War, Performance, and the Survival of Foreign Ministers

Abstract

Why do some foreign ministers stay longer in office than others? Are they punished when the country loses a war? Several scholars have focused on the tenure of leaders as an important predictor of foreign policy outcomes, such as war onset, creating an interest in leadership survival. We here shift the focus to the survival of other important politicians in cabinet—foreign ministers, hypothesizing that their tenure depends on their performance in office. For example, we expect that foreign ministers stay longer in office when the country experiences an armed conflict resulting in a win or in a compromise agreement. We evaluate and find support for several of our hypotheses using an original historical dataset, which comprises all foreign ministers of the world's thirteen great powers from the early modern period to the present, covering about 1,100 foreign minister-terms of office.

Introduction

In line with a long-standing tradition in the field of foreign policy analysis (e.g., Hermann 1980; Byman and Pollack 2001), recent literature has increasingly recognized the importance of individual leaders when explaining foreign policy outcomes (e.g., Horowitz, Stam, and Ellis 2015). In the literature, several scholars have focused on the expected survival of leaders as an important predictor of war onset. For example, Debs and Goemans (2010) show that compared to autocratic leaders, the tenure of democratic leaders does not depend on war outcomes and offer an explanation to why democratic leaders are more willing and able to avoid war. We contribute to this literature, following the work by Quiroz Flores (2009, 2016), by shifting focus to the survival of another important politician in cabinet: the foreign minister (or secretary of state). The question we ask is as follows: Does the survival of foreign ministers, similarly to the head of government, depend on foreign policy outcomes, such as a win or a loss of war?

The specific aim of this paper is to inquire into why some foreign ministers sit longer on their posts than others. Although obviously not the sole outcome of interest, survival in office can be seen as a proxy for prominence and influence. “Evidence at hand,” according to Modelski (1970, 144), “indicates that longer-tenured foreign ministers are better known by more of their colleagues, are more highly regarded by them, and tend to be more active in the international arena.” In this sense, we are trying to explain what makes some foreign ministers more successful than others.

We contribute to the previous literature by evaluating original hypotheses about the features that influence foreign ministerial survival. We expect that specific background features of the individual foreign ministers influence their tenure, for example, hypothesizing that foreign ministers with a diplomatic background have a lower risk of losing their posts because they are expected to perform well in office. We also hypothesize that the survival of foreign ministers depends on situational features, and that foreign ministers are less likely to survive after a loss of war, either because this can be perceived as a “poor performance” on their part or because heads of governments (HoGs) will shift blame within cabinet to avoid losing office. However, we expect that foreign ministers are not only punished for poor performance, but also “rewarded” for performing well. Thus, when a country has experienced an armed conflict resulting in a win or in a compromise agreement, the foreign minister's tenure will be longer.

We evaluate and find support for several of our hypotheses by analyzing a new and unique comparative historical dataset on foreign ministers’ background and reasons for leaving office since 1789. The dataset comprises all foreign ministers of the world's thirteen great powers from the early modern period to the present (Levy 1983): Austria, Britain, China, France, Italy, Japan, the Netherlands, Prussia/Germany, the Ottoman Empire, Russia, Spain, Sweden, and the United States, covering about 1,100 foreign minister-terms of office.

Overview of the Field

The Literature on Foreign Policy Leaders

In this paper, we draw on the previous literature that assumes that foreign policy outcomes that are to be explained are the result of human decision-making (Hudson 2005). Much work in this field focuses on the decision-making of political leaders, where several scholars have found that “the core psychological characteristics of presidents and prime ministers affect their personal policy preferences and the policies adopted by the states they lead” (also see Hermann 1980; Byman and Pollack 2001; Crichlow 2005, 179).

A number of scholars have more recently gathered and analyzed comparative systematic data on the personal features of individual leaders. For example, Horowitz, Stam, and Ellis (2015, 12) present results that evaluate, on the basis of the background experience of 2,400 leaders (1875–2004), “the probability that a leader will engage in interstate military conflicts while in office.” Later work shows that features such as combat and rebel experience affect leaders’ dispute selection and the effectiveness of coercive threats (Horowitz et al. 2018). These scholars build on the important work of the creators of the Archigos dataset, focusing mainly on modes of leader entry and exit (Goemans, Gleditsch, and Chiozza 2009).

As described by Crichlow (2005, 180), most previous research on foreign policy leaders has focused on prime ministers and presidents. However, these individuals are not the only important politicians in the foreign policymaking process. Chrichlow (2005, 180) argues that if we are to fully understand the impact of leaders on foreign policy outcomes, we need to investigate the role of subordinates of presidents and prime ministers, suggesting that:

History is replete with examples of extraordinarily influential foreign ministers, from Talleyrand to Schuman and Adams to Kissinger […]. Foreign ministers are officially charged with directing foreign policy […] foreign ministers typically have a great deal of discretion, and their institutional resources and authority, plus the legitimacy they possess as the government's primary voice on foreign affairs, allow them to move policy in directions they favor.

There are thus a number of reasons for focusing on individual leaders when aiming to explain foreign policy outcomes. So far, however, relatively little systematic comparative work has focused on foreign ministers, and their personal characteristics. One important exception is the early work by Modelski (1970) that presents information on the background and “interactions” of the 175 foreign ministers who held office in 1965. Modelski, for example, shows that most of these individuals (about 80%) have a college or university education, often law and legal training, and that relatively few foreign ministers (about 20%) have military experience. Modelski (1970, 149) also shows that when it comes to occupational experience, many foreign ministers have a background in political office or within the party, but also that “the diplomatic service is another clear source of recruits.”

Oppermann, Brummer, and Willigen (2016) suggest that foreign policy research should open up the “black box” of coalition governance in foreign affairs as this type of endeavor “promises more fine-grained insights into the drivers and characteristic of coalition foreign policy.” More specifically, they suggest that foreign policy is shaped by which party is allocated the foreign ministry, and whether the foreign minister has a high level of policy discretion. In a recent article, Oppermann and Brummer (2018) focus on predicting which party is allocated the highly prized foreign minister post in coalition negotiations.

The Literature on Leadership Survival, Foreign Policy, and War

Some of the conflict literature has focused on understanding leadership survival, since leaders’ expected survival should influence their foreign policy decisions, such as their willingness to go to war or to make bargaining concessions (e.g., Bueno de Mesquita and Siverson 1995). This literature suggests that whether leaders wage war or agree to concessions depends on political institutions, with democratic leaders having a propensity to fight wars that they can win because victory in war is considered a public good, which is essential for survival in office in large coalition systems (Bueno de Mesquita et al. 2003).

In terms of concessions, the logic of this argument is that leaders understand not only that concessions at the bargaining table could lead to peace, but also that such concessions could affect their survival as leaders. Therefore, the “size of the concessions he or she is willing to make depends on the sensitivity of his or her survival to the share of the pie obtained internationally” (Debs and Goemans 2010, 430). This argument stresses the importance of understanding when and why leaders survive. Even though this literature focuses on the survival of leaders, we suggest that it is highly relevant for our argument about foreign ministers. Since foreign ministers are likely to take part at the bargaining table, their survival is highly relevant to research aiming to explain war and other foreign policy outcomes.

Whereas some of these previous studies have argued that the postwar tenure of leaders in democratic regimes is more sensitive to war outcomes (see, e.g., Bueno de Mesquita et al. 2003), recent empirical analyses seem to suggest the opposite (see, e.g., Debs and Goemans 2010). Croco and Weeks (2016, 578) attempt to solve this “puzzle,” suggesting that previous work has overlooked the leaders’ “culpability” for a conflict. That is, when analyzing leaders’ tenure and likelihood to survive, it is important to consider that the “domestic audience's willingness to sanction a leader” may vary both across and within countries. Here, a leader's perceived culpability for war is affected by whether the leader was in power when the war began. The degree to which a domestic audience is capable to punishing a leader also influences the leader's vulnerability and their likelihood to survive in office.

Another potential explanation for why democratic leaders do not lose their posts after a defeat in a war, also related to the argument about “culpability,” is that other actors may be taking the blame for “poor performance” in such systems, such as foreign ministers, acting as policymakers or advisors of HoGs. The underlying idea is that leaders can “shift blame” to other actors when losing a war, a point which we elaborate on in this paper. This kind of blame-shifting has important implications for reputational theories of leaders and international politics (e.g., Wu and Wolford 2018).

The Literature on Ministerial Survival

The comparative literature on ministerial selection departs from a view of parliamentary democracies as based upon a chain of delegation in which the prime minister (PM) acts as “principal” of the ministers who are acting as “agents,” a view that is transferable to other systems, for example, presidential ones, but with somewhat more complex delegation relationships. According to principal–agent theory, the principal employs control mechanisms to mitigate “agency loss.” For example, ex ante control mechanisms such as extensive screening of potential candidates for ministerial posts can be applied. The most important ex post measure is to end the principal–agent relationship, that is, to dismiss ministers that are deemed incompetent, disloyal, or exceeding their range of discretion, and this is typically seen as an effective instrument for the PM to sanction agency loss (e.g., Strøm 2003).

The previous literature has recognized that PMs may use cabinet reshuffles for several reasons, such as blaming policy failures or scandals on individual ministers (e.g., Dewan and Dowding 2005). One important task of the head of government is to monitor the performance of ministers in order to detect “agency loss.” Berlinski, Dewan, and Dowding (2010, 559) argue that the PM “evaluates her ministers according to information available to her that is related to their performance,” and test this argument by analyzing the number of resignation calls for a minister as an indicator of individual performance.

Focusing specifically on foreign minister survival, Quiroz Flores (2009) argues that affinity and loyalty toward a leader—particularly in autocracies—as well as the uncertainty brought about by leadership change and time dependence, strongly determine a minister's tenure. Quiroz Flores and Smith (2011) formalize this argument, arguing that leaders face internal and external threats to their hold on power. Internal threats such as coups are more salient in autocracies and therefore autocrats remove high-performing cabinet ministers, as they are potential challengers to the leader.

Theory and Hypotheses

Drawing on these literatures, we here elaborate on our hypotheses. We are here specifically interested in the personnel decisions made by HoGs. Hence, we should take our starting point in the goals of these leaders. Following the previous literature, HoGs (prime ministers, presidents, kings, etc.) are assumed to be primarily interested in staying in office. In democratic systems, this implies that they are instrumentally vote-seeking, that is, they should to some extent aim to “please their voters”. In addition, leaders in any regime may also be policy-seeking in the sense that they are interested in implementing a specific policy program in order to remain in office (Bueno de Mesquita et al. 2003). For office-, policy-, or vote-seeking reasons, leaders should thus be interested in hiring and keeping high-performing ministers and should aim to make personnel decisions that minimize agency loss.

However, as has been noted in the literature on ministerial selection and survival, HoGs are not always free to staff the cabinet as they see fit, without the approval of some other actors. Comparative research also shows that the risk that a minister gets fired is lower in coalition governments, suggesting that the PM is constrained when making personnel decisions in such cabinets (Huber and Martinez-Gallardo 2008). Hence, even if we restrict ourselves to parliamentary democracies, it is not always clear that the PM is the main principal in the ministerial appointment process, and ministers in coalition governments may clearly have competing principals (e.g. Bäck, Debus, and Müller 2016).

The literature on decision units and foreign policymaking make similar claims. Here, scholars have suggested that it is important to consider that some governments are characterized by a “sharp fragmentation of political authority within the decision unit” (Hagan et al. 2001,170), arguing that any actor in the decision unit may be able to “block” or “veto” initiatives of the other actors, for example, by threatening to leave the coalition, thereby bringing down the government (Hagan et al. 2001, 170). This literature also gives us some suggestions on other types of regimes that may be characterized by fragmented decision units. Hagan et al. (2001, 172), for example, suggest that authoritarian regimes with power dispersed over separate factions or groups may be seen as fragmented decision units.

Hence, it is clearly a simplification to say that the HoG is always making decisions about ministers’ tenure. Even though reality is more complex, we believe that when it comes to the hiring and firing of foreign ministers, other actors—such as the party (in both democratic and authoritarian single-party regimes) or the ruling coalition undergirding a leader's tenure in office in more personalized dictatorships (Svolik 2012) —should have similar interests as the HoG when it comes to the performance of foreign ministers. That is, they are likely to reward positive performance and punish poor performance.

We present two types of hypotheses relating to the performance of foreign ministers. The first set of hypotheses address ex ante performance, which is strongly determined by the individual background features of foreign ministers. The second set explores ex post performance, determined by situational features such as the level and outcomes of conflict. As argued by Quiroz Flores and Smith (2011), new ministers by definition have not revealed their level of performance—competence or performance, however defined or measured, is an unknown quantity. Yet, some ministers are more experienced than others—all else equal, ministers with a long career in the diplomatic service are expected to perform better during international conflicts than ministers without any diplomatic experience. In this light, expected performance keeps ministers in office, at least until a sufficient quantity of actual performance is observed. This leads to our situational hypotheses, which center on the notion that foreign ministers may be blamed or rewarded for the situation that the country is in. In other words, we refer to actual performance and the perception it produces about a minister's competence.

We acknowledge that numerous actors observe ministerial performance and update their beliefs or perceptions about a minister. We also note that perceptions of observed performance can be easily manipulated, particularly by HoGs. For instance, victory in a war is similar to a public good in that relevant parties can always argue that they added to a successful war effort; in this context, there is little need for manipulation of perceptions of observed performance. Foreign ministers can correctly argue that war consists of an alternating sequence of negotiations and battles (e.g., Werner 1998) and that as the country's top diplomats they contributed to finding a negotiated settlement that is preferable to the continuation of war. HoGs can argue that they had the vision to hire such successful ministers. However, in the case of failure in conflict, HoGs have incentives to manipulate public perceptions of ministerial performance and may blame a foreign secretary—rightly or wrongly—for a failed conflict.

Hypotheses about Individual Background Features and Foreign Minister Survival

In the literature on cabinet reshuffles, a number of background features of the ministers are expected to influence ministerial duration. Berlinski and colleagues (2007, 248) argue that “Ministerial performance is likely to be related to personal abilities and these might be indirectly related to their characteristics.” Much of the literature on cabinet reshuffles highlights cabinet experience as an indicator of ministerial ability, also allowing for intensive screening of potential candidates. This might lead one to expect that experienced ministers will survive longer in their posts (see, e.g., Huber and Martinez-Gallardo 2008).

This literature suggests that the extensive screening of potential cabinet members serves to effectively minimize the risk of agency loss, which implies that the HoG's need to fire ministers is reduced. In particular, the legislative and party venue allows ex ante screening of potential ministers in terms of their competence and loyalty, assuming that the screening of prospective parliamentarians and potential cabinet members is performed by centralized, cohesive, policy-oriented political parties. Hence, having a background within political parties, in parliament, or in the executive, should increase the expected performance of ministers and should decrease the risk that ministers are forced to leave their posts. We thus hypothesize that:

 

Hypothesis 1:

H1:Foreign ministers with a political background have a lower risk of losing their posts.

In addition, we suggest that the diplomatic experience of foreign ministers should be especially important to expected performance since these ministers have a specific role within the cabinet, leading negotiations with other countries’ representatives. As described by Quiroz Flores (2009, 118), “as the highest diplomats in government, [foreign ministers] represent the sovereign state in one of its most important functions, that is, external relations.” Having diplomatic experience and a potentially positive “track record” in terms of external relations and international negotiations could therefore decrease the minister's risk of being forced out of office because they are expected to perform well in this position. We thus hypothesize that:

 

Hypothesis 2:

H2:Foreign ministers with a diplomat background, have a lower risk of losing their posts.

Specific backgrounds may thus signal their potential for high performance in the foreign minister post in any system. However, as mentioned before, the previous literature on foreign ministers argues that performance can lead to minister deposition in autocracies and some democratic parliamentary regimes. We here add to this literature and qualify the effect of performance or “competence” as follows. Although highly competent cabinet ministers will perform excellently across ministerial portfolios, the average minister tends to perform well only on one or two positions. More importantly, good performance in one cabinet position does not necessarily imply that a minister is a credible challenger to the incumbent leader.

In this light, we argue that some backgrounds, resulting in competence on some portfolios, might lead to ministerial replacement in autocracies, as demonstrated by Quiroz Flores (2009, 2016). Nevertheless, competence in other portfolios might not necessary lead to cabinet change, even in autocracies. For instance, we expect that foreign ministers with a diplomatic background, who are likely to perform well in office, are not necessarily credible challengers to the incumbent leader, and thus we do not expect that diplomatic experience reduces their tenure, even in autocracies. Considering the importance of this argument, we evaluate the impact of background features conditional on the level of democracy, expecting that:

 

Hypothesis 3:

H3:Foreign ministers with a political background face a lower risk of losing their posts in democratic systems, but face a higher risk of losing their posts in autocratic systems.

Hypothesis 4:

H4:Foreign ministers with a diplomat background have a lower risk of losing their posts in both democratic and autocratic systems.

Hypotheses about Conflict Level and Outcomes and Foreign Minister Survival

Following the previous literature on leadership survival, we expect that the international conflicts that a country is involved in matter for foreign minister survival. Bueno de Mesquita and Siverson (1995) argue that there are strong reasons to believe that there is a close connection between war and the domestic fate of political leaders. Specifically, Bueno de Mesquita et al. (2003, 2004) contend that leaders maximize tenure in office by providing an optimal mix of public and private goods. The particular mix depends on political institutions and particularly on the size of the selectorate and the winning coalition. The selectorate is the group of individuals with the institutional prerogative to select the leadership of their country, while the winning coalition—a subset of the selectorate—is the group of people whose political support is necessary to hold office. As the size of the winning coalition increases relative to the selectorate, leaders provide more public goods in order to keep the loyalty of their supporters.

In the work of Bueno de Mesquita et al. (2003, 2004), as well as in further extensions (e.g., Bueno de Mesquita and Smith 2009), this argument is used to explain why leaders fight wars. Bueno de Mesquita and colleagues assume that victory in war is a public good. Since leaders in large coalitions need to provide more public goods to stay in office than their autocratic counterparts, democrats tend to fight wars that they can win, as fighting a losing war will only bring deposition. This logic should also apply to foreign ministers.

Quiroz Flores (2016) has explored part of this connection in his empirical analysis of foreign ministers, basing his measure of ministerial performance on a well-known result that shows that governments can reach pre-war bargains leaving them as well off as fighting a costly war. In this light, high-performing foreign ministers are those who are better able to reach these ex ante bargains, producing longer periods of peace. Here we take a somewhat different approach by exploring the effect of performance during conflict. Indeed, the initiation of hostilities could be the result of events exogenous to the bargaining process, which provides an opportunity to explore the effect of performance during war. Moreover, we examine Quiroz Flores's untested argument that the worst performing foreign ministers will not only fail to avoid war, but also start a conflict that they will lose.

The question is as follows: How does being involved in international conflict affect the survival of foreign ministers? We suggest that HoGs, that are likely to be seen as being “culpable”, have an opportunity to “shift blame” within the cabinet, and to make foreign ministers responsible for the country being at war. As shown by Dewan and Dowding (2005, 46), the government as a whole can “pin blame on individual ministers and deflect criticism and subsequent falls in popularity by sanctioning or removing the minister concerned.” It may of course be that foreign ministers “have a great deal of discretion” (Crichlow 2005, 180), allowing them to actually make important policy decisions that partly explains why a country is involved in conflicts, and in this case, blame is “rightly” attributed to the foreign minister. Regardless if a foreign minister is to blame for a country being involved in conflict, or if the foreign minister is used as a “scapegoat,” foreign ministers should have a higher risk of losing their posts when the conflict level is high. We thus hypothesize that:

 

Hypothesis 5:

H5:The higher the level of conflict, the higher the risk that foreign ministers lose their posts.

As pointed out above, the previous literature has found mixed results regarding conflict outcomes and leadership survival (e.g., Bueno de Mesquita and Siverson 1995; Croco and Weeks 2016). This may partly be due to the fact that these studies do not differentiate between individual leadership survival and regime survival, where the latter is likely to be more directly influenced by war outcomes in that (autocratic) leaders often lose office through “violent means,” such as rebellions or coups (Debs and Goemans 2010, 435). We suggest that foreign minister survival may be somewhat easier to predict, considering that we are here not studying “regime survival,” but rather the individual survival of foreign ministers (as the HoG stays in office), and their performance should be more clearly linked to their tenure. We suggest that conflict outcomes are generally “positive” or “negative” and that a minister is likely to be “blamed” or “credited” for these outcomes, being the “government's primary voice on foreign affairs” (Crichlow 2005, 180). For example, conflicts resulting in “win” or a “compromise agreement” should benefit a foreign minister. The latter is likely to be characterized as a positive performance by the foreign minister since part of the job lies in leading negotiations with other countries’ representatives, and reaching a compromise could thus be seen as the foreign ministers “doing their job.” We thus hypothesize that:

 

Hypothesis 6:

H6:When a country has experienced an armed conflict resulting in a positive outcome (“win”), there is a decreased risk that foreign ministers lose their post.

Hypothesis 7:

H7:When a country has experienced an armed conflict resulting in a compromise agreement, there is a decreased risk that foreign ministers lose their post.

Lastly, we follow the previous literature on leadership survival in expecting that losing a war is negative for leaders. One reason for this is that a loss may “signal incompetence” and that “the domestic audience cares about keeping competent leaders” (Debs and Goemans 2010, 431). Hence, a defeat in war is likely to signal poor performance of the government in general, which may lead to that HoGs “pin the blame” on the foreign minister—therefore a “loss” of a war should decrease their tenure. Or, a defeat in war may also signal that the foreign minister, as a policymaker has “performed poorly.” Regardless of the mechanism, we expect that:

 

Hypothesis 8:

H8:When a country has experienced an armed conflict resulting in a negative outcome (“loss”), there is an increased risk that foreign ministers lose their post.

Methods and Data

Research Design and Case Selection

We here draw on a unique dataset covering foreign ministers or secretaries of state in thirteen former or current great powers: Austria (the Habsburg Empire/Austria–Hungary), Britain, China (Qing Empire/Republic/People's Republic of China), France, Italy, Japan, the Netherlands, the Ottoman Empire, Prussia/Germany, Russia/USSR, Spain, Sweden, and the United States. Although our data only cover the post-1789 period, we thus include all thirteen great powers in the international system from the early modern period to the present (Levy 1983). Apart from the fact that we are thus arguably capturing the most influential foreign ministers in modern times, this sample of countries also provides ample variation both across countries and over time in regime characteristics and conflict propensity.

The starting point of the data collection for each country, if later than 1789, depends on what the term “foreign minister” more exactly is interpreted to mean. A broad definition would be the highest official in a country/state “exclusively or at least mainly concerned with the formulation and carrying-out of foreign policy” (Anderson 1993, 73). This definition leaves two problems unresolved. The first is how functionally differentiated the role of foreign minister must be from other government posts, most importantly, the one as HoG itself. Otto von Bismarck is here a case in point, simultaneously holding the position as Reichskanzler (under the Emperor) and Prussian Minister of Foreign Affairs, the latter de facto implying foreign minister for the whole German Empire. Second, a foreign minister, to count as such, must have at least a minimal amount of policymaking authority. Purely administrative personnel that only carry out the decisions of rulers higher in the hierarchy should be excluded.

Our solution to both these problems is to operationally only consider foreign ministers as such if they (a) hold such a title (possibly together with other titles) and (b) preside over a ministry of foreign affairs. This would seem to ensure both a minimal level of functional differentiation and policymaking authority. As a consequence, Bismarck and all other German Chancellors up until the fall of Imperial Germany in 1918 except one¹ are considered Germany's foreign ministers. The German “state secretaries,” who formally headed the foreign office but in practice were subordinate to the chancellors in this regard (Doss 1982, 230), are thus excluded. By the same criteria, Russia/USSR enters our series in 1802 with the establishment of the Imperial Ministry of Foreign Affairs, preceded over by a minister (Uldricks 1982, 517); Sweden in 1809 when the title of “State Minister for Foreign Affairs” was first introduced (Carlgren 1982, 458); the Netherlands in 1814 (Wels 1982, 366–68); Italy upon its unification in 1861 (Serra 1982, 298); the Ottoman Empire with the first creation of a foreign ministry in 1836 (Kuneralp 1982, 500); China (Qing Empire) upon the same in 1861 (Hsu 1982, 122–24); followed by Japan in 1869 (Nish 1982, 328).

The Austrian Empire presents a borderline case: although a ministry of foreign affairs was not established until 1848, Clemens von Metternich had held the title of foreign minister since 1809, together with that of state chancellor since 1821. If he were to be included, Metternich would, with 39 years in office, clearly be one of the most enduring foreign ministers in our sample, if not ever (cf. Quiroz Flores 2009, 127). However, since the function of the State Chancellery he presided over was not only to formulate foreign policy (Rumpler 1982, 52), we have by the criterion of functional differentiation from other government positions cautiously excluded Metternich from our sample.²

The UK, United States, France, and Spain present no special problems, all having foreign ministers (or “secretaries of state”) by title and in charge of a foreign office since our starting year in 1789 (Cromwell 1982; De Santis and Heinrichs 1982; Dethan 1982; Smyth 1982). With the exception of the Ottoman Empire, which drops out of our sample at its dissolution in 1922, our data include all foreign ministers up until May 2017.

Comparative Data on Foreign Ministers’ Background and Survival

Through a team of research assistants knowledgeable in the language of the country they coded, we have, based on web searches and biographies, collected biographical information on 1,103 regular foreign-minister terms (some serving multiple terms) from the thirteen selected countries. Some basic personal characteristics of the ministers are presented in figure 1 (for codebook and operational definitions, see the online appendix 3). Much as in Modelski's (1970) global sample from 1965, the modal foreign minister in our sample is a middle-aged man with university education. Gender, age, and education are here included as control variables.

Figure 1.

Personal characteristics of foreign ministers (percentages).

We have also surveyed the personal backgrounds of foreign ministers before entering office. Again in line with Modelski (1970), the majority of foreign ministers lack military background (in terms of education, service, or professionally). There is, however, variation across countries with respect to that rule: Prussia/Germany has predominantly had a preference for putting someone from the military ranks on the post as foreign minister and both France and the United States has had almost as many foreign ministers with military background as without one. We include military experience only as a control variable in our analysis as we do not have any expectations as to whether this feature should increase or decrease tenure.

Perhaps somewhat surprisingly, foreign ministers overall have predominantly not had personal experience with diplomatic service abroad, particularly not in the UK, United States, Spain, and Italy (only in Russia/USSR and the Ottoman Empire is diplomatic experience a modal feature). Yet even in cases like the United States, where diplomatic experience is the exception, there are prominent examples of secretaries of state with such experience. For example, Madeline Albright was Ambassador to the United Nations before being appointed to become Secretary of State and even Thomas Jefferson, the first US Secretary of State in our sample, had diplomatic experience before appointment, having been Minister to France for five years.

In terms of political experience more generally, foreign ministers are clearly no rookies. When counting membership in a political party or background in a legislature, in the cabinet, or other higher government office, only 10 percent of the foreign ministers in our sample lack any such experience. A modern-day example that has received some attention is the appointment of Rex Tillerson, a former energy executive without any political experience, to US Secretary of State by President Donald Trump in 2017.

The mean days of tenure for a foreign minister in our sample is 800 days, but the median is only 412 days, so there is a long tail of unusually long-tenured foreign ministers. In table 1, we present the “hall of fame” of the ten most long-termed ones in our sample. The most long-lived minister is count Karl Nesselrode who served as a foreign minister for the Russian Empire for more than thirty years. Next in line is the Soviet Cold War architect Andrei Gromyko (twenty-eight years). Third is Otto von Bismarck, who first became foreign minister of Prussia in 1862, then kept this post in the North German Confederation from 1866 and, in combination with the post as chancellor, in the unified German Empire from 1871 to 1890 (twenty-eight years).

Table 1.

Foreign minister's “hall of fame”

Name	Country	Date of entry	Date of exit	Days
Karl Nesselrode	Russia/USSR	August 8, 1822	April 15, 1855	14,482
Andrey Gromyko	Russia/USSR	February 15, 1957	July 2, 1985	10,364
Otto von Bismarck	Prussia/Germany	November 23, 1862	March 20, 1890	9,979
Aleksandr Gorchakov	Russia/USSR	April 27, 1856	April 9, 1882	9,478
Aixin-Jueluo Yiyin (Prince Gong)	China	January 20, 1861	April 8, 1884	8,479
Aixin-Jueluo Yikuang (Prince Qing)	China	April 12, 1884	July 24, 1901	6,311
Östen Undén	Sweden	August 12, 1945	September 20, 1962	6,248
Verstolk van Soelen	Netherlands	December 1, 1825	September 13, 1841	5,765
Lars von Engeström	Sweden	March 15, 1809	June 8, 1824	5,564
Count de Floridablanca	Spain	February 25, 1777	February 28, 1792	5,481

Not all foreign ministers leave their posts for the same reason, however. We have therefore also collected information on the primary modes through which they exited office, presented in table 2. Some die by natural causes on their post, such as Nobel Laureate Gustav Stresemann who suffered from a stroke on October 3, 1929, while still foreign minister of the Weimar Republic. Some face a more violent ousting, such as revolution (Max von Baden in Germany in 1918), deposition by a foreign power (Wilhelm Wolf during “die Anschluss” in Austria in 1938), or even assassination (Walter Rathenau of Germany in 1922, Englebert Dolfuss of Austria in 1934, and Anna Lindh of Sweden in 2003). A more common exit is retirement due to ill health or other types of unforced resignations, but the most common exit is simply that the term of the entire cabinet reaches its end, primarily due to term limits, elections or other constitutionally mandated government terminations.

Table 2.

Why foreign ministers lose office

Mode of exit	N	%
Death by natural cause	22	2.0
Violently	37	3.3
Retirement/unforced resignation	196	17.8
End of government term	427	38.7
Forced resignation	341	30.9
Missing:
Incumbent	12	1.1
Unknown	68	6.2
Total	1,103	100%

Notes: The “forced resignation” category consists of nine subcategories (number of occurrences in parentheses): “Political scandal” (14), “Policy disagreement between minister and premier/PM” (36), “Policy disagreement between minister and monarch/president” (55), “Policy disagreement between minister and own party/other minister” (34), “Personal/departmental error or low personal performance” (24), “Move to other post within cabinet (only concerning the foreign minister)” (68), “Move to another post within cabinet in the general context of a reshuffle” (52), “Loss of eligibility for the post” (3), and “Other reason” (55).

In line with our theorizing of the principal–agent relationship between foreign ministers and their HoG, the mode of exit that mainly concerns us here is what we call forced resignation, comprising slightly less than a third of the sample. This is when the foreign minister is involuntarily removed from office by the HoG. The reasons for such removals may vary, but most importantly consist of political or policy disagreements—about 11 percent of the foreign ministers leave their post due to a disagreement with the premier, president (or monarch), or the party leadership. US Secretary of State Colin Powell is a prominent example of an individual leaving office due to a policy disagreement with the president. In a 2004 Washington post article, Powell expressed discontent with the George W. Bush administration's decision to invade Iraq in 2003 and was after that not reappointed when Bush was reelected in 2004.

Considering that scandals involving cabinet members are given so much attention in the media, the number of foreign ministers leaving office due to a scandal is surprisingly small—only about 1 percent of the ministers are forced to leave their post due to such a reason. Somewhat more common are exits from this post due to, for example, personal or departmental error—about 2 percent of the foreign ministers leave their post due to this reason. A fairly common exit reason (10 percent of the exits) is that foreign ministers leave their post in some kind of reshuffle in the cabinet. Either they are the only cabinet members being moved to another post or there is a general reshuffle involving several ministers.

Situational Measures of Conflict Events

We use data on militarized interstate disputes (MIDs) (Palmer et al. 2015) to construct a collection of measures of the foreign policy environment and performance of the foreign minister. MIDs have a variety of characteristics. We focus on two: hostility level and outcome. The variation in these independent variables is described in figure 2. The hostility level for every MID ranges from “no use of force” (1) to “war” (5). For each day the foreign minister is in office, we took the maximum hostility level of their ongoing MIDs. We then took the mean of that maximum hostility level, generating a measure that signifies the average hostility level over the tenure of that foreign minister, ranging from 0 to 5. Because countries have different propensities for conflict, there is substantial variation in average hostility level by country.

Figure 2.

Situational conflict variables (percentages).

Because it is an average, weighted by day, there is one structural difficulty with our measure. The only way a foreign minister can have a value of 0 for their average hostility level is if there were no MIDs during their tenure. The chances of having absolutely no MIDs increase as tenure duration gets shorter, so while we expect that tenure will get shorter as average hostility increases, our measure has a statistical relationship between the value 0 and tenure that cuts the other way. As a result, we use a dummy for whether there was any MID experienced by the country during the tenure of the foreign minister. We include that dummy in one model and use it to filter on several others. Put differently, for our measure of average hostility to work properly, there needs to be at least one dispute. Accordingly, we limit some of our models to those observations where there was at least one dispute.

MIDs are coded with a variety of potential outcomes. We focus on three: victory, loss, and the resolution through compromise. In each case, the variable is just a dummy for whether the country won, lost, or resolved a MID through compromise during the tenure of the foreign minister. As can be seen in figure 2, each outcome variable is predominantly characterized by zero values. Foreign Ministers who experienced clearly coded MID outcomes are relatively rare. Similar to the dummy variable for whether a MID existed, we use a dummy variable for whether any MID ended during the minister's tenure as a control variable.

The Statistical Model and Censoring

We apply an event history model, also called a “survival model,” which is the standard approach used when analyzing cabinet duration (see, e.g., Diermeier and Stevenson 1999) and cabinet reshuffles or ministerial survival. We are here interested in the duration of individual ministers on their posts and we estimate the effects of various features that increase or decrease the length of time foreign ministers survive at their post (Berlinski, Dewan, and Dowding 2007, 247). An especially important issue when dealing with event history data is “censoring,” which occurs when we cannot observe the entire event history of a unit, and which we can take into account using event history analysis (see, e.g., Box-Steffensmeier and Jones 1997). Here, we have coded the “exits” of all foreign minister appointments up until May 2017, which means that appointments where the foreign minister is in office, that is, where they have not experienced the “event” of leaving their post, at that point in time, are “right-censored”—they are censored since we do not know how long they will be in office.

In addition to treating cases that have not experienced the terminating event when our data collection ended as censored, we also focus on specific terminations or “exits” that are especially important for answering the research question (see, e.g., Diermeier and Stevenson 1999). We here censor all ministerial spells where foreign ministers died at their post, left the post due to health reasons, or where the minister survived until the end of the government. All of these exits are less interesting for us since we are specifically focusing on the personnel decisions made by the HoG. For the same reason, we have also chosen to censor all exits that are not “forced,” that is, when the foreign minister leaves the post voluntarily, for example, taking another post outside cabinet for retirement.³

In event history analysis, we are in general interested in modeling the so-called hazard rate, which can be interpreted as “the instantaneous probability that an event occurs given that the event has not yet occurred” (Box-Steffensmeier and Jones 1997, 1427). Here, the hazard rate describes the probability or “risk” that an individual foreign minister exits his or her post at time, conditional on that the foreign minister has not exited the post before that time. The model that we apply here is a semi-parametric Cox proportional hazards model, which allows us to study the relationship between various features and the duration of ministerial spells without making any specific assumption about the shape of the hazard function.⁴

Multivariate Analysis

Tables 3 and 4 present our estimates of the effects of our covariates on foreign minister survival in office.⁵ We present hazard ratios, so coefficients above one represent increased hazard rates, meaning shorter tenures, while coefficients below one suggest decreased hazard rates, meaning longer tenures. Besides our main independent variables, we include several controls in our models, including the age, gender, and education of the foreign minister, whether they are themselves the HoG, whether they previously served in the post, whether they had any military experience, and whether the state is a democracy or not (Marshall, Jaggers, and Gurr 2009).⁶ Models 2, 4, and 6 use the sample limited to those ministers who experienced an international dispute. All models allow shared fragility by state.⁷

Table 3.

Cox proportional hazard models of foreign minister survival

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Previous term	1.228	1.293			1.305*	1.292
	(0.176)	(0.253)			(0.190)	(0.256)
Political experience (Hypothesis 1)	1.651*	1.224			2.356***	1.281
	(0.451)	(0.477)			(0.678)	(0.518)
Diplomatic experience (Hypothesis 2)	0.657***	0.534***			0.676***	0.535***
	(0.0923)	(0.112)			(0.0935)	(0.113)
Military experience	0.940	1.225			1.034	1.231
	(0.136)	(0.230)			(0.151)	(0.234)
Gender	1.199	1.374			1.131	1.053
	(0.562)	(1.013)			(0.528)	(0.782)
Age	1.013	1.020*			1.009	1.010
	(0.00786)	(0.0124)			(0.00765)	(0.0125)
Head of government	1.037	0.733			1.143	0.763
	(0.193)	(0.207)			(0.219)	(0.231)
Education	0.825***	0.853			0.822***	0.772**
	(0.0574)	(0.0870)			(0.0591)	(0.0819)
Level of democracy (Polity)	0.961***	0.961**	0.975**	0.986	0.967***	0.971*
	(0.0112)	(0.0170)	(0.00971)	(0.0143)	(0.0110)	(0.0168)
Average dispute level (Hypothesis 5)			1.271***	1.320***	1.326***	1.374***
			(0.0641)	(0.0695)	(0.0725)	(0.0797)
Any dispute			0.847		0.728
			(0.205)		(0.187)
Win dispute (Hypothesis 6)			0.429***	0.350***	0.491***	0.424***
			(0.0912)	(0.0771)	(0.109)	(0.0994)
Compromise dispute (Hypothesis 7)			0.676	0.577**	0.526**	0.489**
			(0.181)	(0.157)	(0.152)	(0.149)
Lose dispute (Hypothesis 8)			0.773	0.671*	0.769	0.659*
			(0.159)	(0.141)	(0.166)	(0.149)
Any dispute end			0.407***	0.339***	0.414***	0.330***
			(0.0974)	(0.0849)	(0.104)	(0.0878)
Theta	0.527	0.801	0.482	0.784	0.502	0.969
	(0.2128)	(0.375)	(0.195)	(0.375)	(0.207)	(0.440)
Observations	912	507	988	538	912	507

Notes: Exponentiated coefficients (hazard ratios); Standard errors in parentheses. ^*p < .1, ^**p < .05, ^***p < .01.

Theta is the estimated variance of the group level frailty.

Table 4.

Interactive model of performance and political institutions

	Model 7
Previous term	1.283*
	(0.188)
Political experience	2.296**
	(0.825)
Polity × political experience (Hypothesis 3)	0.989
	(0.0562)
Diplomatic experience	0.678***
	(0.0968)
Polity × diplomatic experience (Hypothesis 4)	0.978
	(0.0200)
Military experience	1.044
	(0.154)
Gender	1.146
	(0.537)
Age	1.008
	(0.00768)
Head of government	1.158
	(0.225)
Education	0.827***
	(0.0602)
Level of democracy (Polity)	0.981
	(0.0556)
Average dispute level	1.329***
	(0.0735)
Polity × average dispute level	0.998
	(0.00562)
Any dispute	0.738
	(0.191)
Win dispute	0.472***
	(0.110)
Polity × win dispute	1.026
	(0.0287)
Lose dispute	0.778
	(0.173)
Polity × lose dispute	1.013
	(0.0267)
Compromise dispute	0.527**
	(0.152)
Polity × compromise dispute	1.019
	(0.0364)
Any dispute end	0.397***
	(0.102)
Theta	0.502
	(0.208)
Observations	912

Notes: Exponentiated coefficients (hazard ratios); Standard errors in parentheses. ^*p < .1, ^**p < .05, ^***p < .01

Theta is the estimated variance of the group level frailty.

Models 5 and 6 provide the most comprehensive test of our hypotheses, on the full data and the 1-dispute-minimum subset of the data. We do not find evidence in favor of Hypothesis 1, that ministers with political experience will have a lower risk of being forced to resign. In fact, in the models that use the full sample, we find the opposite. Foreign ministers with political experience have a statistically significant but higher hazard rate. In the MID limited sample, the coefficient is in the same direction, but no longer statistically significant. We return below to a discussion of what might explain this result.

In contrast, we find robust support for our second hypothesis, that diplomatic experience decreases the likelihood of forced resignation. Across every model in which it is included, the coefficient is statistically significant. Hence, having diplomatic experience is clearly positive for a foreign minister's survival. This result suggests that individual ministers who are expected to perform well in office are less likely to lose their post through forced resignation.

We also find support for Hypothesis 5 as average hostility level is positively and significantly associated with higher hazard rates in Models 5 and 6, and all prior models. Hence, it does appear that the higher the level of conflict, the higher the risk that foreign ministers lose their posts. We also find statistically significant support for Hypothesis 6, which argues that a positive⁴ outcome in a dispute (e.g., a win) decreases the risk that foreign ministers lose their post. That finding is consistent across all models that include conflict variables. Hypothesis 7 is also supported: resolving a dispute through compromise is associated with a lower risk of removal from office. In contrast, we find no support for Hypothesis 8, that a loss in a dispute increases the foreign minister's risk of removal.

In light of the previously mentioned argument suggesting that some high-performing ministers may be seen as challengers to the leader, resulting in higher hazard rates for these ministers, particularly in autocracies, we also assess the extent to which political institutions influence our results. Table 4 presents a survival model taking a first step in testing this argument by interacting background features with political institutions, here with regime type as measured in Polity. We do so for our fully specified Model 5 in table 3. Estimation results for a Cox model with country-based shared fragility is presented in table 4.

The interactions need to be interpreted carefully because the marginal effects of our key variables require a linear combination of coefficients, which are available in our replication files. The interactions show, first, in line with Hypothesis 3, that the effect of competence varies across different types of background characteristics. For instance, political experience might indicate that a minister is a credible challenger to the incumbent and, therefore, it should increase a minister's hazard rate. This is precisely what we find in the linear combination of political experience and its interaction with Polity—it increases the hazard rates for small values of Polity, but then it increases the hazard at decreasing rates for larger values of Polity. Hence, ministers with political experience in autocracies face a high risk of deposition, but as the regime becomes more democratic, that risk is also high but not at the same level.

As expected (Hypothesis 4), we find exactly the opposite result with diplomatic experience, which suggests that ministers with diplomatic experience are not necessarily seen as credible challengers. Our test for the linear combination of coefficients indicates that diplomatic experience reduces the hazard rate in autocracies and the reduction is larger in democracies. In other words, foreign ministers with a diplomatic background have lower risks of deposition and this risk is even lower in democracies. Hence, hypotheses 3 and 4 are borne out in the results, showing that it is important to distinguish between autocratic and democratic regimes when analyzing the impact of performance-related features on foreign ministers’ tenure.

Concluding Remarks

We have in this paper asked whether the survival of foreign ministers depends on their performance in office. Analyzing original historical data on foreign ministers, we find several performance-related features that matter for foreign minister tenure. First, diplomatic experience decreases the likelihood of forced resignation. This result is particularly interesting since diplomatic experience should be important for foreign ministers as the “highest diplomats in government” (Quiroz Flores 2009, 118), and since no previous studies have shown that this type of experience may be important for tenure.

However, contrary to what we expected, foreign ministers with political backgrounds are more likely to be fired from their posts. Although this result is in line with the literature on what to expect in autocracies (e.g., Quiroz Flores and Smith 2011), there may be other possible explanations for why we find the same (but weaker) effect in democracies. It may be that foreign ministers with political backgrounds are ill-suited to the post, but are selected for reasons unrelated to their qualifications. If political background makes someone more likely to be appointed foreign minister, independent of whether they would be good at it, the result could be a population of ministers who are, on average, worse in ways our performance measures do not capture. It could also be the case that a political background is poor training for diplomacy, and having it actually makes the foreign minister worse at their job, again in ways that our performance measures do not capture.

Just as in some of the previous literature on leadership survival, we find that conflict matters for foreign minister tenure. First, we find support for the hypothesis that foreign ministers are significantly more at risk of losing their posts if the level of conflict is high during their time in office. This may be due to that HoGs “pin blame” on individual ministers and deflect criticism by removing the foreign minister when the country is involved in a conflict. We also find support for the idea that positive conflict outcomes, such as winning a conflict or ending it by reaching a compromise agreement, increases the foreign ministers’ tenure, which could suggest that “good performance” is rewarded by the HoG.

However, we do not find strong support in our data for the hypothesis that losing a militarized dispute influences the tenure of foreign ministers—hence, they do not seem to be “punished” for a loss. This is not fully in line with the results found in the literature on leadership survival (e.g., Debs and Goemans 2010). Here, part of the explanation of the differing results may of course lie in the fact that we are studying foreign ministers and not HoGs, and that the features influencing these politicians’ fate are slightly different.

Finally, it may be that, while foreign ministers get partial credit for policy decisions that lead to good outcomes, and for compromising solutions to disputes, they do not get blamed for lost militarized disputes. Foreign ministers might help make policy, but are rarely responsible for the performance of the military—that lies with defense ministers. Thus, it may be that defense ministers get blamed for the loss of militarized disputes, while foreign ministers get partial credit for foreign policy success. If so, that would explain why wins and compromised disputes decrease the likelihood of forced resignation, while dispute losses do not affect the likelihood that foreign ministers lose their posts. Whether foreign policy decision-making lies at the hands of foreign ministers or defense ministers may not only clearly vary across countries, but also vary over time. The United States is an important example of the latter, where the chain of command has changed on several occasions (see, e.g., Flynn 2014). Future research should address this possibility more directly, ideally by performing analyses similar to ours, focusing on whether the survival of defense ministers is influenced by conflict outcomes.

Even though we are here mainly interested in foreign ministers and the impact of foreign policy outcomes, we should also discuss how transferrable our results are to other policy domains and other types of ministers. Many of the arguments we make here should also be valid for other policy domains and other ministers. For example, we know from previous research that Prime Ministers are likely to use ministerial dismissals in a number of policy domains to “pin blame on individual ministers and deflect criticism and subsequent falls in popularity” (Dewan and Dowding 2005, 46). Some previous research has also analyzed the impact of economic features on ministerial survival, for example, showing that economic growth reduces the risk of minister turnover (Camerlo and Pérez-Liñan 2015). Hence, similar arguments as we make here about how foreign minister survival and the role of important “events” should be transferrable to other policy domains. However, the ability and incentives of a HoG to “scapegoat” individual ministers may clearly vary along a number of parameters, for example, depending on how important the events are and how much control both the HoG and the individual minister have over a specific policy area. Here, further research is needed, analyzing other types of ministers and policy domains, for example, focusing on whether finance ministers are “scapegoated” for economic downturns.

Another avenue for research is to explore the interaction between various features included in our model. For example, it may be the case that only individuals with a specific type of background are punished or rewarded for certain conflict outcomes. Are foreign ministers with a diplomat background more likely to get punished if they are not able to reach a compromise agreement? This may be expected if diplomat foreign ministers are seen as being skilled in negotiations and reaching such agreements—failure may then be more likely to be attributed to a minister who has a background within diplomacy and they should be more likely to be removed from their posts. Our data allow scholars to evaluate such conditional hypotheses.

Another type of explanation that our data allow scholars to explore is the role of “affinity” between leaders and foreign ministers. In this article, we have not added any information on the political leader who selects and deselects the individual foreign minister, but this is clearly possible to do by combining our data with data drawn from the LEAD dataset, which includes information on the background of individual HoGs, for example, on their educational background and military experience (Horowitz et al. 2015). One hypothesis that can be evaluated by doing so is whether politicians with similar backgrounds to their leaders are more likely to become foreign ministers, since leaders may expect that such ministers will have similar preferences to their own. Principal–agent theory would suggest that leaders can minimize agency loss by selecting such ministers as agents. For example, are leaders with military experience more likely to appoint foreign ministers with a military background?

By making our data public, we hope to stimulate future research along these lines, allowing scholars to compare the background of foreign ministers to that of their leaders, thereby providing important information that could settle arguments about who is likely to become the closest advisor to presidents, prime ministers, kings, and dictators.

Funding

Research for this article was made possible through funding from Riksbankens Jubileumsfond, grant no. M14-0087:1.

Supplementary Information

Supplementary information is available at the Foreign Policy Analysis data archive.

Footnotes

¹The exception is Prince of Hohenlohe.

²We have, however, erred on the side of inclusion when also including all foreign ministers (many of whom were also federal chancellors) in Austria from 1922 to 1959, although the Austrian Foreign Ministry during this time, for cost-reducing reasons, was formally only an “office” within the Federal Chancellery (Derndarsky 1982, 61, 68).

³As described in table 1, there are several different types of “forced” resignations, which we now group together and use to create our dependent variable. One type of “exit” which should be discussed here is the one where a foreign minister is forced to leave their post, but remains in cabinet. This is not as “severe” of a punishment as being completely removed from cabinet, but we believe that since most other cabinet posts are less prominent than the foreign minister post, these exits can be treated in a similar manner. As a robustness check, the online appendix 4 presents an extension of Model 5 in table 3, our preferred model, to a competing risks model of forced resignation, retirement/unforced resignation, violent exit, and end of government term. Generally, our substantive results remain unchanged in this model, see the online appendix 4.

⁴Several empirical analyses of leader and minister tenure in office use fully parametric survival models (e.g., Bueno de Mesquita et al. 2003; Quiroz Flores 2009). The online appendix 1 presents results for our specifications of table 3 using the same choice of distribution for duration time as Bueno de Mesquita et al. (2003) and Quiroz Flores (2009); our substantive results remain unchanged.

⁵As a robustness check, we also run these models broken down by time period: pre–, and post–World War I. Our substantive results remain unchanged.

⁶We use Polity to measure regime type and treat it as a control. The average level of Polity in our estimation sample is 2.5, but there is a 7.0 standard deviation around that mean. The online appendix 5 controls for economic conditions. Specifically, we extend Model 5 and control for the natural logarithm of real GDP per capita using data from the Maddison Project (Bolt et al. 2018). The results for our main hypotheses are substantively the same.

⁷Tests for non-proportionality using Schoenfeld residuals indicate that a covariate in Models 3 and 5 is not proportional. The online appendix 2 presents estimation results that correct for non-proportional covariates (e.g., Park and Hendry 2015). Our substantive results remain unchanged.

Notes

Hanna Bäck is a Professor of Political Science at Lund University. Her research mainly focuses on political parties, governments, and cabinet ministers in parliamentary democracies. She has published extensively on these topics, for example, in journals such as British Journal of Political Science, Comparative Political Studies, and European Journal of Political Research. She is the coauthor of Political Parties, Parliaments, and Legislative Speechmaking (Palgrave MacMillan, 2016).

Jan Teorell is a Professor of Political Science at Lund University. His research interests include political methodology, history, and comparative politics, comparative democratization, corruption, and state-making. His work has appeared in journals such as the American Political Science Review, Comparative Political Studies, and European Journal of Political Research. He is the author of Determinants of Democratization (Cambridge University Press, 2010) and coauthor of Varieties of Democracy (Cambridge University Press, 2020).

Alexander S. Von Hagen-Jamar is an independent scholar. He finished PhD at the University of Michigan in 2014 and has held positions at Carleton College, Lund University, and NUY Shanghai. His research interests include foreign policy, political attitudes, state capacity, and political methodology. His work, for example, has appeared in the Journal of Politics.

Alejandro Quiroz Flores is a Reader at the Department of Government and Chief Scientific Adviser at the Institute for Analytics and Data Science at the University of Essex. His research focuses on political economy and econometrics. His work has appeared at Global Environmental Change, British Journal of Political Science, and International Studies Quarterly, among others. He is the author of Ministerial Survival during Political and Cabinet Change: Foreign Affairs, Diplomacy and War (Routledge 2016).