Introduction
Donald Trump’s renewed candidacy for the US presidency, which he held from 2017 to 2021, has sparked public interest in the politics of non-consecutive second or even multiple terms as chief executive in representative democracies. While there has been only one other ‘comeback president’ in the history of the United States before Trump’s comeback in 2024—Grover Cleveland in the late nineteenth century—, the phenomenon is much more widespread in the presidential democracies of Latin America, with one or more instances in most countries.Footnote 1 The possibility of presidents returning to office after a while is greatly facilitated by the fact that most presidential democracies have constitutional term limits, precluding incumbents from immediately continuing their tenure after reaching a maximum number of consecutive terms, two or often just one (Baturo and Elgie, Reference Baturo and Elgie2019). Therefore, even the most popular and powerful leaders need to take a break before being re-elected and resuming the presidency.
But what about chief executives in parliamentary democracies? Do they tend to come back to office after a break as well? One might immediately think of some illustrious figures from different countries and eras who returned to the prime ministership once or even twice after an interim period, such as Winston Churchill (UK) in the 1950s, Gro Harlem Brundtland (Norway) in the 1980s, or Silvio Berlusconi (Italy) in the 2000s. But there are also parliamentary democracies that have never seen any ‘comeback prime minister’ (CBPM), such as Germany or Spain. Even this cursory glance suggests that the frequency of chief executive comebacks tends to be more variable in parliamentary democracies than in presidential democracies, which seems to make sense given the distinct structure of parliamentary systems: unlike their counterparts in presidential systems, prime ministers are not directly elected for a fixed termFootnote 2 , but can be dismissed any time by a parliamentary majority through a vote of no confidence (e.g. Strøm, Reference Strøm, Strøm, Müller and Bergman2003: 64). This institutional dependence of the chief executive on parliament explains not only why prime ministers face no constitutional term limits; the parliamentary responsibility of prime ministers also gives the political parties a decisive role in their selection and removal (Samuels and Shugart, Reference Samuels and Shugart2010).
The return, or non-return, of former incumbents to office depends indeed mainly on the decision of the party nominating a prime-ministerial candidate. However, the possible reasons for such decisions have been widely ignored in comparative political research and remained largely unknown. Apart from very rare case studies (e.g. Chin, Reference Chin2018), there is no systematic evidence on the frequency of CBPMs, let alone the particular circumstances of their reinstatement. Some recent studies deal with ‘ministerial comebacks’ (e.g. Pedrazzani and Vercesi, Reference Pedrazzani and Vercesi2022), but not prime-ministerial comebacks. Even the evolving literature on the ‘afterlife’ of prime ministers only considers other political (or non-political) positions taken after leaving office, but not their return to the premiership (Baturo, Reference Baturo2017; Theakston, Reference Theakston2012; Theakston and De Vries, Reference Theakston and De Vries2012).
This conspicuous gap in the literature is even more surprising as a more profound knowledge about prime-ministerial comebacks could contribute a lot to a better understanding of how parliamentary democracies work. Since democratic rule is always ‘government pro tempore’ (Linz, Reference Linz1998: 19), the very possibility of prime ministers returning to office after a while can be seen as a crucial indicator for the genuinely democratic nature of the regime, distinguishing it from various types of autocracy in which former power-holders are usually either locked-up or dead, if not hiding in exile. Beyond this general perspective, the occurrence of CBPMs could also have different effects on the functioning of parliamentary democracy, depending on which dimension of the key credentials of ‘good governance’ is being emphasized. If the virtue of parliamentary democracy is mainly seen in its ability to allow for genuine alternation of power and renewal in terms of both policy and personnel, then the return of former prime ministers to the top job may seem to betray the promise of ‘government pro tempore’, and the fair chance for different actors to rise to the top. Frustration with this is likely to grow to the extent that the parties concerned do not allow much participation in the selection of prime-ministerial candidates, and when the governments eventually formed appear to be only loosely related to the electoral performance of the parties involved. Alternatively, if parliamentary democracy is specifically valued for its inherent emphasis on selecting the ‘best qualified’ people for leadership positions, former prime ministers returning to power may seem a precious commodity, as they have unique first-hand experience in the top post, which may prove an invaluable resource when it comes to performing successfully.Footnote 3
This paper presents the first comparative study of CBPMs, exploring their occurrence in 18 established democracies from 1945 to 2024. For this purpose, we refer to parliamentary democracies as regimes in which the prime minister is the effective political chief executive. ‘Comeback prime ministers’ are those actors who regain the premiership after losing it for whatever reason, be it an election defeat of their party or government, a successful vote of no confidence, a coalition break-up, or sickness.Footnote 4
As noted above, a characteristic feature of parliamentary democracies is that political parties are the main actors in the selection and removal of prime ministers, and thus determine the politics of their possible reinstatement. This may be different in a particular type of semi-presidentialism, which not only meets the two standard features of this regime type, that is, parliamentary responsibility of the prime minister and direct election of the head of state (Elgie, Reference Elgie1999), but in which the latter also tends to be the unchallenged political chief executive with wide-ranging powers in the process of government formation and termination (i.e. the head of state operating as ‘executive president’; Elgie, Reference Elgie and Magone2015: 326). In those regimes, of which the Fifth French Republic is widely seen as the prototype, the president is usually a major player in the process of reinstating former prime ministers. However, this does not necessarily mean that political parties are marginalized; in particular when there is split party control of parliament and the presidency, the majority parties in the legislature may be the dominant actor (see note 8). More generally, constitutional rules as such can be a poor guide to presidential power in constitutional practice. Take Austria, where the popularly elected presidents have usually not made use of their considerable powers in government formation and beyond, leaving issues of political leadership, including the selection of the chancellor, to the main parties (Helms and Umek, Reference Helms, Umek, Praprotnik and Perlot2023).Footnote 5 Therefore, our sample also includes some semi-presidential regimes with ‘executive presidents’ alongside ‘classic’ parliamentary democracies to explore if presidential powers systematically matter for the occurrence of CBPMs.
In the next section, we not only elaborate on how the general conditions for a return of chief executive differ between parliamentary and presidential democracies but also look into the specific contextual features that may motivate parties in parliamentary democracies to reinstate a former prime minister in the top executive position. As a result, we identify three ideal-typical settings in which CBPMs are particularly likely to occur: replacements of ‘failed’ prime ministers in contexts marked by frequent replacements of prime ministers, post-electoral comebacks in contexts where prime-ministerial parties alternate at short intervals, and prime-ministerial returns from ‘presidentialized’ parties with a high internal power concentration. The following section provides a systematic review of CBPMs in 18 established democracies since 1945; it reveals that, while prime-ministerial comebacks are remarkably common overall, their occurrence also varies considerably between countries and over time. We then show how the concrete contextual settings help to make sense of these patterns. In general, established democracies witnessing frequent occurrences of CBPMs fall into two groups: those with high cabinet instability specifically warranting prime-ministerial replacements by experienced leaders, and those with alternating government parties that keep their former prime ministers in a ‘waiting position’ as opposition leaders. However, the overall decline in CBPMs since the 1990s may be best understood as an ambiguous consequence of the ongoing ‘presidentialization of politics’ (Poguntke and Webb, Reference Poguntke and Webb2005b). On the one hand, presidentalization implies greater electoral personalization. Most prime-ministerial parties tend to choose ‘fresh faces’ rather than former incumbents to regain the top executive position. On the other hand, presidentialization fosters the emergence of personalized parties with a high concentration of intraparty power. Some of these parties have also repeatedly won parliamentary elections, which allowed their (internally uncontested) leader to return to the premiership. The conclusion discusses the main implications of these findings and identifies several avenues for future research.
Reasons for prime-ministerial comebacks
Becoming chief executive is undoubtedly the pinnacle of a political career in representative democracies (Müller-Rommel et al., Reference Müller-Rommel, Kroeber, Vercesi, Andeweg, Elgie, Helms, Kaarbo and Müller-Rommel2020). Most incumbents will try to stay in office as long as possible, and many of them may even want to return if being forced to leave early, provided they still have political ambitions and are in reasonably good health. However, there are various mechanisms of ‘time-related control’ over incumbents, and would-be incumbents aspiring to return to office, which are characteristically different in presidential and parliamentary democracies.
In presidential democracies, chief executives are directly accountable to the voters through their popular mandate, but not to legislative assemblies, which have no right to remove them for political reasons.Footnote 6 Therefore, political parties also have very little leverage to control incumbent presidents. Indeed, they even have limited influence on the selection of presidential candidates. Primaries, which have been applied in several presidential democracies (Kemahlioglu et al., Reference Kemahlioglu, Weitz-Shapiro and Hirano2009), make the outcome of intraparty selection processes highly uncertain, with party functionaries having little to no advantage over outsiders of any kind. Once in power, most presidents tend to strongly dominate their parties, a phenomenon Samuels and Shugart (Reference Samuels and Shugart2010) refer to as a ‘presidentialized’ variant of intraparty politics.
Since presidents cannot be effectively held accountable by either parliament or their parties and may even aggrandize their power beyond the established institutional checks and balances, most presidential constitutions provide for term limits precluding immediate re-election of incumbents after one or two consecutive terms (Baturo and Elgie, Reference Baturo and Elgie2019; Heyl and Llanos, Reference Heyl and Llanos2022). Apart from some cases that rule out another attempt to win the presidency after reaching the maximum number of terms, most constitutional term limits do not imply a definite farewell for presidents, but rather a ‘cooling off’ period that allows for a political reset at the next election, giving candidates a distinct opportunity to demonstrate their ability to regain voter support in the absence of an incumbency advantage.
Overall, it is the combination of two characteristic features of presidential democracies that plausibly explains the rather high number of ‘comeback presidents’ in Latin America and elsewhere. First, direct election to the presidency favors the rise and success of charismatic candidates who can largely bypass the party establishment on their way to the top. Second, constitutional term limits ensure that even the most popular incumbents must take a ‘cooling off’ period before running again and possibly returning to office.
The institutional structures of parliamentary democracies establish fundamentally different mechanisms of time-related control over the chief executive. Crucially, prime ministers are not directly accountable to the electorate but rather to parliament. However, prime ministers are not only agents of parliament but also of their respective party. Samuels and Shugart call these parties ‘parliamentarized’ because they hold their prime ministers accountable ‘through an internal deselection process’ similarly as parliament holds them accountable ‘through the confidence procedure’ (Samuels and Shugart, Reference Samuels and Shugart2010: 16).
This mechanism of intraparty accountability is most straightforward in the Westminster-type democracies, such as the United Kingdom, Australia, and Canada, where incumbents hold the premiership in their role as leader of the largest parliamentary party (Rhodes et al., Reference Rhodes, Wanna and Weller2009). Since the latter marks the prerequisite for the former, internal leadership contests in a party that controls the premiership are effectively contests to win the premiership, while losing the party leadership entails losing it. However, even in parliamentary democracies outside the Westminster hemisphere, where holding the party leadership is not a constitutional requirement for becoming prime minister, and where the parties’ top candidates are not necessarily party leaders, such as in Germany (Helms, Reference Helms2020b), it is the parties that control the staffing of the chief executive position. As outlined in the introduction, this is to some extent even true in semi-presidential regimes, where directly elected presidents have formal or informal control over prime-ministerial selection and removal (Elgie, Reference Elgie1999, Reference Elgie and Magone2015). While such ‘dual-executive hybrids’ (Samuels and Shugart, Reference Samuels and Shugart2010: 98-99) complicate the accountability relationships, they only mitigate, but not eliminate, instances of intraparty scrutiny of the prime minister.
Overall, time-related control of chief executives in parliamentary democracies is both more continuous and more immediate than in presidential systems. As a result, there is no apparent need to impose constitutional term limits on prime ministers. At the same time, their tenure is much more variable than that of their presidential counterparts. Parliamentarized parties tend to stick with their prime ministers as long as they are politically successful. If ineffective or highly unpopular prime ministers become an electoral liability, their parties are likely to replace them even before the end of the term. By contrast, prime ministers considered able to help their party to retain governmental power beyond the next election are likely to stay on. As a result, some prime ministers may serve much longer than their presidential counterparts, precisely because of the conspicuous absence of any formal term limits for chief executives in parliamentary systems. Typical examples are Helmut Kohl and Angela Merkel, who each served as German chancellor for 16 years. With other chancellors also serving two or more consecutive terms, Germany has seen very rare alternations in the chief executive office since 1949 as compared to other parliamentary democracies (Grotz and Schroeder, Reference Grotz and Schroeder2023: 368-369).
Irrespective of the length of tenure, the relationship between parliamentarized parties and their executive agents seem to suggest that prime ministers complete their mandates without interruption. When incumbents leave the top job after losing a parliamentary election, through early replacement or even a voluntary resignation, this usually indicates a loss of political authority, appeal, and support, which is likely to disqualify them in the eyes of their party for a later comeback. At the same time, former prime ministers have also gained unique leadership experience in the top executive position, which may make them particularly attractive candidates for their party when it comes to filling the premiership again. Given these competing rationales, a former prime minister’s comeback can be considered likely especially in contexts where the reinstatement of an ‘experienced incumbent’ seems, overall, more beneficial to the party than avoiding the public perception of a ‘loser’s return’.
To systematically identify such contextual settings, we first look at the two ideal-typical situations in which prime ministers take office (Grotz and Weber, Reference Grotz, Weber, Harfst, Kubbe and Poguntke2017): (1) In ‘post-electoral’ situations, with the entire parliamentary term still ahead, prime ministers ‘have considerable time to put through an ambitious policy agenda and “earn” the hoped-for rewards’ (ibid.: 235). Hence, they are primarily expected to deliver on their party’s manifesto in order to maximize its re-election chances. (2) In ‘replacement’ situations, a prime minister takes office during the election period when the previous incumbent had to leave office. This poses particular challenges for the new leader, as the remainder of the term may be too short for implementing a coherent policy agenda, and may also be overshadowed by the failures of the outgoing cabinet. As a result, incoming prime ministers can be expected above all ‘to “bring the government home” without a lot of fuss’, to avoid political damage to their party before the next election (ibid.: 235-236).
Considering these different contexts of cabinet formation, former prime ministers seem to be a reasonable choice for their parties particularly in replacement situations, as their leadership experience makes them perfectly qualified to meet the difficult challenges of restabilizing the government. Thus, we would expect CBPMs to be most common in parliamentary democracies with high levels of cabinet instability, where frequent replacements seem to require experienced leaders. A typical example of the latter context is the French Fourth Republic (1946–1958) with its notoriously unstable cabinets and a high number of prime ministers. In contrast, Spain has seen very few changes in the prime minister’s office since the early 1980s (see Table 1).
Table 1. Contextual conditions and frequency of CBPMs
By contrast, prime-ministerial comebacks in post-electoral situations seem realistic only under specific circumstances. Plausibly, a former prime minister will be the ‘natural candidate’ for the top job after the party’s election victory if that party was in opposition before and the former prime minister occupies a prominent intraparty position, such as party leader or parliamentary group leader. In this scenario, the party is likely to allow former prime ministers to return to office under two conditions. First, prime-ministerial parties in that system do not usually remain in power for several consecutive terms but alternate at shorter intervals. In this case, parties may be much more willing to stick with the previous incumbent as their leader in opposition to have a ‘tried and tested’ but still reasonably fresh top candidate available for the upcoming election. Second, electoral competition should not be too strongly ‘personalized’. If the personal characteristics of top candidates are crucial for the parties’ electoral success, former prime ministers would be a particularly ‘risky target’, as they are likely to be blamed for past policy mistakes, thus limiting their parties’ chances to regain power. A case in point is Norway, which has seen several changes in the prime minister’s office during legislative periods and an overall limited degree of ‘presidentialization’ of politics (Kolltveit, Reference Kolltveit2012); an opposite case is Germany with extended terms of chief executives and a higher degree of personalization, which creates a preference for ‘fresh faces’ when the pendulum eventually swings (see above).
Beyond these different contexts of government formation, the probability of prime-ministerial returns may also be influenced by the intraorganizational context of the respective party. Specifically, if intraparty power and decision-making authority is strongly concentrated on its (often charismatic and popular) leader, he or she is likely to remain at the top of the party even after losing the premiership and will also be the party’s undisputed choice for the next opportunity to fill that position. In other words, prime-ministerial returns are more likely in parties with ‘presidentialized features’ (Samuels and Shugart Reference Samuels and Shugart2010, 16) than in parliamentarized parties, which have less concentrated decision-making structures and are not critically dependent on their leader’s popularity. An example for a party with a high degree of internal power concentration is Forza Italia, built around its founder Silvio Berlusconi, while the Social Democrats in Sweden may represent the latter, decentralized type of party (see Table 1).
The extent to which presidentialized parties occupy the premiership in parliamentary democracies remains a matter of debate. While Samuels and Shugart (Reference Samuels and Shugart2010: 17) argue that such parties are ‘at odds with the core logic of parliamentarism and [… are therefore] likely [to be] the exception rather than the rule or a trend across all systems’, Poguntke and Webb (Reference Poguntke, Webb, Poguntke and Webb2005a) contend that politics in parliamentary democracies has overall become more ‘presidentialized’ in recent decades, as various structural changes—from the erosion of traditional social cleavages, an internationalization of politics to increased mediatization, and others—have combined to expand the power of prime ministers in the electoral arena, within their parties, and across the executive. However, the chapters from Poguntke and Webb (Reference Poguntke and Webb2005b) do not provide full and up-to-date systematic evidence for their ‘presidentialization of politics’ thesis (see also Elgie and Passarelli, Reference Elgie, Passarelli, Andeweg, Helms, Elgie, Kaarbo and Müller-Rommel2020). While comparative research has shown that the effects of personalized leadership on electoral behavior have increased significantly in the recent past (Garzia et al., Reference Garzia, Ferreira da Silva and De Angelis2021), it is not clear yet whether there is a similarly consistent personalization within prime-ministerial parties.
Indeed, the suggested ‘presidentialization of politics’ could well have highly ambivalent and divergent effects on the occurrence of CBPMs. On the one hand, leader-centered parties may have become more common and electorally stronger, making them significantly more likely to nominate the prime minister after a general election. As leaders of presidentialized parties can even regain the premiership if they have previously lost it, CBPMs from these parties would become more common. On the other hand, as presidentialization also involves increasing levels of personalization of parliamentary elections, the advanced personal exposure of top candidates resulting from this may ultimately increase their political vulnerability (Helms, Reference Helms2012: 66; Poguntke and Webb Reference Poguntke, Webb, Cross, Katz and Pruysers2018). From this perspective, ‘fresh faces’ may seem more than ever a reasonable choice for parliamentarized parties seeking to regain control of the premiership, which implies that, all else equal, CBPMs will become less frequent under conditions of electoral presidentialization.
Table 1 summarizes the contextual conditions, under which CBPMs are likely to occur. These include two cabinet formation contexts that may be typically found in different parliamentary democracies: first, contexts of high cabinet instability, where former prime ministers are possibly sought after as particularly qualified replacements for prematurely ousted incumbents; second, post-electoral contexts characterized by shorter intervals between alternations of prime-ministerial parties and low electoral personalization. Furthermore, the emergence of prime-ministerial parties with a high personal power concentration may facilitate prime-ministerial comebacks of their leaders. While the likelihood of CBPMs from such presidentialized parties has plausibly increased in the recent past, the simultaneously higher level of electoral personalization may make the return of a former prime minister from parliamentarized parties a less frequent phenomenon.
Patterns of comeback prime ministers in established democracies
With these theoretical considerations in mind, we now proceed to look at CBPMs in established parliamentary democracies, including Western Europe, Canada, Australia, and New Zealand. We begin with prime ministers who were in office from 1945 or later and consider only those who had left office by 30 June 2024. Caretaker governments are omitted both for CBPMs as well as for the other prime ministers serving as reference category. Countries with a population of less than one million (i.e. Iceland, Luxembourg, Malta, and Liechtenstein) were excluded, because their elite structures and recruitment patterns may differ significantly from those of larger countries (for example, due to the smaller pool of would-be candidates and a different political culture), undermining attempts to make first generalizable judgments. For similar reasons, the post-communist democracies of Central and Eastern Europe were also excluded, given the idiosyncrasies in their party organization and elite recruitment (from considerably ‘leaner’ party organizations to the much greater presence of political entrepreneurs). The period covered for Greece, Portugal and Spain is limited to the democratic chapters of their postwar histories, that is, since the mid-1970s. As mentioned above, the sample also includes semi-presidential regimes with ‘executive presidents’, that is, France (since 1958) and Finland (until 1999).Footnote 7 As these two countries have experienced periods of both semi-presidential and ‘classic’ parliamentary democracy, they offer a unique opportunity to examine if presidential powers make a difference to the occurrence of CBPMs. As a result, the sample comprises a total of 311 prime ministers from 18 democracies, which seems sufficiently large for the present purpose of a first exploratory study of prime-ministerial returns based on the assumptions outlined above.
Table 2 presents the CBPMs in relation to all prime ministers from the 18 countries covered in this study. To facilitate a first assessment of their political significance, the table also shows whether they were party leaders during their tenure, which indicates their intraparty power status (Dowding Reference Dowding, Strangio, Hart and Walter2013), as well as their total tenure in the chief executive office, which can be considered a proxy for ‘prime-ministerial strength’ (Müller and Philipp, Reference Müller, Philipp, Blondel and Thiébault1991; Baylis, Reference Baylis2007).
Table 2 Comeback prime ministers in 18 parliamentary democracies (1945–2024)
Sources: Authors’ calculations based on Döring et al. (Reference Döring, Huber and Manow2024) and Müller-Rommel et al. (Reference Müller-Rommel, Vercesi and Berz2022).
Annotations: Figures in parentheses refer to those prime ministers that (re-)entered office since 1990. Data for Greece, Portugal and Spain start with their democratization since the mid-1970s. Total numbers refer to all CBPMs/PMs excluding caretakers. Duration (in days) refers to the total tenure of the respective prime ministers, considering only completed terms until 30 June 2024.
Overall, the data reveal several notable findings. Most basically, around one in five prime ministers since 1945 returned to office after a break (19.2 percent), which qualifies the potentially widespread assumption that there is little to no room for CBPMs in established democracies. Moreover, 55 percent of them were party leaders during their time in office, and their average tenure was significantly longer than that of other prime ministers who led either one or several consecutive governments (2,242 days vs. 1,398 days). As a result, most CBPMs in established democracies do not appear to have been marginal figures but rather political heavyweights in the eyes of their respective parties. Consistent with this, there is only one woman among them, Gro Harlem Brundtland of Norway, confirming the more general finding that women are rarely selected for the premiership and other senior cabinet positions (Müller-Rommel and Vercesi, Reference Müller-Rommel and Vercesi2017; Kroeber and Hüffelmann, Reference Kroeber and Hüffelmann2022). This is generally because men selectors tend to favor ‘like-minded’ and ‘trusted’ individuals with gender-specific social capital for the most powerful and prestigious political offices (Verge and Claveria Reference Verge, Claveria and Waylan2017). Moreover, the almost complete absence of women CBPMs in our sample may be explained by two fundamentally different, indeed outright opposite, characteristics of first-time women premierships: While some women prime ministers had strikingly short tenures with unhappy endings (e.g. Kim Campbell, CA; Édith Cresson, FR; Lizz Truss, UK), effectively discrediting them for a comeback, others (such as Margaret Thatcher, UK; Angela Merkel, GER; Helen Clark, NZ) had exceptionally long uninterrupted tenures so that their eventual departure was generally perceived as a final farewell.
Beyond these general patterns, the occurrence of CBPMs varies considerably over time and space. Most notably, their share among all prime ministers has fallen sharply since 1990, to 9.6 percent. Since then, Italy and Norway have been the only countries from our sample with more than two prime-ministerial comebacks. There is also considerable variation across countries over the entire 1945–2024 period. While eight countries had between three and nine ‘returnees’ among their prime ministers (Denmark, Belgium, Finland, France, Ireland, Italy, Norway, and Sweden), in another eight countries CBPMs remained the exception (Australia, Austria, Canada, Greece, the Netherlands, New Zealand, Portugal, United Kingdom). They are completely missing in Germany and Spain, which is very much in line with our assumption about the strong correlation between long incumbency of prime-ministerial parties and the likely absence of CBPMs. In fact, supposedly helped by the ‘constructive vote of no-confidence’, most heads of government in Germany and Spain served between two and four consecutive terms (Müller-Rommel et al., Reference Müller-Rommel, Vercesi and Berz2022; see Table 2).
Apart from the frequency of their occurrence, CBPMs in the various parliamentary democracies also differ in terms of their political importance or power status, measured by his or her control of the party leadership. In this regard, the countries with three or more CBPMs fall into two groups. In Denmark, Ireland, Norway, and Sweden, almost all of them were party leaders before retaking office, indicating their crucial role in picking the party’s candidate for the top government’s position. In Belgium, France, Finland, and Italy, however, the share of party leaders among the CBPMs was rather low, suggesting that they did not play a dominant role in the intraparty candidate selection, or were selected by other actors. Specifically, the semi-presidential regimes of Finland and France have seen rather few CBPMs who were party leaders, which might point to the powerful role of their ‘executive presidents’ in cabinet formation.Footnote 8 The low share of powerful party leaders in Italy’s First Republic was very much a reflection of the notoriously high level of intraparty factionalism within the predominant Democrazia Cristiana (Leonardi and Wertman, Reference Leonardi, Wertman, Leonard and Wertman1989).
The average tenure of CBPMs is not only significantly longer than that of other incumbents for the overall sample, but also for most countries. This is partly due to the fact that no less than 13 prime ministers returned to office twice or more.Footnote 9 Even in Belgium and Italy, which have had very low levels of cabinet stability, CBPMs have served longer than other Belgian or Italian prime ministers. The most notable exception to this pattern is France, whose CBPMs—the large majority of which being creatures of the infamously unstable Fourth rather than the Fifth Republic—served for even shorter periods than their ‘non-returning’ colleagues.Footnote 10
The overall pattern of CBPMs in established parliamentary democracies thus reveals two main variations. First, there is a clear-cut cross-national variation with two groups of countries hosting multiple CBPMs that differ in terms of average tenure. More than half of all CBPMs (31 out of 60) are found in four countries with the shortest average tenure of prime ministers, providing initial evidence that our argument about executive-experienced leaders as candidates for replacement holds. At the same time, in countries with more stable cabinets and longer average prime-ministerial tenures, there are quite a few cases of long-serving CBPMs, most of whom have also been party leaders. This suggests that the scenario of CBPMs’ takeover in post-electoral situations in systems with alternating prime-ministerial parties could hold as well. Second, there is a pronounced temporal variation, namely a significant decline of relevant cases after 1990. This suggests that the ‘presidentialization of politics’ has no sweeping ‘boost effect’ on CBPMs but rather the opposite, while it is not clear whether presidentialized parties are actually more prevalent among the recent cases of prime-ministerial returns. We will therefore take a closer look at the cabinet formation contexts in which prime ministers return to office, before turning to the personal power concentration of the CBPMs’ parties as an indicator of their internal ‘presidentialization’.
The role of cabinet formation contexts and intraparty power concentration
Table 3 shows the two contexts of cabinet formation—post-electoral and replacement—of the CBPMs for their first and last uninterrupted terms. The majority of them (60 percent) returned to office in a post-electoral situation, suggesting that they had a central role within their party. Interestingly, exactly half of all CBPMs in this category also won office after a general election in the first place, while the other half initially became prime minister as a replacement. In other words, what might be seen as a less prestigious start of a prime-ministerial career did not undermine a candidate’s chances of a later comeback after a general election victory of his or her party. That said, winning the office in the first place after a general election carries the greatest chances to win the office back again after a later general election; only about 22 percent of first-time post-election prime ministers ended up as a replacement CBPM.
Table 3. Comeback prime ministers and cabinet formation contexts (1945–2024)
Source: Authors’ calculations based on Döring et al. (Reference Döring, Huber and Manow2024).
* In the case of PMs with multiple returns, the comeback period is classified according to the last relevant tenure.
Of the CBPMs who started their prime-ministerial careers as replacement candidates, around half returned in the wake of another replacement succession, while the other half made a comeback in the aftermath of a general election. In other words, some prime ministers were able to use the ‘side-entrance’ to the premiership to build up sufficient political capital to later become their party’s top candidate and eventually regain office. In some cases, this was helped by the short first stints as prime minister, with their ‘real premiership’ stretching over several consecutive terms beginning only after their comeback. This is true in particular for Menzies (Australia), Holyoake (New Zealand) and Stoltenberg (Norway), but also for Jorgensen (Denmark). While many replacement CBPMs had a more balanced pattern of first-time and comeback tenures, a few had a considerably more extended tenure as first-time prime minister. The showcase among the latter is clearly Trudeau (Canada).
Looking at the larger picture from the perspective of replacement CBPMs, the first thing to note is that more than three-quarters of them also started their prime-ministerial careers as replacements, which does however not necessarily say much about their leadership qualities (or lack thereof). More importantly, the lion’s share among the CBPMs who returned in a replacement situation—18 out of 24 (75 percent)—occurred in the four countries with the lowest average tenure of PMs (Belgium, Finland, France, and Italy).
The marked relevance of replacement situations for selecting CBPMs in the latter group of countries is also illustrated in Figure 1. It shows that Belgium, Finland, France, and Italy not only have the highest frequency of replacement situations per observation year but also a correspondingly high frequency of CBPMs. At the same time, frequent replacements are not a necessary condition for the occurrence of CBPMs. Rather, the countries with relatively ‘stable’ PMs during the parliamentary term fall into two groups with a notably high or low reoccurrence score. This once again confirms the assumption that post-electoral situations in systems with regularly altering prime-ministerial parties provide a distinct context conducive to frequent prime-ministerial returns.
Figure 1. PM replacements and CBPMs in established democracies (1945–2024).
Source: Authors’ calculations based on Döring et al. (Reference Döring, Huber and Manow2024). X-axis shows average annual number of replacements for the countries’ respective investigation period, y-axis shows share of CBPMs in all prime ministers per country.
Another crucial difference between post-election and replacement CBPMs concerns the length of their tenures. With few exceptions—especially Norway, featuring some highly esteemed and tenacious comeback leaders (Brundtland and Stoltenberg)—the individual tenures of replacement CBPMs were much shorter than those of their counterparts returning after a general election. Arguably even more important, most prime ministers returning in a post-electoral situation had a period of three to four years between their tenures, suggesting that the ‘alternation model’ applied to them. Indeed, in quite a few countries where some incumbents lost parliamentary elections, they retained their party leadership in opposition and returned to power after the next election. This applies mainly to countries with the institutional characteristic of an official Leader of the Opposition (UK, Canada, New Zealand, and Ireland; see Helms Reference Helms2024) or, alternatively, an established tradition of minority government (Denmark, Norway, and Sweden; see Rasch Reference Rasch, Persson and Wiberg2011), in which opposition leaders tend to be powerful players as governments depend on the parliamentary support of (parts of) the opposition. Prominent cases include Churchill and Wilson (UK), Trudeau (Canada), Holyoake (New Zealand), de Valera, Costello, FitzGerald, Haughey, and Lynch (Ireland), and—from the other side of the aisle—Brundtland and Stoltenberg (Norway), Hedtoft, Krag, and Jørgensen (Denmark) as well as Palme, Carlsson, and Fälldin (Sweden). Finally, the exploration of cabinet formation contexts also uncovers notable time-related features. Importantly, most of the few post-1990 CBPMs were deliberately chosen by their parties as top candidates and re-entered office after general elections (Kurz, AT; Rudd, AU; Rasmussen, DK; Berlusconi and Prodi, IT; Bondevik and Stoltenberg, NO).
The prominence of post-electoral returns in the post-1990 period may also have to do with altered intraparty power constellations in the era of presidentialization. Specifically, following our argument, we suppose that the prime-ministerial parties of this second period had a higher degree of internal ‘presidentialization’ in terms of leader-centeredness than the parties of CBPMs from previous decades. To establish the suggested party-nexus empirically, we look into the internal power concentration of those parties that reinstated a former PM before and after 1990, using the new V-Party dataset providing relevant data since 1970 (Düpont et al., Reference Düpont, Berker Kavasoglu, Lührmann and John Reuter2022). A particularly appropriate indicator to capture the central aspect of ‘party presidentialization’ or ‘personal power concentration’ in our context is the index of party personalization, which measures the extent to which the party is ‘a vehicle for the personal will and priorities of one individual leader’ (ibid.: 2). For each CBPM, we take the average of their party’s personalization scores for the final year of the first period and the first year of the second period and classify the pre- and post-1990 cases in three categories (strong/medium/weak) based on a 0–4 scale.
Table 4 shows these patterns, which overall confirm our expectations. Most prime-ministerial comebacks before 1990 were launched by parties with a low level of internal personalization. Remarkably, this equally applies to post-electoral/alternation and replacement/unstable cabinet contexts. In the former group, there were several countries where the incumbents lost parliamentary elections, retained their party leadership in opposition and eventually returned after the next election (see above). Intriguingly, these prime ministers were affiliated to parties from both sides of the ideological spectrum, such as the Social Democrats in Denmark (Hedtoft, Krag, Jørgensen) and in Sweden (Palme, Carlsson) on the left, and Fianna Faíl (de Valera, Haughey) in Ireland and the Centre Party (Fälldin) in Sweden on the right. In the latter group, Italy is arguably the most prominent case. Until 1990, all Italian CBPMs came from the Democrazia Cristiana, which governed almost continuously from 1945 to 1993. Since DC-led cabinets tended to be notoriously unstable, the party repeatedly relied on experienced PMs to lead a new cabinet, with Amintore Fanfani performing no less than four comebacks. The only clear exception to this pattern is Pierre Trudeau (CA) with his strongly personalized Liberal Party in the 1970s and 1980s.
Table 4. Intraparty power concentration and comeback prime ministers (1970–2024)
Sources: Authors’ calculations based on Döring et al. (Reference Döring, Huber and Manow2024) and Düpont et al. (Reference Düpont, Berker Kavasoglu, Lührmann and John Reuter2022).
Among the prime-ministerial parties after 1990, the share of ‘strongly personalized’ organizations is much higher and includes such different cases as Forza Italia with Silvio Berlusconi, who managed to re-enter the PM office twice after periods in opposition, Andreas Papandreou with his PASOK and, more recently, Sebastian Kurz with the Austrian People’s Party (ÖVP). Kurz’ ability to nearly completely reshape the traditional Christian democratic ‘catch-all’ party to his own liking, including wide-ranging organizational reforms specifically designed to enhance the power of the party leader, could be seen as a paradigmatic case of a ‘hyper-presidentialized party’ of the new age, and not just by Austrian standards (see, e.g. Puller, Reference Puller2018: 13–18).
However, as the Scandinavian experience suggests, advanced levels of intraparty personalization are not a sine qua non of CBPMs. A few prime ministers from relatively decentralized parties also returned in the post-1990 period, as to be observed in Denmark and Norway. Importantly, it would seem, however, that not only the paths back to office have been very different in Berlusconi’s Italy, Papandreou’s Greece and Kurz’ Austria on the one hand, and in the traditionally more consensus-oriented political systems of Scandinavia on the other; the personalities of the CBPMs from both groups are strikingly different as well. That the Nordic countries have continued to witness CBPMs after 1990 does not indicate that they have turned into hyper-personalized democracies. Rather, CBPMs have been part and parcel of a distinct brand of Scandinavian parliamentary democracy, whose most prominent feature has long been the prevalence of minority government (e.g. Rasch, Reference Rasch, Persson and Wiberg2011; Field and Martin, Reference Field and Martin2022). For the benefit of avoiding an understanding of individual countries that stresses their alleged uniqueness, it seems more appropriate to consider Norway and Italy—the only countries from our sample featuring more than one post-1990 CBPM—as the exemplary cases of two fundamentally different political contexts that have facilitated the emergence of CBPMs for opposite reasons: While Italy’s early Second Republic was marked by a unique combination of a power vacuum left by deeply discredited parties and record levels of personalization (Musella, Reference Musella2020), Norway has continued its long-standing tradition of picking leaders and leadership teams with the broadest possible experience in party and parliament (Kolltveit, Reference Kolltveit2012).
Conclusion
Prime-ministerial comebacks are much more common in established parliamentary democracies than might be expected in view of the conspicuous ignorance of this peculiar feature of leadership succession. With nearly one-fifth of all prime ministers from the 18 countries under study, the overall share of CBPMs since 1945 is substantial. At the same time, there is a strong cross-national variation, with strongholds and deserts easily identified, as well as a significant temporal variation, with the overall number of CBPMs declining significantly after 1990. Beyond these distinctive patterns, it is possible to identify a certain profile of CBPMs in the democracies under study. The typical CBPM, if there was one, is much more likely to hold an active status as party leader than not, regains the premiership after a general election, and tends to have a considerably longer tenure than ‘normal’ prime ministers. Most of these features are true both for the larger group of pre-1990 CBPMs and their colleagues of the post-1990 period. However, one important difference relates to the status within their party. While most pre-1990 CBPMs came from lowly personalized parties, some of their very few recent successors have been strongly personalized (party) leaders.
A key message of this article is that context-related rationales of the parties when selecting their prime-ministerial candidates demonstrably matter a lot. Specifically, we were able to uncover different patterns of prime-ministerial returns depending on systemic and intraparty features: replacements in contexts of high cabinet instability, post-electoral returns in contexts of alternating governments, and in parties with high levels of internal personalization (‘presidentialized’ parties). As the French and Finnish experience suggests, even presidents from constitutionally similar regimes may act very differently with regard to prime-ministerial (re)selection, which—in addition to the well-known effects of patterns of party control—may reflect different leadership cultures or even particular personal preferences of a president.
There obviously may have been additional, difficult-to-nail-down factors at work favoring, in particular, the documented decrease of CBPMs over time. Reflecting the sweeping acceleration dynamics in politics and the public sphere more generally, and the withering of a ‘deference culture’, contemporary citizens, media, and parties may simply expect leaders losing an election to make space for someone new, while previous generations of citizens, journalists and party officials considered it much more normal for accomplished leaders to continue if they wanted to. Very much in line with these tentative assumptions, recent comparative research suggests that the widely perceived incumbency bonus has turned into an ‘incumbency burden’. The greater public exposure of incumbents, compared to leaders of the opposition, and the decidedly critical approach of most media to government power means that incumbents (including former ones possibly willing to run again) are likely to constitute an electoral liability (Thesen et al., Reference Thesen, Green-Pedersen and Mortensen2024).
While there have been fewer CBPMs in our sample over time, we should by no means expect them to die out completely in the foreseeable future. In fact, ‘presidentialization’ and the closely related phenomenon of personalization with their deeply ambivalent features and effects, look very much as possible driving forces that have come to stay. Moreover, the recent and ongoing ‘rejuvenation’ of the executive elite, and of prime ministers in particular (Helms, Reference Helms, Keating, McAllister, Page and Guy Peters2023: 100-101), may well become a new and independent factor in the game of former leaders jockeying back to power. The prime example demonstrating the ruthlessness of some younger ‘personalized leaders’ is certainly Austria’s Sebastian Kurz, who—operating in a historical context marked by several long-term leaders and strongly age-based notions of authority—became the Second Republic’s first and only returning head of government in early 2020 at the age of 34. When losing the office for the second time in late 2021, he installed a personally close ‘placeholder’, Alexander Schallenberg, who willingly resigned just a few weeks later when it became clear that there would be no return of Kurz anytime soon.
There is obvious room for broadening the empirical scope of future research on this subject by enhancing the sample of countries, including those with supposedly different logics, or cultures, of party governance, elite recruitment, and political succession, such as the post-communist democracies of Central and Eastern Europe (CEE). Have CBPMs—recent prominent cases including Robert Fico in Slovakia and Donald Tusk in Poland—been more frequent in this region, as a cursory glimpse suggests, due to the eminent role of political entrepreneurs in establishing genuinely new parties (Hloušek et al, Reference Hloušek, Kopeček and Vodová2020; Haughton and Deegan-Krause, Reference Haughton and Deegan-Krause2020)? Especially in contexts with weakly institutionalized party systems, such as CEE, increasingly ‘presidentialized’ parties have also become more successful in parliamentary elections, allowing their leaders to return to the premiership despite their limited success in the office at previous occasions.
Last but not least, one could reverse the perspective applied here and treat CBPMs as the central independent rather than the dependent variable. As prime ministers may have a significant impact on the evolution of parliamentary democracy (Grotz and Weber, Reference Grotz and Weber2025), the key question then would be if CBPMs perform differently, that is, more or less successfully as compared to their non-returning counterparts, or their own first tenure. Historical experience suggests that the ‘comeback track’ of prime ministers may look fundamentally different than the first part of a premiershipFootnote 11 , though how exactly different patterns are shaped by the prevailing circumstances and the actors involved has very much remained a mystery. Such a perspective can be easily married with deep-running normative considerations about the nature of parliamentary democracy. In fact, the more evidence there is that comeback premierships primarily satisfy the ambitions of power-conscious parties and overzealous politicians, rather than the hopes and demands of their countries or the wider international community, the more they come as a challenge to the very idea of parliamentary democracy and its ever-alluring promise of democratic innovation and change.
Acknowledgments
We thank Marko Kukec, David M. Willumsen, and the three anonymous reviewers for their most helpful comments on earlier drafts of this manuscript, and David-George Rusu for his valuable research assistance.
Open access
