Political leverage of the presidents

In presidential systems, cabinets depend on the president's assessment of what makes government rule most viable. Under this institutional design, the president can unilaterally decide the composition of the team and the survival of each partner. Of course, the support of a legislative majority is not a trivial factor, particularly since party systems have become increasingly fragmented – and have even collapsed in some instances (Chaisty et al. Reference Chaisty, Cheeseman and Power2018). At the same time, the self-recruitment of presidential candidates has also increased (Carreras Reference Carreras2012; Inácio and Llanos Reference Inácio, Llanos, Andeweg, Elgie, Helms, Kaarbo and Müller-Rommel2020). Minority presidents secure interparty support in fragmented legislatures through coalition governments – a recurrent phenomenon in Brazil, Chile, Colombia and Uruguay. However, ‘ghost coalitions’ involving legislative majorities without the formation of multiparty cabinets are also observed in Ecuador (Mejía Guinand and Botero Reference Mejía Guinand, Botero, Camerlo and Martínez-Gallardo2018) and Peru (Vera and Carreras Reference Vera, Carreras, Camerlo and Martínez-Gallardo2018). Majority presidents usually lead one-party cabinets, particularly in countries with less party fragmentation, such as Argentina and Mexico.

In addition, rules for the presidential election differ and were modified during the third wave of democratization in most Latin American countries, with only Honduras, Mexico, Panama, Paraguay and Venezuela using the plurality rule since 1978. The adoption of run-offs altered the conditions for presidential competition and impacted cabinet composition as well (Freudenreich Reference Freudenreich2016; Pérez-Liñán Reference Pérez-Liñán2006). In fact, run-offs drive the formation of electoral alliances before the second round, which may later become the basis for coalition governments. However, they also increase party-system fragmentation, may encourage outsiders to participate, give (small and new) parties a chance to test their strength and invite citizens to vote sincerely in the first round for the candidate they like most. All this increases the chances of electing a president who was not the first choice for voters, but probably the least disliked of the first-round winners. Such a president will face strong rivals.

Certainly, the president's control of the cabinet and the risks of moral hazard depend on the president's partisan status in Congress and their own electoral performance in the presidential race. However, these do not define in isolation the bargaining conditions between the branches of government and the command of the executive branch. According to Daniel Ponder (Reference Ponder2012), a hostile political environment – whereby the president faces popular disapproval and there is a lack of trust in government – reduces presidential leverage and may spur a cabinet reshuffle, as such an environment can put extra pressure on the performance of individual ministers and the cabinet as a whole. In 2002, for example, when Peruvian president Alejandro Toledo saw his popularity rating shrink to 18% and as his government faced a credibility crisis, he carried out a profound reshuffling of his cabinet – firing the experts in charge of core ministries (Vera and Carreras Reference Vera, Carreras, Camerlo and Martínez-Gallardo2018).

In view of all this, we built an indicator of the president's political leverage as likely impacting cabinet formation and turnover. This variable is based on three political factors considered by the scholarship to be critical drivers of cabinet change. First, the total share of votes the president won in the presidential race – that is, in the first electoral round; second, the level of presidential support in the legislative branch, which may come from a single party or a from a coalition; and, third, the popularity of the president per se. We expect that presidents scoring strongly on these three aspects are more likely to run the cabinet smoothly, to overcome internal conflicts and thus to feel less pressure to implement a reshuffle. Hence, if we observe their relations with ministers, we expect that:

Reputational assets of presidents

While political leverage refers to the political conditions that accompany presidents in power, there are other conditions – those concerning one's personal profile – also potentially defining the chances presidents have in case of challenges to their rule. Existing studies focus on the interaction of legal and party variables to explain different outcomes in terms of presidential profiles, which particularly vary in the extent to which presidents are party-loyal and depending on what kind of previous political experience they have (Inácio and Llanos Reference Inácio, Llanos, Andeweg, Elgie, Helms, Kaarbo and Müller-Rommel2020). For instance, David Samuels and Matthew Shugart (Reference Samuels and Shugart2010: 67) define ‘insiderness’ as the nature and extent of a prospective executive's links to a central party organization, measuring it as a combination of partisan and political experience. In general, the lack of political experience at the time of selection to office is regarded as problematic, as newcomers are difficult to work with, may display inconsistent or unpredictable behaviour or take poor decisions (Baturo and Elkink Reference Baturo and Elkink2021).

We assume that presidents' prior experience regarding the diversity and timing of their political activities in partisan, elective or appointed offices generates important reputational gains for their career and political influence. Hence, we consider reputation to be an asset that allows the president to nurture and respond to the expectations of voters and supporters, differentiates them from opponents and creates barriers of protection against competitors' threats. Ludger Helms (Reference Helms2016) considers a leader's reputation a critical aspect of leadership capital, defining their opportunity structure in politics.

Presidents may come into office with different levels of reputational assets. We propose that presidents who do not count on such assets are more exposed to intra-cabinet conflicts, which may be related to party, policy and/or personal divergences between the president and their ministers. Simply put, outsiders, newcomers or presidents with little previous political experience will tend to compensate for this reputational weakness by appropriating the reputation and expertise of their ministers. However, these ministers are at the same time the president's competitors, so sooner or later tensions will mark their relations. Presidents who have built robust reputational assets before their path to the presidency, however, are less dependent on the performance of individual ministers and will engage in fewer controversies with them, so that:

Operationalizing reputational assets is challenging, but recent works have proposed measures allowing for comparison across countries and regimes. Miguel Carreras (Reference Carreras2017) uses biographical data on political positions occupied by democratic leaders before coming to power in the period from 1945 to 2015. Alexander Baturo and Johan Elkink's (Reference Baturo and Elkink2021) measure accounts for the length, breadth and quality of leaders' involvement in politics. The authors use a latent variable model to enumerate the indicators and estimate PolEx, their measure of ex ante political experience, which is applied across regimes. We rely on the latter measure for the empirical analysis of Latin American presidents' reputational assets.

Political leverage and presidents' reputational assets are our core variables to capture the effects of political and reputational factors involved in the presidential calculus on cabinet changes. They seek to grasp how the presidents' strength or weakness drives them towards more costly strategies, such as changing their governing team. However, we argue that the effects of these variables are relational: that is, they are conditional on the potential threat the ministers pose to the president. We incorporate into our model this potential threat via ministers' ‘walk-away value’ – or their chances of political survival outside the government as well as personal factors prospectively driving greater autonomy from their side, thus leading eventually to ministerial opportunism within the cabinet.

Walk-away value of ministers

Ministers' attributes, expertise and political or professional networks matter: if such factors are desired by the president, they increase ministers' bargaining power; if the latter are dismissed from the cabinet, these factors serve to reduce the losses they experience. This defines the minister's ‘walk-away value’Footnote ¹ – a minister with a high ‘walk-away value’ is one more likely to act in a discretionary way and pursue their own policy positions, even if they diverge from those of the president. Ministers with high ‘walk-away values’ can be seen as a threat to a president with few reputational assets.

Studies on cabinet conflicts have increasingly focused on the incentives for ministerial discretion (Laver and Shepsle Reference Laver and Shepsle1990) and the risks of ministerial opportunism (Indridason and Kam Reference Indridason and Kam2008), by which ministers can defy the president even in the face of the threat of being fired. For ministers, being in office is playing a game with mixed motivations for participation, one in which collective responsibility and the success of the government as a whole are balanced against ministers' opportunities to leverage their own electoral futures or those of their respective parties. Opportunities to leverage one's career or the party's reputation are more attractive in contexts where the ministerial position offers greater resources and increased chances to influence policymaking. In Latin American presidentialism, the ministers who are ultimately appointed reflect not only presidential choices but also the strategies of parties, politicians and interest groups alike.

For the heads of the executive, the risks of agency losses are connected with the ministers' partisan ties or expertise. On the one hand, ministers belonging to other parties – or even to a different faction of the presidential party – are more likely to cause ministerial drift in response to the demands of their clientele (Martínez-Gallardo and Schleiter Reference Martínez-Gallardo and Schleiter2015) or in an attempt to differentiate themselves from other parties in the cabinet (Fortunato Reference Fortunato2019; Martin and Vanberg Reference Martin and Vanberg2011). On the other, non-partisan ministers hold a diverse range of profiles, such as technocrats, ‘technopols’ or businesspeople, with previous trajectories and political (Domínguez Reference Domínguez1997; Inácio Reference Inácio2013) or corporate (Alexiadou Reference Alexiadou2015; González-Bustamante Reference González-Bustamante2020) career plans that may encourage them to seek greater independence as ministers. Presidents weigh up the costs of removing a minister connected to the political networks that back them, especially in contexts where parties are not the most effective channel for doing so – as is the case, indeed, in many Latin American countries. We assume that prior political experience and ties to different interest groups tend to make ministers less dependent on the president and on their own ministerial position.

The attention to a minister's attributes and whether they are a threat in the eyes of the president allows us to estimate the individual risks of the minister being dismissed. We will use this dimension to test a set of hypotheses about the interaction between the president's reputational assets and the ministers' ‘walk-away values’. The relational and contingent nature of ministerial turnover is also examined here.

We argue that presidents with greater reputational assets feel less threatened by ministers with a high ‘walk-away value’, who tend to survive longer in the cabinet. Their party and extra-parliamentary group ties, as well as their expertise, may be seen by presidents as additional gains on top of their own assets. Contrariwise, presidents with limited reputational assets may see these kinds of ministers as a potential threat to their leadership of government or their political future. If this is the case, they may prefer to get rid of them when they make it convenient to do so or when staying together becomes untenable. Thus, we hypothesize that:

Presidents staff cabinets with members of their own party, but also with ministers selected due to political or personal alignments with the president. Even in coalition cabinets, presidents seek to surround themselves with their ‘own people’ as a deliberately constructed barrier of political protection and a way to pursue their own priorities vis-à-vis other parties or even their own party. We argue, ceteris paribus, that presidents with greater reputational assets are more likely to retain their own people, since they are less pressured to negotiate these positions with other political parties or groups.

Indeed, presidents' relations vis-à-vis their ministers are embedded in the political context at hand and will inevitably be affected by unexpected situations of crisis during their term in office. Since in presidential systems collective responsibility for government performance is more diluted than in parliamentary cabinets, if an external shock causes a crisis then responsibility will primarily fall to the president. Economic crises, health emergencies or corruption scandals can pressure the president to undertake policy shifts – ministerial turnover being an immediate signal of such turns (Camerlo and Martinez-Gallardo Reference Camerlo and Martinez-Gallardo2018; Camerlo and Pérez-Liñán Reference Camerlo and Pérez-Liñán2015; Martínez-Gallardo and Schleiter Reference Martínez-Gallardo and Schleiter2015). We assume that if a president's responsibility is considered directly related to those crises, their political leverage will decline and the reputational assets of both president and ministers will become more apparent. As chief executives, presidents will assess the costs of the crisis vis-à-vis those derived from sacking a minister or an entire party from the cabinet.

Research strategy: ministerial turnover

To give an overview of ministerial changes in presidential cabinets, we begin by testing the arguments concerning the effects of the president's political leverage and reputational assets – H1 to H2 – on ministerial turnover in Latin America. The dependent variable Ministerial Turnover measures the proportion of ministers who left the cabinet each year, meaning the overall ministerial turnover rate. We exclude the turnover of the first year only to avoid counting the changes related to the inaugural cabinet. The ministerial turnover rate captures the most radical ministerial makeover because it does not include other types of changes such as the reassignment of ministers across portfolios, which is a common measure to modify the pay-offs of ministers/parties or to signal government policy shifts. To construct this variable we took the ministerial cabinet retention rate from the WhoGov Data (Nyrup and Bramwell Reference Nyrup and Bramwell2020), which is calculated as the proportion of ministers who remained from one year to the next, adjusting for the size of the ministerial cabinet. However, as the objective is to analyse the extent of ministerial turnover, we used the inverse of the retention rate as a dependent variable. Figure 1 below presents the annual proportion of ministerial changes in 17 Latin American countries from 1979 to 2018. We include only the years in which the countries scored as democratic according to Polity2. Different line lengths for each country are also due to the availability of data on the two main independent variables: Political Leverage and Reputation Asset.

Figure 1. Turnover Cabinet in 17 Latin American Countries, 1979–2018

Source: Authors' own compilation, based on Nyrup and Bramwell (Reference Nyrup and Bramwell2020).

As explained in the previous section, the president's political leverage is a metric that combines three dimensions. We consider the Political Leverage measure simply as the average of these variables. In turn, to measure the president's Reputational Assets, we relied on Baturo and Elkink's (Reference Baturo and Elkink2021) PolEx measure, which, as previously explained, captures the leader's political experience prior to their term in presidential office.

We also control for the number of years the president has been in office (Mandate Course). We are agnostic about the direction of the effect of this variable. We added two political variables: namely, the type of government (whether a single-party or coalition government) (single-party is the reference category), and the country's democratic status (Polity2).Footnote ² We also control for economic aspects (GDP Growth and Terms of Trade), which in the unstable economic environments of the region can change the configuration of political dynamics.Footnote ³ Descriptive statistics for all these variables are available in Online Appendix 1.

Estimation method

Figure 2 below presents the results of the regression model with the dependent variable Turnover Cabinet, which, as explained above, measures the proportion of ministers who left the presidential cabinet in each year in any given country. The unbalanced panel consists of information from 17 countries.Footnote ⁴ Due to the nature of the data, it was possible to test different specifications on the relationship between ministerial turnover and our two variables of interest: Political Leverage and Reputational Assets. Here, we present the model that best fits the robustness tests. In this way, Figure 2 presents the estimates of the panel model with random effects.Footnote ⁵

Figure 2. Effects of Political Leverage and Reputation Assets on Turnover Cabinet: Regression Models

Note: The coefficients are random-effects panel-data estimators with clustered robust standard errors by country. For the significance level in the estimations: *p < 0.1; **p < 0.05; ***p < 0.01. All diagnostic test results are provided in Online Appendix 1.

Figure 2 reveals that a one-point increase in the Political Leverage index reduces by 0.28 points the proportion of ministers who leave office each year, this being the strongest predictor of cabinet turnover. Considering the components of this index, we can state that strong presidents – that is, those with high political leverage – retain their ministers more often, while those who are electorally weak and have less legislative as well as popular support adopt the radical strategy of removing their ministers. This finding is consistent with claims that a lower level of presidential political leverage causes government instability and, based on this, ministerial turnover – as posited by H1. Importantly, the president's political leverage has also proven to be one of the main predictors of ministerial turnover in a region marked by a growing number of minority presidents. The high level of ministerial turnover is linked to the strategies adopted by minority presidents to keep their legislative strength via reshaping their coalitional cabinets, as often happens in Brazil (Inácio Reference Inácio, Camerlo and Martínez-Gallardo2018), and to a president's attempts to reconcile with different factions of their own party (Basabe-Serrano et al. Reference Basabe-Serrano, Polga-Hecimovich, Mejía Acosta, Camerlo and Martínez-Gallardo2018).

Regarding the president's reputation assets, meaning the political experience with which they arrive at presidential office, the effect of our proxy (the PolEx measure) on the yearly proportion of ministers leaving the cabinet is significant and performs in the expected direction – as suggested by H2. Figure 2 shows that this effect is negative: that is, ministerial turnover tends to be lower when the president's reputational assets are higher. In our view, these results support the idea of testing the effect of this variable further, from a relational perspective with ministerial reputations. Our expectation is that this variable is relevant for predicting the chances of micro-level changes: namely, those in the president's relationship with ministers who threaten the former's reputational assets. We explore this further in the next section. Figure 3 pictures the marginal effects of our core variables, president's reputation assets and political leverage.

Figure 3. Marginal Effects of a President's Reputational Assets and Political Leverage on the Minister's Risk of Dismissal

In addition, the model depicted in Figure 2 includes a series of controls. These are contextual or systemic factors potentially conditioning the president's decision to dismiss ministers, thus having an impact on turnover among the latter. The results indicate that the number of years the president has been governing (Mandate Course), type of government, Polity2 and terms of trade have no significant impact on turnover. Interestingly, of the variables controlling for exogenous shocks only GDP Growth has an impact in the expected way. This is somehow strengthened if we assume that shocks indirectly impact political leverage via presidential popularity rates (as covered in the variable Political Leverage). In any case, we see that Political Leverage and Reputation Assets are stronger predictors of cabinet turnover, thus suggesting that this is a tool used more to carry out political adjustments in view of internal (president–minister) conflicts than as a means of directly reacting to exogenous (economic) shocks.

From this first empirical analysis, we conclude that the president's political leverage and reputational assets, as in H1 and H2, are the main reasons for the overall ministerial turnover in the course of such an administration. Next, we move further with the connection of the individual features of both the ministers (what type of minister is more likely to be dismissed) and presidents (what type of president fears a reputational loss in keeping such a minister).

Research strategy: ministerial tenure

In this section, we provide evidence on H3 and H4, which connect the president's reputational assets and the ministers' ‘walk-away values’. We test them as follows: we assume that presidents' fear of reputational losses is related to the reputational assets they hold when coming into office (which we already presented and measured as PolEx in the previous analysis). This is a key element to understand how presidents handle their relations with ministers, and how they decide who leaves the government and when. It can also spur a minister's dismissal based on their personal characteristics.

We use the databases produced by country experts from the Presidential Cabinets Project, which provide a diverse set of information on ministers' personal attributes, career trajectories and political networks.Footnote ⁶ The unit of analysis of the panel data, country and year, is each time he or she was appointed. This analysis was conducted for the following countries and time spans: Brazil (1985–2016), Colombia (1979–2014), Ecuador (1979–2015), Peru (1980–2014) and Uruguay (1984–2015).

We test the role that the president's reputational assets and political leverage play on ministerial tenure, applying survival models (or event-study models). Specifically, we apply non-parametric Cox proportional hazard models because we do not need to specify a shape of the hazard function a priori. Differing from the previous analysis, in this one our dependent variable is individual Ministerial Tenure, or the time elapsed between the minister's appointment and their later dismissal. Our data include the minister's departure due to resignation, reshuffling or the end of the presidential term. Thus, the dependent variable, which is essentially a dichotomous outcome (the event has occurred or has not), captures the different risks to which ministers are exposed from reassignment to another position (either a promotion or demotion) in the cabinet to being removed from the latter entirely.

We account for the fact that the effects of our independent variables may vary depending on the type of ministerial departure. On the one hand, the president can decide to make more moderate changes, reassigning ministers between portfolios, promoting or demoting them. We recognize that these are important strategies signalling the president's satisfaction or dissatisfaction with the minister or own party and they may threaten removal in future. On the other hand, acting more radically, the president can remove the minister from office. These diverse presidential strategies indicate different motivations for the ultimate removal of ministers, which in turn we expect will be related to the latter's ‘walk-away values’. To deal with this heterogeneity, we first estimate models with the whole sample of modalities of cabinet changes (Figures 4 and 5 below) and then for the subset of ministers who are removed from the cabinet (Figure 6).

Figure 4. Ministerial Tenure: Kaplan–Meier Survival Probability

Source: Presidential Cabinet Project

Figure 5. Cox Proportional Hazard Models of Minister Survival (Total Sample)

Figure 6. Cox Proportional Hazard Models of Minister Survival (Restricted Sample)

To analyse ministerial survival, we need to define a cut-off point in time (data censorship) and we consider the ministers who left their posts – regardless of the reason – as events. However, if the minister left office at the end of the presidential mandate, this departure is not considered an event. One way to analyse the ministerial-survival pattern is to consider the Kaplan–Meier survival-estimates process. This allows us to observe temporal behaviour and, in general, illustrates changes in cumulative probability as the curve advances over time. Kaplan–Meier Survival Probability is an initial way to explore ministerial survival, as shown in Figure 4.

Figure 4 shows that the survival rate is not uniform across countries. Uruguay leads in ministerial endurance while Peru exhibits the greatest instability; these are, indeed, cases of opposing extremes in terms of political personalization, with Uruguay characterized by strong parties and a stable party system (Chasquetti and Buquet Reference Chasquetti, Buquet, Camerlo and Martínez-Gallardo2018) and Peru by weak parties that are either unstable or context-contingent structures formed around the president (Vera and Carreras Reference Vera, Carreras, Camerlo and Martínez-Gallardo2018). Additionally, the Peruvian Congress has prerogatives to confirm and dismiss ministers – an institutional feature that evinces further checks on the president and is unusual in comparison to the other presidential systems of the region. Politics is also personalized in Ecuador, the country with the second-highest level of ministerial instability – together with marked presidential instability in the 1990s and first decade of the new century, as well as a party-system collapse in the latter time period (Basabe-Serrano et al. Reference Basabe-Serrano, Polga-Hecimovich, Mejía Acosta, Camerlo and Martínez-Gallardo2018).

Estimation method

We build five models to estimate the hazard rate of ministerial survival in connection with our main predictors of cabinet changes: presidents' political leverage and reputational assets.Footnote ⁷ To test whether the minister's ‘walk-away value’ tempers the effects of our main variables, we use a set of binary variables on ministerial characteristics. The variable President's Own People captures ministers' closeness to the president, classifying in this way the ministers of the president's own party as well as the non-partisan ministers who were directly selected by the president. Minister's Party Attachment indicates whether the minister is non-partisan (reference) or partisan. Minister's Policy Expertise classifies the ministers as a non-expert (reference) or policy expert. We use four binary variables indicating Minister's Political Network: namely, the minister's prior links to extra-parliamentary groups (union, religious, economic and military). In addition, we use two control variables. Type of Government takes a value of 1 if the government is a multiparty coalition and 0 otherwise. Gender, as a binary variable, has a value of 1 for male and of 0 for female ministers. Descriptive statistics for all these variables are available in Online Appendix 1.

To interpret the results, it should be noted that the values correspond to hazard ratios. Thus, values above 1 imply an increase in hazard ratios and values below 1 indicate a reduction thereof. The estimates of the basic and extended specifications of our models can be consulted in Online Appendix 1. Figure 5 presents the estimates of the full model (number 5 in the table in the Online Appendix), which includes all variables and three interaction terms between the president's reputational assets and the minister's ‘walk-away value’ (own, party-related and expertise). These are meant to estimate a minister's chances of survival by associating variations in his or her profile with the president's personal incentives to fire them. This model includes the whole sample of actions that the president can undertake to remove the minister from their position.

The variable Reputational Assets performs as expected in Models 1–4 (see Online Appendix 1): that is, presidents with greater such assets mean reduced risks of ministerial changes. A different picture emerges when we interact Reputational Assets and ministers' ‘walk-away values’, as depicted in Figure 5. Our findings show that the greater reputation of the president reduces the risk of dismissal for policy-expertise ministers – confirming partially H3 – and the president's own people – as posited by H4, a finding that is line with other studies (Camerlo and Pérez-Liñán Reference Camerlo and Pérez-Liñán2015). This effect is not clear in cases of partisan ministers. In fact, in the basic and other models (see Online Appendix 1), partisan ministers face a greater risk of dismissal compared to non-partisan peers. But when we interact this variable and the president's reputational assets the direction of this effect changes, as we expected, but it is not statistically significant. As the effect of the ‘reputational assets * own people’ interaction indicates, the protective effect of party attachment on ministerial tenure seems to be unambiguous when treating ministers who are members of the president's own party and ministers who are loyal and friends.

Political leverage has been shown to be a strong predictor of ministerial turnover. However, its effect is not clear when it comes to identifying which minister profile is more at risk of dismissal. The results indicate that the president's higher degree of political leverage tends to reduce the risk of ministers' resignation, as expected in H1, but this effect has statistical significance only in the simple models.

A final interesting result indicates to what extent the dismissal of ministers with extra-parliamentary ties is sensitive to the type of group to which the minister is close. Ministers connected to unions are the ones at greater risk of being displaced, perhaps indicating the large chances that policy orientation and social alliances will experience shifts within the course of a presidential term. Results for other ministers' connections resulted without statistical significance. The results partially confirm our hypothesis that reputationally strong presidents tend to retain ministers with high ‘walk-away values’, particularly policy experts, and those personally close to them.

Now we move to the results we obtained when accounting solely for the removal of a minister. As previously explained, removals can be understood as a more radical and costly strategy for the president than displacing a minister through a promotion, demotion or reassignment to ministries of different political or administrative relevance within the cabinet. The models presented in the previous figures checked all those cases, but how do the ministers' and the presidents' attributes behave if we consider only that they remove ministers from the cabinet? We estimate the effects of the same variables considering only the sample of those who were fired – either individually or as a consequence of a broader cabinet reshuffle. Figure 6 presents the results corresponding to Model 5, the one with all the controls (all models can be consulted in Online Appendix 1).

On the one hand, the effects of the Political Leverage variable are quite interesting in this analysis. Its effects are not significant when all types of ministerial dismissals are considered (total sample), whereas they are very clear when the president decides to remove the minister from the cabinet. The higher the political leverage of the president, the lower the risk of ministers being removed permanently from the cabinet, as in H1, thus reiterating results we had obtained in the first empirical strategy (Figure 2). This confirms that presidents with higher parliamentary support and popular approval seem to prefer moderate changes to their cabinet, such as reassigning ministers to other positions, realigning their team or signalling to a minister that they are at risk of dismissal.

On the other hand, the interactions reveal that in the case of cabinets led by a reputationally strong president, ministers with high ‘walk-away values’ due to their party ties and their proximity to the president (President's Own People) are at lower risk of leaving the cabinet. Interestingly, this discovery indicates that while the effect of partisan attachment is not significant in the previously shown analysis of all types of ministerial changes, it is important when it comes to removing a minister from the cabinet outright (the restricted sample). In this case, party affiliation seems to increase the bargaining power of ministers, and probably of their parties too. The effect of protection is unclear in the case of ministers who are policy experts when the decision is to push them out of office. Regarding network linkages, ministers close to trade unions and economic groups are more likely to be removed from office.

A final remark on two variables. First, the effects of the variable Type of Government for ministerial survival as estimated using these two samples of dismissals. This variable has null effect when all ministerial changes are considered, even though it reduces the risk of removal from the cabinet. These results are consistent with the literature on coalition presidentialism, which points to the smaller leeway of the president who decides to share the government with other parties. The use of the more radical strategy – that is, to definitively dismiss a minister – seems to be avoided out of presidential fear of hurting parties and losing their support. Second, regarding Gender, Figures 5 and 6 indicate that women are less likely than men to be displaced or demoted to other positions.

Overall, the results presented here raise some important questions for the study of cabinet changes. In particular, we found evidence that the risks of a minister being reassigned or fired vary considerably when we consider the interplay between the profiles of presidents and ministers. We identified how the president's reputational assets matter and have a potential protective effect for some ministerial profiles. Even ministers who may represent a challenge to the chief executive's leadership – due to their party ties and policy expertise – may survive longer in the cabinet when the president has large reputational assets. Different from the expectation that any president protects their own staff, we showed how this is more likely to occur among those with strong reputations since they are less pressured to move their allies in ministerial changes. A further implication of our findings is that the president tempers the strategies of cabinet changes, from the more moderate to the more radical ones, according to their reputational fragility or strength and to the ‘walk-away values’ of their ministers.