-ministerial performance
In light of the outstanding position of chief executives in parliamentary democracies, PMs have to fulfil a wide range of tasks (Strangio et al. Reference Strangio, ‘t Hart and Walter2013; ‘t Hart and Schelfhout Reference ‘t Hart and Schelfhout2016). The characterization of parliamentary democracy as a chain of delegation and accountability between voters, legislature, executive and state administration (Strøm Reference Strøm2000) indicates that prime-ministerial tasks fall into two broad dimensions (Grotz et al. Reference Grotz2021: 4–5). First, the PM is responsible for running the state affairs, which includes managing the cabinet, shaping government policies, managing exogenous crises and securing the national interests at the international level. Second, as agent of the parliamentary majority and her own party, the PM has to secure the support of these two principals. We build our theoretical argument on this concept of prime-ministerial performance, as it provides a comprehensive understanding of the position of the PM vis-à-vis other actors, as well as offering a differentiated catalogue of tasks that a PM ought to accomplish.
Previous research highlights the role of political experience as well as the party constellation in government as determinants of a PM's success in fulfilling these manifold tasks. For instance, Florian Grotz et al. (Reference Grotz2021) find that PMs who were party leaders before coming to office perform considerably better than their counterparts without such experience. Moreover, conflicts among key party elites or disunity of the parliamentary party group were found to destabilize the PM's political position (Baylis Reference Baylis2007). Minority cabinets and multiparty governments further increase the complexity and uncertainty of the PMs' political environment, which might also affect their performance (Bergman et al. Reference Bergman, Ersson and Hellström2015).
Beyond these factors, the PM relies on the cabinet ministers as heads of government departments, and by that her immediate political agents, to accomplish her tasks. Therefore, frequent movements of ministers within the cabinet as well as their departures from the cabinet may undermine prime-ministerial performance. Ministers have, or are expected to acquire, substantive expertise and political skills to lead their portfolios (Dewan and Hortala-Vallve Reference Dewan and Hortala-Vallve2011; Indriðason and Kam Reference Indriðason and Kam2008). The value of ministerial expertise for the PM is not restricted to the development of policy along her own agenda, but also extends to other tasks delegated to the chief executive, such as crisis management and foreign affairs. Concerning the former, the PM can at best provide an interpretation of the crisis event and strategic direction for the response, but the ministers are responsible for the operational level of crisis management. Similarly, the head of government is often engaged in relations with other countries or international organizations, but depends on the foreign minister and her office for relevant information and advice. Finally, ministers also support the PM in parliament and at meetings of her own party.
Cabinet reshuffles diminish these ministerial contributions to prime-ministerial performance. Most obviously, ministers who remain in office for short time periods cannot develop the substantial policy expertise on which the PM depends (Bright et al. Reference Bright, Döring and Little2015; Hansen et al. Reference Hansen2013). Short incumbency by ministers hinders their control of civil servants within the department, who might often have policy preferences that diverge from those of the minister and the PM (Huber Reference Huber1998: 578). Ministerial reshuffles may also contribute to tensions between the PM, her own party and the coalition partners, as the original line-up of cabinet often results from painstaking negotiations during government formation (Bovens et al. Reference Bovens, Brandsma and Thesingh2015; Fischer et al. Reference Fischer, Dowding and Dumont2012).
Despite the benefits of ministerial continuity, PMs in parliamentary democracies are usually confronted by cabinet reshuffles that are beyond their control. Ministers might leave the cabinet voluntarily for non-political reasons, such as health problems, retirement or an interest in more attractive career options outside politics (Bovens et al. Reference Bovens, Brandsma and Thesingh2015). Or there are political reasons for ministerial resignations such as disagreements with government policy or scandals that involve the minister herself or her close associates (Dewan and Dowding Reference Dewan and Dowding2005). Aside from the PM, cabinet ministers have several other principals who might exert pressure on them (Bucur Reference Bucur2017). Ad hoc parliamentary majorities composed of coalition partners and opposition might react to policy disagreement and scandals by triggering a parliamentary motion of no-confidence against a minister that forces her to step down. In semi-presidential systems where state presidents have considerable powers over ministerial appointment and dismissal, ministers might fall victim to an intra-executive conflict between PM and president (Bucur Reference Bucur2017; Sedelius and Mashtaler Reference Sedelius and Mashtaler2013).
In some cases, the PM might not only be a passive bystander in cabinet reshuffles, but assume a more active role in firing and hiring ministers. Nevertheless, the evidence for such active reshuffling predominantly originates from Westminster parliamentary democracies, which are characterized by single-party governments headed by party-leader PMs. Within such ideal-type parliamentary chain of delegation and accountability, the PM might be freer to deliberately reshuffle the cabinet to address serious ministerial drift (Dewan and Dowding Reference Dewan and Dowding2005; Indriðason and Kam Reference Indriðason and Kam2008). Such a move could help the PM to realign government departments with her agenda and ensure the accountability of individual ministers, which enhances prime-ministerial performance. However, in multiparty governments, the PM frequently meets resistance from her own party, coalition partners or parties supporting a minority government, which may threaten to bring down the whole cabinet (Berlinski et al. Reference Berlinski, Dewan and Dowding2010: 561). For these reasons, cabinet reshuffles initiated by the PM are quite rare events in most parliamentary democracies. Furthermore, active cabinet reshuffles remain rather costly for the PM even when the minister is eventually removed from the cabinet. For instance, the PM may not be successful in finding a suitable replacement within or outside the cabinet (Dewan and Myatt Reference Dewan and Myatt2010). Moreover, frequent firing of ministers by the PM might send a signal to the remaining ministers that the PM would not protect them in the event of public scrutiny, which might hamper their policy activism (Dewan and Myatt Reference Dewan and Myatt2007). Hence, since active reshuffling by PMs also bears considerable costs, we expect cabinet reshuffles to have an overall negative effect on prime-ministerial performance.
Hypothesis 1: The higher the number of cabinet reshuffles, the less successfully will PMs perform.
Furthermore, the negative effect of cabinet reshuffles on prime-ministerial performance may be particularly pronounced in specific constellations, which can be derived from two models of cabinet governance. The ‘ministerial government model’ assumes that ministers have a wide scope of influence over their departments, which ensures credibility and accountability in policymaking (Laver and Shepsle Reference Laver and Shepsle1990). Consequently, prime-ministerial performance primarily depends on the relation between the PM and the ministers heading the most powerful departments.
Following the ministerial government model, we argue that ministerial reshuffles in some portfolios, such as finance, foreign relations, defence and internal affairs, are more consequential for prime-ministerial performance than others. The outstanding role of these core portfolios derives from their control over large resources, responsibility for the key functions of the state, as well as coordination or even veto rights over policy initiatives from other portfolios (Druckman and Warwick Reference Druckman and Warwick2005). Running such a portfolio requires political expertise and acumen, which are lost once the minister departs from the cabinet. As new ministers require time to familiarize themselves with these complex portfolios, policymaking and bureaucratic control might be less effective following reshuffles. Additionally, core portfolios carry more political weight within and outside the cabinet, so they are usually headed by more important politicians. The departure of these heavyweights from cabinet might trigger instability in the relations between the PM and her supporting parties in parliament (Bright et al. Reference Bright, Döring and Little2015; Hansen et al. Reference Hansen2013). Taken together, we expect reshuffles in core portfolios to undermine prime-ministerial performance to a greater extent as compared to reshuffles in other portfolios.
Hypothesis 2: Reshuffles of ministers in core portfolios affect prime-ministerial performance more negatively than reshuffles of ministers in other portfolios.
The ‘coalition compromise model’, in turn, proposes that government policy is developed collectively between coalition partners with the help of ex ante (coalition agreement) and ex post (monitoring and control) instruments (Muller et al. Reference Muller, Bergman, Ilonszki, Bergman, Ilonszki and Müller2019; Strøm et al. Reference Strøm, Müller and Smith2010). According to this model, prime-ministerial performance is crucially affected by the relationship of the PM and the ministers of her party on the one hand and the ministers of the other coalition parties on the other hand. Following a cabinet reshuffle, it is much easier for the PM to impose her political agenda when new ministers come from her own party and should hence share the same basic policy goals (Saalfeld Reference Saalfeld2000: 357). In cases of reshuffles in portfolios controlled by other coalition parties, the PM has much less leverage over the newly appointed ministers. Hence, the adversarial effect of cabinet reshuffles on prime-ministerial performance will likely be diminished when ministers from the PM's party are reshuffled.
Hypothesis 3: Reshuffles of ministers from other coalition parties affect prime-ministerial performance more negatively than reshuffles of ministers from the PM's party.
Research design
We test these propositions on data for 131 cabinets in 11 new democracies in CEE between 1990 and 2018.Footnote 2 We identify new cabinets if the PM changes, new elections take place or the party composition of the government changes (King et al. Reference King1990: 856). Interim and caretaking governments are not included in the data, since PMs have limited options to perform in these cases and are thus not comparable with full governments. Moreover, we excluded cabinets that lasted less than six months to ensure that PMs had sufficient time to show their performance in office, as well as to ensure that experts were able to gather sufficient information to assess the performance of PMs.Footnote 3
Dependent variable
To operationalize prime-ministerial performance, we make use of an index introduced by Grotz et al. (Reference Grotz2021). The measure captures a PM's success in office defined as the ability to manage the cabinet and direct domestic affairs as well as to ensure the permanent support of the parliamentary majority and the PM's party. Overall, 11 aspects of prime-ministerial performance grouped in six subdimensions form the measure:
(1) Settling cabinet conflicts (prevailing in PM–minister conflicts, mediating inter-ministerial conflicts)
(2) Shaping government policies (enacting preferred policies for societal concerns, averting non-preferred policies)
(3) Managing exogenous crises (strategizing crisis response, responding to crises in appropriate time)
(4) Securing national interests abroad (securing national interests toward other countries, securing national interests towards the European Union)
(5) Maintaining support of own party (maintaining support of party elites, maintaining support of party base)
(6) Maintaining support of parliamentary majority (maintaining support for government policy).
Each aspect was operationalized by a survey item and about 20 experts per country assessed the performance of the PM in a given cabinet along these aspects. The exact wording of all items is described in Table A1 in the Online Appendix. Next to the headings of the subdimensions in the table, we also report the correlations between individual indicators that make up a specific subdimension of prime-ministerial performance. The substantial but imperfect positive correlations reveal that the individual indicators validly capture a particular subdimension. The responses to subdimensions are weighted equally to create an additive index of prime-ministerial performance. A more detailed inspection of the internal consistency of the index is available from the authors in Florian Grotz et al. (Reference Grotz2021: appendix D). Theoretically, the index ranges between 0 and 4. Empirically, we observe average ratings for prime-ministerial performance in a cabinet between 1.07 (Gyurcsany III in Hungary) and 3.45 (Nastase I in Romania, Orban II in Hungary).Footnote 4 The mean evaluation of all PMs is 2.48 and the standard deviation 0.53.Footnote 5
Independent variables
Four different variables capture the number of reshuffles: the overall frequency of reshuffles, frequency of reshuffles in core portfolios, frequency of reshuffles of PM party ministers and frequency of reshuffles of coalition partner ministers. All variables build on a full list of cabinet members, including the names of all ministers, portfolio(s), party belonging and the time each person remained in a given position.Footnote 6 To operationalize the overall reshuffles, we counted the number of appointees who took over a portfolio from another person. This is a conservative estimate of reshuffles, since it excludes ministers without portfolio and those cases in which a new portfolio was first assigned to a person when the cabinet was already in office. To identify the reshuffles in core portfolios, we follow the same strategy but count only reassignments in the areas of defence, interior, foreign affairs, finance and economy. While the salience of portfolios certainly varies over time and between countries, these five portfolios appear to be of the highest relative importance (Krook and O'Brien Reference Krook and O'Brien2012).Footnote 7 The variable reshuffles of PM party ministers contains the number of reassignments of ministers who belong to the party of the PM, while the reshuffles of coalition partners' ministers captures the number of reassignments of ministers who neither belong to the party of the PM nor are independent. To make the number of reshuffles comparable over cabinets with different durations (186 to 1,491 days), we standardized the count of reassignments per 200 days.Footnote 8
Table 1 provides summary statistics for all independent variables. The overall frequency of reshuffles per 200 days ranges between 0 and 8.333. A total of 10 cabinets did not witness any reshuffles.Footnote 9 At the other extreme, the Ponta I cabinet reshuffled 8.333 per 200 days, which equals 9 replacements over its 216 days in office. On average, 1.695 positions were reshuffled within 200 days. For the nuanced counts of reshuffles, the average value and maxima are lower. Reshuffles of ministers who belong to the coalition partner occur least frequently, with an average of 0.459 replacements per 200 days, compared to 0.608 replacements of PM party ministers and 0.519 reshuffles in core portfolios. As Figure 1 clarifies, the distributions for all variables counting the frequency of reassignments are strongly skewed to the right, with only a few observations with very high counts. In particular for ministers not belonging to the PM's party and in charge of core portfolios, larger numbers of reassignments are extremely rare.
Figure 1. Kernel Density of the Frequency of the Four Variables for Reshuffling (Overall, Core Portfolios, PM Party Ministers, Coalition Partner Ministers)
Notes: Based on average data at cabinet level, bandwidth = 0.2.
Table 1 Descriptive Statistics of Main Independent Variables
Control variables
All models include a series of control variables at the cabinet and country level that might impact both cabinet reshuffles and prime-ministerial performance. Moreover, we introduce country- and decade-fixed effects to capture differences in the relationship between reshuffles and prime-ministerial performance that occur due to country-specific factors or temporal dynamics. Additionally, we introduce further control variables as a robustness check.Footnote 10
Overall reshuffles at same day
Multiple reshuffles at a single point in time might indicate that the PM is able to make a clean break after continuous struggles within cabinet. In such a scenario, we observe a comparably high number of reshuffles, but the liberating effect for the PM might actually enhance her performance in office. These rather rare but influential events might lead to an incorrect (depressed) estimation of the average linear effect of cabinet reshuffles on prime-ministerial performance. We therefore control for the timing of reshuffles by including a variable that counts the overall reassignments that occurred on the same day and identified the maximum value per cabinet. We observe that up to seven replacements took place the same day, but the average is considerably lower with 2.160 simultaneous events.
Single-party cabinets
Cabinet reshuffles are considerably more likely to occur in single-party than in coalition cabinets (Berlinski et al. Reference Berlinski, Dewan and Dowding2010: 561; Huber and Martinez-Gallardo Reference Huber and Martinez-Gallardo2008; Kam and Indriðason Reference Kam and Indriðason2005: 336). At the same time, it might be easier for PMs to perform successfully in single-party cabinets than coalitions, since governing with multiple parties entails higher transaction costs and higher complexity, as PMs are required to negotiate continuously with other coalition partners (Grotz and Weber Reference Grotz and Weber2012; King et al. Reference King1990: 858). To take this factor into account, we introduce a control variable that takes the value ‘0’ for coalition cabinets and ‘1’ for single-party cabinets (based on Döring and Manow Reference Döring and Manow2019).
PM appointment and dismissal powers
The PM's institutional powers to appoint and dismiss individual ministers are a key instrument for proactive cabinet reshuffles. However, the degree to which the PM can make such decisions without consulting the head of state or parliament varies across countries (Andeweg Reference Andeweg2000: 381–382). At the same time, these powers also shape the PM's ability to perform successfully in office, since the institutional empowerment of multiple actors creates a more complex and uncertain environment (Elgie Reference Elgie and Helms2012; Goplerud and Schleiter Reference Goplerud and Schleiter2016). We therefore include an additive index for prime-ministerial powers to reshuffle ministers based on Bairett (Reference Bairett2015). The index is a summary of two measures: the power of the PM to appoint ministers (ranging between 0, where the head of state appoints cabinet without assembly confirmation, to 4, where the head of government appoints cabinet without assembly confirmation) and the power of the PM to dismiss ministers (ranging between 0, where the cabinet or ministers can be removed by the head of state only, and 4, where the assembly may not remove ministers). Hence, the index ranges from 0 to 8, with 0 indicating extremely limited powers, while 8 stands for unrestrained powers.
Number of ministers in a cabinet
Larger numbers of reshuffles are more likely to occur in cabinets with many ministers (Kam and Indriðason Reference Kam and Indriðason2005: 343). At the same time, we expect that performing successfully, in particular managing cabinet conflicts, is considerably easier in small cabinets in which responsibilities are less likely to overlap. All models therefore control for the absolute number of ministerial posts per cabinet, approximated as the number of the first portfolios listed for a minister in a cabinet, excluding ministers without portfolio.
Empirical analysis
To study the effects of the frequency (of different types) of reshuffles on prime-ministerial performance, we first explore the bivariate relationship between these variables. Figure 2 displays scatter plots and fits for prime-ministerial performance and the four cabinet reshuffle variables. In all figures, the linear and quadratic fits differ considerably for large numbers of replacements, suggesting that the effects of these variables on prime-ministerial performance change as the number of reshuffles increases.
Figure 2. Scatter Plot and Fit of Prime-Ministerial Performance and the (a) Reshuffles, (b) Reshuffles in Core Portfolios, (c) Reshuffles of PM Party Ministers and (d) Reshuffles of Coalition Partner Ministers
Notes: Based on cabinet-level data. Solid black line shows linear fit. Dashed black line shows quadratic fit. Dashed grey line shows median spline.
The top-left figure in Figure 2 displays the relationship between the overall number of reshuffles and prime-ministerial performance. The linear fit shows no clear relationship between the two variables (r = −0.11, p = 0.23), but the quadratic fit suggests a U-shaped relationship. PMs whose cabinets experience around three reshuffles perform less successfully than those who experience no reshuffles, but PMs leading cabinets with more than 5 reshuffles per 200 days perform considerably more successfully than those with fewer reshuffles. Such extremely high numbers of replacements are rare in the CEE context and occur only in Romania. One of these large turnovers took place in Adrian Nastase's second cabinet in connection with the country's accession to the European Union (Euroactiv 2004). In March 2004, the PM dismissed the minister for justice after EU officials severely criticized the lack of independence in the Romanian judiciary, and also reshuffled several other cabinet members to prove his ability to implement the radical reforms required to speed up the accession process. The largest reshuffle happened in Victor Ponta's first cabinet in summer 2013 when the fierce power struggle between the PM and President Traian Basescu culminated in a recall referendum against the president (Stan Reference Stan2013). As the minister of the interior had to admit critical mistakes in the referendum vote count and stepped down on 6 August, Ponta took the initiative and changed six other ministers on the very same day to show that he still held the reins of power. These PMs were able to carry out a reshuffle on their own terms and increase their performance as a result, which reflects many such scenarios in Westminster democracies (Dewan and Dowding Reference Dewan and Dowding2005). However, in the context of CEE, such cases of clean breaks by PMs are rather exceptional.
The top-right figure shows the bivariate relationship between reshuffles in core portfolios and prime-ministerial performance. The linear relationship between the two variables is close to zero (r = 0.01, p = 0.88) and the quadratic fit does not suggest a strong effect of replacements in core portfolios on prime-ministerial performance. The descriptive evidence hence leads us to reject Hypothesis 2.
The figures at the bottom illustrate the effects of replacement of ministers belonging to the PM's and the coalition partners' parties. The lower-left figure shows how replacements of ministers belonging to the PM party affect performance. The linear fit indicates no relationship between the two variables (r = −0.11, p = 0.20), while the quadratic fit indicates that, as the number of reshuffles of PM party ministers increases, PMs perform less successfully. However, this pattern is driven by two idiosyncratic cases in which the PM's position was further weakened by frequent ministerial turnover from his own party. The first is a minority cabinet in Poland formed by the Election Action Solidarity (AWS) that served under PM Jerzy Buzek from June 2000 until the general elections in September 2001. Since the AWS was a heterogeneous alliance of ideologically different organizations with Solidarity chairman Marian Krzaklewski as its leader, Buzek did not proactively reshuffle his cabinet but was rather faced with numerous ministerial replacements that were triggered by the AWS parliamentary faction or by ministers joining new parties in the run-up to the elections and leaving the cabinet (Jasiewicz and Jasiewicz-Betkiewicz Reference Jasiewicz and Jasiewicz-Betkiewicz2002). The other case is a cabinet in Romania led by Mihai Tudose who was installed as PM in June 2017 by the powerful chairman of the Social Democratic Party (PSD) Liviu Dragnea, who was not allowed to become head of government himself due to a suspended prison sentence (Stan and Zaharia Reference Stan and Zaharia2018). As several PSD ministerial positions were filled by loyal friends to Dragnea, Tudose replaced them in order to emancipate himself from Dragnea's predominance. However, this cabinet reshuffle did not strengthen the PM's position but rather increased tensions between him and the PSD leader, which eventually resulted in Dragnea replacing Tudose in early January 2018.
The lower-right figure, which displays the effect of reshuffles of ministers belonging to coalition partners' parties on prime-ministerial performance, shows just the opposite pattern. There is no linear relationship between PM's success in office and this type of reshuffling (r = 0.01, p = 0.99), but the quadratic fit suggests a U-shaped relationship. Increasing numbers of reshuffles coincide with decreasing prime-ministerial performance, until a turning point at around 1.5 reshuffles per 200 days, at which point the direction of the relationship turns from negative to positive. Again, frequent reassignments are extremely rare and only observed in the second Nastase cabinet mentioned above. The quadratic bivariate evidence hence lends support for Hypothesis 3, with the exception that some outliers might occur.
To test whether the observed patterns change when considering potential confounders, we estimated multiple linear regression models. We include square terms for all reassignment variables following the bivariate finding that the relationship between reshuffles and performance is likely to be non-linear. Moreover, we calculate cluster-robust standard errors in which individual expert ratings (N = 2,572) are nested in cabinets (N = 131), as recommended for estimated dependent variables by Jeffrey Lewis and Drew Linzer (Reference Lewis and Linzer2005). This allows us to capture the uncertainty about the exact performance of PMs inherent to the expert data.
Table 2 displays results for two regression models. The first model includes the overall number of reshuffles as well as reshuffles in core portfolios, while the second one includes the variables capturing reshuffles of ministers belonging to the PM's or coalition partners' parties. We do not introduce all explanatory variables into a single model, since the latter perfectly predict the former.
Table 2 Linear Regression of Different Types of Reshuffles on Prime-Ministerial Performance
Notes: * p < 0.05, ** p < 0.01, *** p < 0.001. A confidence level of p < 0.10 does not occur. Standard errors are clustered at the cabinet level. Country- and period-fixed effects are omitted from presentation for reasons of simplification.
Starting with the general effect of reshuffles, Model 1 supports the observation based on Figure 2 that reshuffles decrease prime-ministerial performance. Compared to the bivariate analysis, the multivariate models demonstrate a more pronounced negative linear effect of cabinet reshuffles on prime-ministerial performance. The stepwise inclusion of control variables into multivariate model indicates that controlling for the number of reshuffles at the same day particularly clarifies this effect. However, for the few cases in which large numbers of reshuffles occur, this effect turns positive. Figure 3 reveals how the linear and quadratic effects play out together by showing the predicted probability and marginal effects plots. Cabinets without any reshuffles perform considerably better than those with a few reassignments. If the number of overall reshuffles in a cabinet increases from 0 to 1.2 per 200 days (which equals one standard deviation), prime-ministerial performance decreases about 0.33 points (which equals about two-thirds of a standard deviation). However, with every additional reshuffle, the disadvantage for prime-ministerial performance emerging from ministerial reassignments shrinks, and reaches 0 by about 4.5 replacements. For the four cases that lie above this threshold, the model predicts increases in performance if the number of reshuffles increases even further. The large confidence intervals show that this effect for large values of replacements is subject to high levels of uncertainty. However, for the large majority of cases the model lends support for Hypothesis 1, which proposes that higher levels of ministerial turnover lead to lower levels of prime-ministerial performance.
Figure 3. Linear Prediction and Marginal Effects Plot of Overall Reshuffles of Ministers (Based on Model 1)
Notes: Grey shaded areas indicate 95% confidence intervals.
Models 1 and 2 also underpin the descriptive finding that there is no clear link between reshuffles in core portfolios and prime-ministerial performance. While the coefficient of the linear effect is slightly positive, it remains weak and does not reach conventional levels of statistical significance. This finding leads us to reject Hypothesis 2, which suggested that the more core portfolios are reshuffled, the less successfully will a PM perform.
Turning to reshuffles of PM party and coalition partner ministers, the regression model shows distinct effects depending on the type of reassignments. Increases in the number of reshuffles of ministers belonging to the PM's party appear to have no clear effect on prime-ministerial performance. Increases in the number of reshuffles of ministers belonging to the coalition partner, by contrast, have a strong negative linear and a weak positive quadratic effect on prime-ministerial performance. Figure 4 clarifies how the combination of the linear and quadratic effects impacts prime-ministerial performance, by zooming into cases with zero to 1.8 reshuffles, which is the large majority of all cases. Within this range, increasing numbers of reassignments of ministers belonging to the PM's party do not significantly change how successfully PMs perform in office. Indeed, a higher fluctuation of ministers belonging to the coalition partners decreases prime-ministerial performance significantly. If, for instance, the number of reshuffles of coalition partner ministers per 200 days increases from zero to 0.82 (one standard deviation), the predicted value for prime-ministerial performance increases 0.36 points (two-thirds of a standard deviation). Even though the weak negative effect we expected to find for reshuffles of PM party ministers did not occur, this pattern lends support to Hypothesis 3. Overall, these insights clarify that the negative effect of overall reshuffles observed above is mostly driven by reshuffles of coalition partner ministers.
Figure 4. Linear Prediction of Reshuffles of PM Party and Coalition Partner Ministers (Based on Model 2)
Notes: Grey shaded areas indicate 95% confidence intervals. Marginal effects plots of PM party and coalition partner ministers on prime-ministerial performance are displayed in Figure A1 in the Online Appendix. Linear predictions and marginal effect plot over the whole range of observed values for reshuffles are displayed in Figure A2 and A3 in the Online Appendix.
Turning to the explanatory power of the models, the adjusted R-squared values range between 0.08 and 0.10, indicating that, beyond reshuffles, a broad variety of factors driving prime-ministerial performance deserve further attention in future research.
These findings are robust to a series of modifications of the models. To begin with, causality might be inverse, with low prime-ministerial performance motivating PMs to reshuffle a lot to strengthen their position in cabinet when they are under political pressure. If this were the case, the number of reshuffles actively enforced by PMs should be strongly linked to prime-ministerial performance, while reshuffles that the PM has to accept more passively (e.g. occurring for minister's personal reasons such as health issues) should have no clear link to prime-ministerial performance. To capture this dynamic, we coded a variable that identifies active and passive reshuffles and introduced it as an additional variable into our model.Footnote 11 Overall, passive reshuffles appear to occur much more frequently in CEE than active reshuffles (62.89% compared to 37.11%). On average, a cabinet experiences 0.95 passive reshuffles per 200 days, but only 0.56 active ones, meaning that reshuffles that PMs cannot enforce are twice as frequent as those enforced by the PM. The results of the additional models indicate that there is no clear relationship between prime-ministerial performance and the frequency with which active reshuffles occur – while the effects of overall reshuffles and of coalition partner ministers remain stable if this variable is introduced into the models (see Test 1.1 Table A2 and 2.1 in Table A3 in the Online Appendix). This insight leads us to conclude that reshuffles affect prime-ministerial performance, not the other way around.
Moreover, we introduced a series of control variables to capture further dynamics at the cabinet level. In these tests, we first take into account that portfolios might move from one party to another during a reshuffle, thereby changing the power balance within cabinet and possibly shaping the chances of PMs to perform successfully in office. We introduce a confounder that counts the number of reshuffles that involve changes in party composition and in which the portfolio moves from a coalition partner to the PM's party (see Test 1.2 Table A2 and 2.2 in Table A3).Footnote 12 Second, we investigate whether the reshuffle of independent ministers confounds the effects of the reshuffles of ministers belonging to the PM's party and coalition partners (see 2.3 in Table A3).Footnote 13 Third, we add a dummy variable to the models that takes the value ‘1’ if not a single replacement occurs and ‘0’ for all other cases (see Test 1.3 Table A2 and 2.4 in Table A3). Possibly, the absence of any reshuffles might be a unique scenario that boosts prime-ministerial performance and leverages predicted values for low values of replacements as well. Fourth, we include a continuous variable counting the number of coalition parties instead of a dummy variable for single-party cabinets (see Test 1.4 Table A2 and 2.5 in Table A3). This modification allows us to identify the changes that occur if the number of coalition partners increases. In a fifth test, a binary variable identifies majority cabinets as opposed to minority cabinets to uncover whether the more reliable parliamentary support of majority cabinets impacts the observed patterns (see Test 1.5 Table A2 and 2.6 in Table A3), while a sixth test identifies such dynamics for surplus coalitions compared to non-surplus coalitions (see Test 1.6 Table A2 and 2.7 in Table A3). Seventh, we add a binary variable for post-electoral cabinets as opposed to replacement cabinets to the models (Grotz and Weber Reference Grotz, Weber, Harfst, Kubbe and Poguntke2017), to see whether the effect of reshuffles might be driven by different dynamics in these settings (see Test 1.7 Table A2 and 2.8 in Table A3). Eighth, the ideological range between the cabinet parties allows us to test for omitted variable bias induced by the potential for political conflict emerging between coalition partners (Jäckle Reference Jäckle2013: 36). For that purpose, we make use of data from Holger Döring and Philip Manow (Reference Döring and Manow2019) (see Test 1.8 Table A2 and Test 2.9 Table A3). For the model studying the differential effects of reshuffles of ministers belonging to the PM's party compared to the coalition partners, we also interact the ideological range with the effect of reshuffles of coalition partner ministers to see whether the performance of PMs presiding over cabinets that span a broader ideological spectrum reacts more sensitively to reshuffles of ministers belonging to the coalition partners (see Test 2.10 in Table A3). Last, an additional test includes a control variable that captures whether a PM was party leader while serving as head of government. Party leadership is a key resource for successful performance in office (Grotz et al. Reference Grotz2021), but also impacts the PM's ability to reassign ministers, in particular those belonging to her own party (see Test 1.9 Table A2 and 2.11 in Table A3).
We also tested for a potential bias emerging through missing country-level factors. For that purpose, we introduce a control variable that identifies the appointment and dismissal powers of the state president (based on data from Bairett Reference Bairett2015) (see Test 1.10 Table A4 and 2.12 in Table A5). The variable is created by summing up the scores for presidential powers in cabinet formation (ranging from 0, where the head of state cannot appoint the cabinet, to 4, where the head of state appoints the cabinet without need for assembly confirmation) and cabinet dismissal (ranging from 0, where ministers may be removed only by the assembly, to 4, where the head of state is unrestrained in dismissing ministers). Similar to the index of prime-ministerial powers, it ranges between 0 and 8, where 8 indicates unlimited powers of the state president in cabinet formation. Possibly, the veto power executed by the president might capture the limitations of PMs' powers more accurately than their own rights (Bucur Reference Bucur2017; Fischer et al. Reference Fischer, Dowding and Dumont2012). Additionally, we checked whether it makes a difference to our results if we consider whether parliament holds the power to dismiss individual ministers, using data from Florian Grotz and Ferdinand Müller-Rommel (Reference Grotz and Müller-Rommel2011) (see Test 1.11 Table A4 and 2.13 in Table A5). Moreover, we added three variables that capture the economic context the year before the cabinet came into power – a key factor determining the ability of PMs to perform successfully (Brändström Reference Brändström2015; Hansen et al. Reference Hansen2013): GDP change (based on the Comparative Political Dataset), unemployment rate (based on the International Labour Organization database) and inflation rate (based on the World Bank) (see Test 1.12–1.14 Table A4 and 2.14–2.16 in Table A5).
In a last set of robustness tests, we introduce some modifications to the sample to see whether the effects are driven by specific cases. To begin with, we include two cabinets in which both the PMs and all ministers had no party membership (Berov in Bulgaria and Ciolos in Romania). The introduction of the single-party cabinet control variable causes their exclusion, but omitting this variable from the analysis and including these two cases does not modify the results (see Test 1.15 Table A4 and 2.17 in Table A5). Moreover, we exclude cabinets in which PMs were selected by certain government parties without being formal members of those parties (Oresharski in Bulgaria; Orešković in Croatia; Bajnai and Medgyessy in Hungary; Skele I in Latvia; and Ciolos, Isarescu, Vacaroiu I and II in Romania).Footnote 14 In these cases, the relationship between PMs and party-affiliated ministers might follow different logics (see Test 1.16 Table A4 and 2.18 in Table A5).
Tables A2 to A5 in the Online Appendix present the results for all robustness checks. The strengths of the key coefficients and their level of statistical significance do not show any substantial changes compared to the models presented in the main text.
As a final step in our analysis, we look for support for the causal mechanisms we proposed as link between the reassignment of ministers belonging to the PM or the coalition partners' parties and prime-ministerial performance. Earlier in this article, we argued that reshuffles of ministers from other coalition parties undermine the PM's ability to perform successfully in office, because the PM has little influence over who is selected for these positions. The consequence is that conflict within the cabinet becomes more frequent and the PMs' ability to shape government policies as well as to maintain the support of the parliamentary majority might diminish. To underpin the plausibility of this causal mechanism, we estimated models predicting the success of PMs on the six subdimensions that together form the index of prime-ministerial performance. We follow the same modelling strategy as outlined for the main models. The six models are presented in Online Appendix Table A6. Figure 5 visualizes these effects by displaying the predicted value for the respective subdimensions of prime-ministerial performance for cases with zero compared to 1.8 reshuffles per 200 days.
Figure 5. Linear Prediction of Reshuffles of PM Party Ministers (0/1.8) and Coalition Partner Ministers (0/1.8) on Subdimensions of Prime-Ministerial Performance
Notes: The circle indicates zero reshuffles in 200 days, the triangle indicates 1.8 reshuffles per 200 days. Models are presented in Table A3 in the Online Appendix. All models include the control variables presented in Table 2, standard errors are clustered at the cabinet level and country- and period-fixed effects.
Regarding reshuffles of ministers belonging to the PM's party, the linear prediction for the subdimensions of prime-ministerial performance does not differ substantially if not a single or 1.8 ministers were replaced per 200 days. This finding is consistent with the null-effect of this type of reshuffle on overall performance. By contrast, reassignments of ministers belonging to the coalition partners impact multiple subdimensions. To begin with, the PM's ability to settle cabinet conflicts – that is, conflicts between individual ministers or between the PM and ministers – is lower if larger numbers of coalition partner ministers are reshuffled. Additionally, the PM's ability to shape government policies by enacting her preferred policies for major societal concerns and by averting non-preferred policies decreases considerably. Moreover, an increase in the number of reshuffles of coalition partner ministers leads to a significant change in the PM's ability to maintain support of the parliamentary majority. With a difference of about 0.60 points, this gap is substantial and statistically significantly different from zero.
By contrast, we observe no comparable effect of reshuffling of coalition partner ministers on the PM's ability to maintain support of her own party; performance on this subdimension decreases by 0.34 points if 1.8 compared to no reshuffles occur, but the effect does not reach conventional levels of statistical significance. Given that reshuffles of coalition party ministers only have indirect consequences for the PM's party, this finding is logically consistent with the causal mechanism we propose. Lastly, reshuffles in core portfolios do not have an effect on the PM's ability to manage crisis and perform at the international level that reaches conventional levels of statistical significance. In these two subdimensions of prime-ministerial performance, PMs fulfil their tasks more independently, with less dependence on the expertise of the ministers and the parties involved in the government. The absence of an empirical link between reshuffles of ministers belonging to coalition partners on the one hand and the PM's success in managing crises and protecting national interests abroad on the other hand is therefore again consistent with the causal mechanism proposed above. Overall and at large, these patterns support our theoretical proposition that the link between the reassignment of ministers belonging to coalition partners and prime-ministerial performance works through increasing levels of conflict between PMs and coalition partner ministers.Footnote 15
Conclusion
In parliamentary democracies, government ministers are in charge of directing their departments in accordance with the will of the PM, who runs the state affairs. Ministerial turnover may thus affect the operation of parliamentary governments in general and the performance of PMs in particular. While many studies have explored when and why ministers are reshuffled, the consequences of such reshuffles have received surprisingly little attention. Therefore, this article has investigated the effects of cabinet reshuffles on prime-ministerial performance in the post-communist democracies of CEE, which have been characterized by a relatively high degree of political dynamics and government instability. In this region, frequent ministerial turnover has a negative impact on PMs to fulfil their office tasks successfully. This effect is driven by the turnover of ministers belonging to other coalition parties, which reduces prime-ministerial performance. These results hold independently of whether a ministerial turnover was actively enforced by the PM or happened outside her influence. Therefore, cabinet reshuffles in CEE cannot be regarded as last-ditch moves of PMs to demonstrate their leadership and thus be indicative of their weak performance but rather seem to have a distinct, detrimental effect on the fulfilment of prime-ministerial tasks.
Overall, our findings imply that the partisan dimension of cabinet reshuffles is decisive for prime-ministerial performance in post-communist democracies. This insight corresponds to other recent research, which showed that cabinet decision-making in most CEE countries follows the ‘coalition compromise model’ in which legislation in each policy domain requires the consent of all government parties (Bergman et al. Reference Bergman, Ilonszki, Müller, Bergman, Ilonszki and Müller2019: 555–558). Our analysis suggests that prime-ministerial performance suffers from the turnover of ministers from other coalition parties because they are in charge of implementing the joint government policies in their domains but are hardly under the PM's political control.
Against this background, the impact of ministerial reshuffles on prime-ministerial performance could be further explored in various ways. To begin with, our expert-based measurement of different functional dimensions of prime-ministerial performance might be complemented by public opinion surveys on prime-ministerial approval that are carried out on a regular basis in the individual CEE countries. These quantitative approaches could be followed by process-tracing studies that seek to uncover the precise sequence of events following a cabinet reshuffle.
One might also look into the interaction between reshuffles and other factors to see if their influence on prime-ministerial performance is more or less pronounced under certain conditions. For instance, the power relation within the coalition – that is, whether one party is clearly larger than the others – might shape the effect of turnover of coalition partner ministers. Likewise, the economic situation might moderate the relationship between prime-ministerial performance and reshuffles in core portfolios, as the budgets of ministers in these portfolios might increase, making it harder to replace them.
Moreover, it remains open whether our results are only confined to post-communist countries or are of a more general nature. To answer this question one could explore the effects of cabinet reshuffles on prime-ministerial performance in other contexts. Relevant cases include Westminster democracies such as the UK where not only is the overall number of ministerial changes relatively high, but active reshuffles are particularly frequent (Thompson Reference Thompson2020). Finally, future studies could expand the scope of our article by investigating the influence of ministerial turnover on government performance in different policy domains (Alexiadou Reference Alexiadou2015). In this regard, special attention may be paid to the career attributes of ministers that shape their office stability (Morgan-Collins and Meyer-Sahling Reference Morgan-Collins and Meyer-Sahling2021) but could also impact their policy influence. This aspect seems to be particularly relevant for CEE democracies where a relatively high number of technocrats and political outsiders have been recruited into cabinet (Semenova Reference Semenova and Costa Pinto2018, Reference Semenova2020). Exploring to what extent the policy performance of individual ministers matches the performance of the respective PMs might contribute considerably to a more comprehensive and differentiated picture of how political executives work both in CEE and in other world regions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/gov.2022.24.
Acknowledgements
We are grateful to the anonymous reviewers, Ludger Helms, Michelangelo Vercesi and the participants of the panel ‘Prime Ministerial Performance in the Age of “Presidentialization”’ at the 2021 IPSA World Congress. We would also like to thank Mažvydas Jastramskis, Darina Malová and Maria Spirova for sharing their expertise on particular cases of cabinet reshuffles. This joint work is a follow-up to the project ‘Career Patterns and Political Performance of Prime Ministers in Central and Eastern European Democracies’, led by Florian Grotz and Ferdinand Müller-Rommel and funded by the German Research Foundation (DFG; Grants GR 3311/3-1 and MU 618/18-1).
Open access
