Variation in institutional rules, cabinet termination and ministerial termination: a descriptive analysis

Table 1 summarizes the instruments by which PMs and presidents may terminate cabinets and ministerial careers. Our sample includes countries with both indirectly (i.e., parliamentary systems) and popularly elected (i.e., semi-presidential systems) presidents, some changing this mode during the period of study. For example, popular presidential elections were introduced in the Czech Republic and Slovakia, but removed in Moldova. In Slovakia, a high degree of party polarization in the late 1990s made it impossible to elect the president from the parliament and the country remained without a president for over a year (Spáč, Reference Spáč and Hloušek2013: 125). After the 1998 elections, the government coalition introduced popular presidential elections via constitutional amendments. In Moldova, in 1999, President Petru Lucinschi initiated a consultative referendum to grant the president the power to dismiss government and tried to force the parliament to either hold a national referendum on the issue or grant him cabinet-dismissal power through constitutional amendments. This backfired, with parliamentary support for the president declining dramatically, and in mid-2000, the parliament removed popular election of the president and decreased his powers (Roper, Reference Roper2008: 120).

Table 1. Institutional rules, cabinet stability, and ministerial terminations in Central and Eastern Europe

^a The values were computed by the authors from Doyle and Elgie’s prespow1 scores (Reference Doyle and Elgie2016).

^b Ministers appointed to caretaker cabinets were excluded.

Authors’ own classification and calculation.

All the national constitutions in the new democracies specify a constrained dissolution of parliament, which can be initiated under specific circumstances (cabinet-building failures or gridlock) and/or requires presidential consent. The high threshold for initiating parliamentary dissolution makes it rare. As of 2018, several countries, both semi-presidential (Croatia, Lithuania, Romania, and Russia) and parliamentary (Hungary) have never held early elections caused by dissolution of parliament.

Where parliamentary dissolutions occur, it is usually due to failure to form a government or because the government has lost a vote of no confidence – as in Slovenia in 2011. The collapse of a governing coalition has also triggered early elections – for example, in Poland in 2007, when corruption allegations against the leader of a junior partner in the coalition caused its downfall. Intra-party conflicts, too, can be responsible for parliamentary dissolutions – the early Slovenian elections of 2014 took place after the PM, Alenka Bratušek, lost her party leadership position to her predecessor, causing turmoil within the coalition and, eventually, the PM’s resignation and an early election. In general, the constitutions in new democracies pose many institutional obstacles to parliamentary dissolution by requiring agreements between various political actors.

Some constitutions grant their presidents discretionary power to dismiss cabinets. In the late 1990s, President Yeltsin of Russia actively used this power to break down the cabinets of PMs Chernomyrdin, Primakov, and Stepashin (Daniels, Reference Daniels1999; Semenova, Reference Semenova, Dowding and Dumont2015). In Ukraine, from independence to 2006 and again from 2010 to 2014, presidents could dismiss governments. During his first presidential term (1994–1999), President Kuchma dismissed PMs Marchuk and Lazarenko (Protsyk, Reference Protsyk2003: 1082). The constitutional amendments of 2004, operating from 2005/2006 to 2010 and again from 2014, deprived the president of this power. The 1990 Croatian constitution granted presidential powers of dismissal, but they had never been used by the time the 2000 constitutional amendments removed them.

The Romanian constitution of 1991 allowed the president to dismiss the PM if the PM is ‘unable to govern’. In 1999, President Constantinescu used this vague rule to dismiss PM Vasile after 10 out of 17 cabinet ministers had resigned. Vasile sat in cabinet without ministers for several days before eventually resigning (Gherghina, Reference Gherghina and Hloušek2013: 265). The 2003 amendments to the constitution explicitly denied the president discretion to dismiss the PM, thus confirming the validity of the previous used power of dismissal.

Finally, some constitutions grant the president the power to appoint ministers. The Ukrainian constitution allows the president to appoint ministers in foreign affairs and defence. In Russia, presidential power to appoint ministers is granted by the federal law on government. From 1992 to 1997, the president could appoint ministers of foreign affairs, the interior, defence, and national security. Since 1997, he can also appoint ministers of justice and emergency situations (Semenova, Reference Semenova, Dowding and Dumont2015). In Poland, until the constitutional amendment of 1997, the popularly elected president shared control with the PM over the nomination of ministers of defence, foreign affairs, and the interior.

Table 1 also reports the average duration of cabinets and ministerial careers. Two important patterns emerge from the descriptive analysis. First, the average cabinet duration is shorter than the average ministerial duration in post-communist countries. Second, the average cabinet duration is slightly longer for popularly elected than for indirectly elected presidents. We now turn to the challenge of explaining how specific institutional rules described in constitutions and legal texts affect cabinet durability and ministerial terminations.

Methodology and the operationalization of variables

We are interested in the effect of institutional rules on both cabinet and ministerial durability, so we perform two analyses. In the analysis of cabinet durability, our dependent variable is the Survival Time of a Government (in months).Footnote ³ Because we are interested in exploring the determinants that affect the risk of early elections and non-electoral replacements separately, we used a competing risks approach (Diermeier and Stevenson, Reference Diermeier and Stevenson1999). We first coded the type of failure (early dissolution and non-electoral replacement) each cabinet faced. Cabinet dissolution is defined as a situation in which the PM and/or the president chooses to call early elections; it distinguishes early elections from constitutionally necessary early elections (e.g., after major constitutional amendments) and from regular elections, which occur when a parliamentary term expires. In some semi-presidential systems, cabinets are replaced at the end of the presidential term. Non-electoral replacements of cabinets define cabinet terminations because of a change of governing coalition or PM which is not the result of elections. Second, we calculate two models, one for each type of risk, in which cabinet failure because of the other risk is censored (Box-Steffensmeier and Jones, Reference Box-Steffensmeier and Jones2004: 168–169). For studying ministerial durability, our dependent variable is the Status and the Survival Time of a Minister (in months). We measure ministerial tenure for each minister in all cabinets and all portfolios to which this minister was appointed.

In both analyses, our major explanatory variables are the termination and appointment powers of the PM and president. Specifically, the effects of the following institutional powers on ministerial and cabinet durability will be assessed: (1) the mode of presidential elections, which defines the type of a political system (a semi-presidential and parliamentary one); (2) the president’s power to dismiss cabinets; (3) the president’s power to appoint ministers, and (4) the president’s combined power to dismiss cabinets and appoint ministers. All these variables are coded from each country’s constitutional and legal texts (1 if the institutional rule is present and 0 otherwise), in effect since 1990 (or the year of the country’s first democratic election).

Finally, we use a set of control variables in each analysis. The control variables for cabinet durability include cabinet- and parliament-specific factors (Strøm and Swindle, Reference Strøm and Swindle2002; Saalfeld, Reference Saalfeld, Müller, Strøm and Bergman2008; Somer-Topcu and Williams, Reference Somer-Topcu and Williams2008; Schleiter and Morgan-Jones, Reference Schleiter and Morgan-Jones2009; Conrad and Golder, Reference Conrad and Golder2010; Nikolenyi, Reference Nikolenyi2014). The features of a specific cabinet strongly influence the value that it has for political actors in the political system (Warwick, Reference Warwick1994). Specifically, the variable Coalition (1 if a cabinet is a coalition) is included; we expect that the existence of a coalition will restrict opportunistic dissolutions and so coalition cabinets should be more durable than single-party ones (Strøm and Swindle, Reference Strøm and Swindle2002: 582). In contrast, Minority cabinets (1 if a cabinet is a minority cabinet) are expected to be less durable because of their limited ability ‘to generate effective public policy’ (Strøm and Swindle, Reference Strøm and Swindle2002: 548), as too are Caretaker cabinets (Warwick, Reference Warwick1994; 1 if a cabinet is a caretaker cabinet).

To account for the high occurrence of technocratic cabinets in post-communist countries (Protsyk, Reference Protsyk2005), we include the variable PM’s party in government (1 if the cabinet includes the PM’s party). We expect the cabinet presence of the PM’s party to increase the policy and office value for PMs, thus decreasing the risk of cabinet failure. The variable Maximum Duration of the Cabinet (in years) measures the period until the constitutional election term at the time of cabinet formation. The benefits for the chief executive in continuing the current government decline as a regular election approaches (Balke, Reference Balke1990: 202–203; Lupia and Strøm, Reference Lupia and Strøm1995: 656; Schleiter and Morgan-Jones, Reference Schleiter and Morgan-Jones2009: 507). Cabinet fractionalization measures the number of coalition partners, to control for the complexity of bargaining among coalition partners over cabinet terminations (Warwick, Reference Warwick1994; Druckman, Reference Druckman1996). We expect greater cabinet fractionalization to decrease cabinet durability. Similarly, Parliamentary fractionalization is expected to decrease cabinet durability (King et al., Reference King, Alt, Burns and Laver1990: 858, 863; Laver and Schofield, Reference Laver and Schofield1990: 161). The fragmentation of parliaments is measured by the effective number of parties in parliament (Laakso and Taagepera, Reference Laakso and Taagepera1979). Finally, in order to control for the effect of democratization on cabinet stability, we include the variable V-Dem Electoral Democracy Index (Teorell et al., Reference Teorell, Coppedge, Skaaning and Lindberg2016), which measures the ‘institutional guarantees’ laid out in Dahl’s (Reference Dahl1973) concept of ‘polyarchy’. The data refer to the level of electoral democracy at the time of cabinet formation. We expect higher cabinet durability in countries with a higher level of electoral democracy.

Our analysis of ministerial durability uses as control variables cabinet- and parliamentary-specific determinants which have been shown to affect the PM’s ability to deal with adverse selection and moral hazard delegation problems. Specifically, coalitions decrease the ability of PMs to dismiss adversely selected ministers, because coalition agreements often prevent personnel changes without renegotiation of the coalition agreement (Indridason and Kam, Reference Indridason and Kam2008). On the other hand, ministers from the PM’s party have a higher risk of dismissal, since losing them does not endanger the survival of government (Huber and Martinez-Gallardo, Reference Huber and Martinez-Gallardo2008: 172–173). Thus, we include Coalition and PM’s Party (1 for serving in a coalition government and for belonging to the PM’s party).

We also add two variables describing the type of the cabinet: Majority (1 in a majority government) and Minority (1 in a minority government). The reference category for these variables will be Non-parliamentary cabinets (i.e., those whose formation is not dependent on parliamentary elections) that have been formed in Moldova, Russia, and Ukraine. In order to decrease the likelihood of delegation problems, the PMs (and political parties) must screen potential candidates for cabinet positions (Berlinski et al., Reference Berlinski, Dewan and Dowding2007, Reference Berlinski, Dewan and Dowding2012; Bucur, Reference Bucur2017). The probability of adverse selection should be lower in majority than in minority cabinets, because the former have a larger pool of candidates to select from (see Huber and Martinez-Gallardo, Reference Huber and Martinez-Gallardo2008: 177). We therefore expect that ministerial terminations will occur more often in minority than in majority cabinets.

We also include a set of individual-level control variables in our analysis of ministerial durability. Two variables indicate whether a minister had held Foreign Affairs or Defence portfolios (1 if appointed to either). We selected these two portfolios because, in virtually all constitutions of post-communist countries, they fall within the presidential areas of influence because presidents are the commanders-in-chief of military forces. They also appoint ambassadors and formulate foreign policy goals. We therefore expect ministers in these portfolios to have lower termination rates because of their strong connections to the president.

Salient portfolios afford incumbents longer tenures than less salient portfolios (Berlinski et al., Reference Berlinski, Dewan and Dowding2007, Reference Berlinski, Dewan and Dowding2012; Quiroz Flores, Reference Quiroz Flores2017), because they are highly prestigious and desirable, so potential candidates are carefully screened, which decreases adverse selection (Huber and Martinez-Gallardo, Reference Huber and Martinez-Gallardo2008: 172). To capture this, we include the variable Portfolio of finance (1 if appointed to this portfolio), which is ranked as one of the most prestigious portfolios in both Western and Eastern Europe (Druckman and Warwick, Reference Druckman and Warwick2005: 30; Druckman and Roberts, Reference Druckman and Roberts2008: 114).

The variable Parliamentary/party experience measures the effect of the minister’s political experience on termination risk.Footnote ⁴ We expect the choice of ministers from parliaments and party offices decreases the risk of adverse selection; they should therefore have lower chances of being dismissed than ministers recruited from outside politics (Blondel and Thiebault, Reference Blondel and Thiebault1991; Semenova, Reference Semenova, Costa Pinto, Cotta and Tavares Almeida2018). Finally, we include Age of the minister at the time of appointment and Gender (1 if a minister is a woman).Footnote ⁵

A number of methodological considerations guide our model selection. Because we want to examine the effects of institutional rules and a set of control variables on cabinet and ministerial survival, we use survival analysis. As we have no specific assumptions about the underlying distribution of hazard function, we choose a semiparametric Cox proportional hazard model (Cox, Reference Cox1972). Mathematically, the survival model estimates the baseline hazard function h_o(t) as well as the estimates of β and its variance–covariance matrix: h(t) = h₀(t) exp(β ₁χ₁ + … + β _kχ_k). The outcomes of a survival model are measured as the hazard ratios: that is, the probability that an individual minister will exit a cabinet at time t provided the minister survives until this point (for the case of cabinets, that an individual cabinet will be terminated provided it survives until the time t).

We assume there are unobserved country-specific risks endangering cabinets and ministerial careers, so we estimate Cox proportional hazard models with country-level shared frailty (random effects) in order to provide some control. We assume that all ministers in a country share the same frailty, but frailty might differ between countries. These models are similar to multi-level models with lower-level units nested into higher-level ones (countries). The frailty term in these models is assumed to have a gamma distribution, with mean 1 and variance θ (theta). In each model, we control for these random effects (theta). In each case, theta was highly significant at P < 0.001.

Institutional rules and the analysis of cabinet durability in post-communist countries

In Table 2, we report coefficients from the Cox proportional hazard model. Coefficients above 1 mean that the determinant increases; below 1 the variable decreases the hazards of cabinet termination. By convention, cabinets that are terminated because elections are censored (Fischer et al., Reference Fischer, Dowding and Dumont2012). Since only 22 out of 195 post-communist cabinet terminations result from early elections (see also Schleiter and Morgan-Jones, Reference Schleiter and Morgan-Jones2009: 498) we could not estimate a separate model for these cases of cabinet terminations. We therefore estimate the survival of governments formed without there being new elections.

Table 2. Cabinet durability between elections (non-electoral cabinet replacements)

* P < 0.10, **P <0 .05, ***P <0 .01.

Based on robust standard errors clustered by country.

Authors’ own calculations.

We see from Table 2 that having popularly elected presidents makes no difference to cabinet durability, in line with our Hypothesis 1a, though contradicting a widely accepted assumption about semi-presidentialism (Model 1c). The ability of the president to dissolve the cabinet at their discretion (by dismissing the PM) between elections also has no effect on cabinet durability, contradicting our Hypothesis 2a drawn from semi-presidentialism research (Model 1c). The power of the president to appoint ministers decreases cabinet durability when we include an interaction between this presidential power and the president’s power to dismiss cabinets and do not control for various other factors, such as coalition governments and how democratic the system is (Model 1b). This suggests that our Hypothesis 3a is overwhelmed by other considerations, such as partisan considerations of coalition building, and other institutional effects such as the maximum duration between mandated elections. The individual effect of the president’s power to appoint ministers remains negative but it is marginally significant (Model 1c). Our Hypothesis 4a is confirmed. The discretionary power of presidents both to dissolve cabinets by dismissing the PM and to appoint ministers increases cabinet durability (Models 1a and 1b; Figure 1).

The effects of control variables on cabinet durability largely support expectations drawn from previous studies (Saalfeld, Reference Saalfeld, Müller, Strøm and Bergman2008; Schleiter and Morgan-Jones, Reference Schleiter and Morgan-Jones2009). Specifically, the risk of non-electoral replacements is significantly higher for caretaker cabinets and if parliamentary fractionalization is high. Other control variables are not statistically significant but generally show the expected tendency of the effects. For example, the maximum duration of the cabinet term does not affect cabinet durability between elections, and minority governments have higher termination hazards.

Institutional rules and ministerial durability

We now turn to the more difficult issue of measuring ministerial durability in relation to cabinet durability. The entry and exit dates for each minister in the 14 countries have been coded,Footnote ⁶ and each minister assigned to one of three categories: ministers (1) reappointed to subsequent cabinet when their initial cabinet fails (upper bar, ‘Cabinet terminates, minister not’); (2) exit during their first cabinet (middle bar, ‘Minister terminates, cabinet not’); and (3) terminated when the initial cabinet fails (lower bar, ‘Minister and cabinet terminate’). Ministers who had not left the cabinet by the end of the observation period were censured, since their duration is unknown.

We see that at least half of all ministers in six countries failed in their first (and only) cabinet (Figure 2, lower bar). This pattern is particularly strong in Bulgaria, Poland, and Slovakia. In contrast, in former Soviet countries (e.g., Lithuania and Moldova), ministers are often reappointed to various cabinets, sometimes under various PMs (Figure 2, upper bar). The duration of ministerial careers in Croatia, the Czech Republic, Latvia, and Estonia is more variable, as shown by the similar numbers of ministers exiting with their first cabinet and those reappointed. The careers of approximately 15% of ministers terminate between government formation and its end (Figure 2, middle bar). The inter-country variation in the number of ministers failing during one cabinet ranges from 5% (Latvia and Ukraine) to approximately 32% (Hungary). These aggregate findings suggest that, in half the countries, when the government fails, its ministers often remain in office and continue to serve under the new PM.

Figure 1. Non-electoral cabinet replacements. Adjusted predictions of the president’s power to discretionally dismiss cabinets and to appoint ministers (with 95% CIs). Based on Model 1b (control variables are set at means). Author’s own calculations.

Figure 2. Ministerial terminations in Central and Eastern Europe (as a percentage of all appointed ministers). Categories are mutually exclusive; ministers appointed to caretaker cabinets and ministers who were still in cabinets by the end of the observation period were excluded from the analysis. Authors’ classification and calculation.

The results of the shared frailty Cox survival model of ministerial durability are displayed as hazard ratios (Table 3).Footnote ⁷ A hazard ratio above 1 means an increase in the risk of failure; a hazard ratio below 1 means a decrease.

Table 3. Determinants of ministerial terminations in Central and Eastern European countries

* P < 0.05, **P < 0.01, ***P < 0.001, †P < 0.10.

^a The reference category consists of non-parliamentary cabinets and minority cabinets.

^b The reference category consists of non-parliamentary cabinets and majority cabinets.

Hazard ratios with standard errors (in parentheses) are reported. Standard errors of hazard ratios are conditional on theta

Authors’ own calculations.

Our Hypothesis 1b is confirmed, as there is no difference between popularly elected and indirectly elected presidents with regard to ministerial durability (Model 2c). Being popularly elected seems to bring no extra authority beyond the de jure powers of presidents. The president’s discretionary power to dismiss cabinets (by dismissing the PM) increases ministerial durability, confirming our Hypothesis 2b (Model 2c). The president’s discretionary power to appoint ministers also increases ministerial durability, confirming our Hypothesis 3b (Model 2c). The sign for ministerial durability where presidents have the combined power to dissolve cabinets (by dismissing the prime minister) and appoint ministers is also in the direction of increasing ministerial durability (Models 2a and 2b), as in our Hypothesis 4b; however, it does not reach normal considerations of statistical significance (Figure 3).

Figure 3. Ministerial durability. Adjusted predictions of the president’s power to discretionally dismiss cabinets and to appoint ministers (with 95% CIs). Based on Model 2b (control variables are set at means). Author’s own calculations.

The effects of our control variables largely confirm previous findings. For example, ministers in coalitions tend to survive longer than their counterparts in other types of cabinets, while high parliamentary fractionalization tends to reduce ministerial durability. Neither majority nor minority cabinets have any significant effect on ministerial durability. Against expectations, the risk of being terminated is not higher for ministers from the PM’s party than for ministers from other parties. This effect may stem from a massive recruitment of politically inexperienced ministers in post-communist cabinets, who suffer higher risks of being dismissed than their partisan counterparts (e.g., Semenova, Reference Semenova, Costa Pinto, Cotta and Tavares Almeida2018: 190–192).

At the individual level, the minister of foreign affairs is more durable than other ministers, but holding the defence or finance portfolio does not increase ministerial durability. As in Western European parliamentary countries, gaining political experience (as an MP and/or a party leader) increases the chances of staying in government. Finally, there are no effects of the major socio-demographic control variables (age and gender) on ministerial termination in post-communist countries.