Incumbent takeovers

Abstract

The expansion of power by incumbent political leaders has become the subject of increased scholarly attention. In democracies, this is known as ‘subversions by the ruling executive’, ‘executive aggrandizement’, or ‘autogolpe’; in autocracies, researchers study ‘personalization’, ‘transition to personal rulership’, or ‘power-grabbing’. While the terminological landscape is rich, there is little conceptual agreement of what leader-driven power expansion is (and is not). Furthermore, we still lack broad data that allow us to investigate the phenomenon systematically across democracy and autocracy. The contribution of this article is twofold. First, it offers a unified approach to study leader-driven power expansion – incumbent takeovers – across the political regime spectrum. Second, drawing from 11 datasets and original data collection and coding, we introduce a new, comprehensive dataset on 495 individual takeover events carried out by 279 political leaders in 132 countries in the period 1918–2019. We provide estimates of the takeover onset years, the time to takeover, the length of the takeover spells, and discuss the differences between distinct indicators, inter alia. Future research may leverage these data for a better understanding of the drivers of incumbent takeovers as well as the role of takeovers in regime change, civil wars, coups, and uprisings.

Keywords

autogolpe democratic breakdown incumbent takeover new data personalism self-coup

Introduction

In recent years, scholars of democratization and comparative authoritarianism have argued that we are witnessing two new, important empirical trends: on one side of the political regime continuum, we see a rising number of incumbent-driven subversions of democracy (Bermeo, 2016; Svolik, 2015, 2019: 61); on the other side of the spectrum, a growth in the share of so-called personalist autocracies (Kendall-Taylor, Frantz & Wright, 2017; Geddes, Wright & Frantz, 2018). Of recent examples, Viktor Orbán’s expansion of power in Hungary represents the former, while Xi Jinping’s rapidly growing authority in China illustrates the latter. However, so far research on these closely related phenomena has been compartmentalized. That is, scholars tend to study the reduction of incumbent constraints that accompanies the erosion of democracy – takeovers perpetuated by elected leaders – separately from the increases in incumbent autonomy that are part and parcel of transitions to personal dictatorships.

In this article, we make the case for a unified approach and refer to both such events as incumbent takeovers. We argue that such an approach is needed as many similarities exist not only in terms of what takeovers are primarily about (incumbent-driven power expansion) and how they manifest themselves (be it through the closing of parliaments or the avoidance of term limits) but also because more protracted takeovers often involve countries traversing regime types (as the case of Hugo Chávez of Venezuela amply demonstrates). To understand the conditions under which incumbent takeovers happen across the political regime spectrum, we, therefore, need two things that are lacking today: conceptual clarity as well as comprehensive data that allow for systematic testing of theories. This article offers both.

First, we clarify what an incumbent takeover is and, equally important, what it is not. Based on a review of existing studies on ‘democratic backsliding’ (Bermeo, 2016; Waldner & Lust, 2018), ‘autogolpes’ (Przeworski et al., 2000: 21), ‘executive aggrandizement’ (Bermeo 2016), and ‘personalization’ (Svolik, 2012; Kendall-Taylor, Frantz & Wright, 2017; Geddes, Wright & Frantz, 2018), among others, we discuss similarities and differences between different forms of incumbent-driven power grabs in different regime settings.

Second, we build a new dataset on incumbent takeovers across the world from 1918 to 2019. Initially, we construct a candidate dataset with events drawn from a number of available sources and then supplement it with an additional data collection of our own. While many takeover events in democracies are covered reasonably well with several independent estimates typically available, takeovers in autocracies – because of the opacity of authoritarian politics – are not. However, we are able to account for takeovers in such contexts by first using the removal of term limits as an observable manifestation of expansion of incumbent power, and second, by leveraging the fact that leaders of regimes categorized as personal and personal hybrids (Geddes, Wright & Frantz, 2014) must have elevated themselves over institutions and elites at some stage during their tenure.

Next, we thoroughly clean the resulting candidate dataset for possible cases that do not fit the conceptual understanding of takeovers and then use the information compiled along with our own investigations of cases to determine the onset year of all incumbent takeovers as well as the takeover spell (the period of years across which the takeover happens).

The resulting dataset is the most comprehensive and accurate to date and includes details of 495 incumbent takeovers carried out by 279 political leaders in 132 countries. It allows researchers to investigate not only which rulers carry out incumbent takeovers but also when, why and how they are most likely to do so and what the consequences of such actions are. The availability of such data is a crucial first step towards a better understanding of when democracies are most at risk as well as when personalist rule is looming. Interestingly, our data lend only limited support to those that argue that incumbent takeovers are on the rise. In fact, we show that in recent decades, the average number of takeover onsets has declined. Only within electoral and minimalist democracies, we see an increase, albeit marginal.

What is an incumbent takeover?

We define an incumbent takeover as an event perpetuated by a ruling executive that significantly reduces the formal and/or informal constraints on his/her power. Our definition draws from the concepts of a ‘self-coup’ or ‘executive coup’ (Marshall & Marshall, 2014; Bermeo, 2016), ‘autogolpes’ (Przeworski et al., 2000: 21), or for that matter, ‘incumbent/executive takeover’ (Svolik, 2019, 2015). It also shares many similarities with related concepts such as ‘consolidation of incumbency advantage’ (Przeworski et al., 2000: 21), ‘executive aggrandizement’ (Bermeo, 2016), as well as more general ‘democratic backsliding’ (Bermeo, 2016; Waldner & Lust, 2018) and ‘personalization’ (Svolik, 2012; Kendall-Taylor, Frantz & Wright, 2017; Geddes, Wright & Frantz, 2018).¹ Next, we discuss the concept of incumbent takeover along four distinctions emphasized in our definition and then place it in relation to other, related concepts.

The key actor: The ruling executive

An incumbent takeover has to have the ruling executive as the main actor and the beneficiary. Many events categorized as takeovers in existing datasets are perpetuated by other civilian elites, such as vice-presidents, for instance. A takeover is an event that is initiated by the sitting president, prime minister, king, or whatever title the de facto leader of the country bears. This contrasts sharply with other forms of events that are also non-democratic in nature such as, for example, coups d’état. While coups obviously also have the sitting executive at the centre of the action, in these events they are the key target rather than the key perpetrator (Powell & Thyne, 2011: 250).

The regime context: A unified approach

We do not restrict incumbent takeovers to particular types of executives or regime contexts but instead embrace a unified approach. In a very general sense, the acquisition of personal power by a leader is a phenomenon that cuts across democratic, partly democratic, and non-democratic regimes. For example, doing away with formal institutions in the form of presidential term limits or parliamentary constraints on the leader’s decision-making are challenges that power-seeking incumbents of both democratic and non-democratic regimes face (Svolik, 2012: ch. 3; Geddes, Wright & Frantz, 2018: 79-85). Treating the tenure extension in democratic Venezuela in 1999 as comparable to the one in authoritarian Uganda in 2005, or the parliament shutdown in democratic Ecuador in 1970 or Peru in 1992 as comparable to the shutdown in autocratic Thailand in 1971 or Jordan in 2001 is, in our view, no more controversial than studying state repression, coups, and civil wars as general phenomena (see e.g. Hegre et al., 2001).

In addition, a unified approach makes our analyses less sensitive to chosen cut-off points between democracies and autocracies, and it allows us to study the many leaders that initiate a takeover first from within a more democratic regime, only to continue the process when the regime context has become non-democratic. Depending on whether we categorize Indonesia in the 1950s as a democracy or dictatorship, the concentration of executive power under Sukarno from 1957 can be studied as either a self-coup (Przeworski et al., 2000) or a personalism onset instead (Geddes, Wright & Frantz, 2014). The choice is arbitrary, however. Over six years, Sukarno engaged in a series of acts (such as the emergency law declaration, dissolution of parliament and a new pro-executive constitution, and the presidency for life) that are observationally equivalent whether they occur among previously democratically elected leaders or autocrats. Because we ‘permit’ incumbent takeovers to occur in different regime types, our approach allows researchers to decide on their own required cut-off point and examine takeovers across regime types or among only, for example, democratically elected leaders, or executives in electoral or closed autocracies. A unified approach that treats incumbent takeover as a general phenomenon still allows researchers to analyse democratic and authoritarian takeovers in isolation, but it also makes it possible to identify differences and commonalities between takeovers across different regime contexts.

The duration of the event: Protracted and rapid actions

As emphasized in the literature on democratic backsliding, the deterioration may happen both because of rapid, one-time events as well as gradual, incremental changes resulting from a series of interconnected actions (Svolik, 2015; Bermeo, 2016; Waldner & Lust, 2018; Lührmann & Lindberg, 2019). Concentration of power in the executive within autocratic settings is no different in this regard as dictators are said to extend personal power through a number of incremental actions that erode norms of accountability or through larger, more easily identifiable ‘power grabs’ (Svolik, 2012: Ch. 3; Geddes, Wright & Frantz, 2018: 79–85).² Again, these similarities across regime types speak in favour of the unified approach, and the differences in the duration of events suggest that incumbent takeovers should be regarded empirically as both intervals and point estimates.³ We, therefore, recognize incumbent takeover as both major single events (as with Olympio in Togo in 1960) as well as longer, drawn-out periods of smaller-scale events that in combination produce significant changes in terms of how political power is distributed, and even span across regime types (as in Ghana under Nkrumah, 1957–63).

The essence of the event: Significant, non-provisional, concentration of power in the executive

An incumbent takeover is an event that, if successful, significantly, and on a non-provisional basis, extends and deepens the political power of the ruling executive, and eliminates or significantly reduces executive constraints. As said, it may be a rapid or a more protracted event. Executives who violate the constitutional order by extending term limits, unlawfully dissolving parliament, abolishing executive elections, or suspending the constitution are clear examples of the rapid, one-time type of incumbent takeovers. In this way, observationally many takeover events resemble ‘autogolpes’, ‘executive coups’ or ‘self-coups’ (Cameron, 1998; Przeworski et al., 2000: 21; Bermeo, 2016) and ‘tenure extensions’ (Baturo, 2014, 2019), or both (Przeworski et al., 2013), which primarily have been studied in democratic settings but obviously also happen in autocratic ones. Importantly, drawing from observational data and/or relevant expert literature, we also include as rapid incumbent takeovers instances where leaders free themselves from informal constraints – for example, in the form of critical shifts in the balance of power between authoritarian rulers and coalitional elites such as when collective forums for decisionmaking are abolished (Svolik, 2012: 61).

The more gradual, protracted takeover events are more difficult to pinpoint exactly because they can consist of different combinations of actions and cover a longer time span. In isolation, each step does not significantly change the power of the executive. However, in combination they do. Though Hungary’s Victor Orbán or Russia’s Vladimir Putin have not dismissed parliament, suspended the constitution, or declared presidency for life, at least for the time being, no one would disagree

Figure 1.

Incumbent takeover in relation to other concepts

that their de facto power has increased dramatically over the years (Tolstrup & Souleimanov, 2021). When identifying protracted takeovers, we emphasize that changes in combination significantly increase the power of the executive but in isolation fall short of the criteria for rapid, one-time takeovers.

In Figure 1, we situate our concept of incumbent takeover (represented by the dashed area in the middle) in connection to the specific event types (in darker grey) it bundles together, as well as to other relevant, but broader, concepts (in lighter grey).

At the most aggregate level, the figure shows that takeovers are a special form of ‘autocratization’, which itself is a general form of regression away from liberal democracy (Lührmann & Lindberg, 2019). Because ‘autocratization’ spans across regime types, takeovers are also a subset of both Bermeo’s (2016) ‘democratic backsliding’ that occurs in democracies, on the one hand, and on the other, of the process of ‘personalization’ in dictatorships, as in Geddes, Wright & Frantz (2018). These concepts are both narrower concepts than that of ‘autocratization’. The two concepts also overlap, as seen from Figure 1, because they cannot be treated as completely distinct. Indeed, many instances of democratic backsliding and personalization entail changes in the amount of personal power of political leaders; a large number of personalist dictatorships emerge – and degenerate from – democracies through the actions of their incumbents.

Compared to Geddes, Wright & Frantz’s (2018) ‘personalization’, incumbent takeovers are best understood as events that happen during the initial stages of such a process. Thus, takeovers do not extend to related aspects such as the establishment and deepening of the cult of personality, which typically occur later during the tenure of many a personal ruler. Nor do we consider all personalization events during earlier stages as incumbent takeovers, such as an establishment of a new support party, for example. An incumbent takeover is manifested in a significant redistribution of power in favour of a political leader away from other ruling elites, such as the dismantling of collective leadership of the Derg in Ethiopia by Mengistu in 1977. That is, incumbent takeover is a special, more restricted form, of personalization. For these reasons, the dashed area in the middle that stands for the concept of incumbent takeover does not extend fully to the borders of ‘personalization’.

The concept of incumbent takeover is also related to that of ‘democratic backsliding’, but with its emphasis specifically on the ruling executive as the main actor and beneficiary, it is, again, a narrower one. For example, Bermeo (2016) recognizes six components of backsliding: ‘election manipulation’, ‘election-day fraud’, ‘classic coups’ and ‘promissory coups’, as well as ‘executive coups’ (self-coups), and ‘executive aggrandizement’. While all these components relate to autocratization generally, only the latter two are related to the incumbent takeovers specifically. Figure 1 shows that we treat ‘executive coups’, or what others term ‘autogolpes’ or ‘self-coups’ (Przeworski et al., 2000: 21), as a proper subset of takeovers. This is also the case for extensions of term limits, or ‘continuismo’, which many scholars include under self-coups (Cheibub, Gandhi & Vreeland, 2010; Przeworski, 2013) or treat as part of personalization (Geddes, Wright & Frantz, 2018). In turn, the concept of incumbent takeover is related to that of ‘executive aggrandizement’ but it is both less restrictive and at the same time narrower than ‘aggrandizement’, as we will explain.

The incumbent takeover concept includes many but not all instances of what Bermeo defines as ‘executive aggrandizement’. This is because the distinction between ‘self-coups’ and ‘aggrandizement’ is not always clear-cut. Indeed, Bermeo’s (2016: 10–11) ‘executive aggrandizement’ encompasses changes that occur at ‘a slower pace’ (than those under the ‘executive coup’). However, the examples of aggrandizement comprise ‘disassembling of institutions that might challenge the executive’ including elected assemblies (2016: 11–12). For example, Bermeo refers to Correa of Ecuador who ‘convinced the newly elected Constitutional Assembly to force the seated Congress into permanent recess and to assume legislative functions itself’ (2016: 12). However, closing down the legislature, as well as extending term limits that occurred at the same time, are also canonical examples of ‘autogolpes’, or what she terms ‘executive coups’.⁴ The distinction between aggrandizement, such as of Correa in 2007 or Chávez in 1999, and autogolpe such as Ibarra of Ecuador in 1970 or Somoza Debayle in 1972, is difficult because in both cases leaders close down their legislatures or convene Constitutional Assemblies. Furthermore, conceptually Bermeo’s ‘aggrandizement’ may also include term-limit changes (Bermeo, 2016: 12). However, ‘Continuismo’ cases are most often part of self-coups as in Cheibub, Gandhi & Vreeland (2010). In other words, while the distinction between ‘executive coups’ and ‘executive aggrandizement’ may be useful analytically to underscore the important changes in executive practice during the Cold War and post-Cold War periods, such a distinction is not easy to sustain conceptually as both ‘executive coups’ and ‘aggrandizement’ may include similar observable implications, even if they occur at a different pace.

At the same time, the concept of incumbent takeover is also narrower than ‘aggrandizement’. Though Bermeo’s (2016: 10–11) more protracted events of ‘executive aggrandizement’ fit many of the cases of incumbent takeovers, ‘aggrandizement’ also includes other changes related to backsliding more generally, such as those that target media freedoms, or judicial autonomy.⁵ These changes are related to backsliding, but not to how we understand takeovers. In this sense, our concept is a more restrictive one.

In summary, Figure 1 shows that the concept of incumbent takeover is narrower than those shaded in lighter grey because it demands that the incumbent executive is the main actor and beneficiary and that he/she is substantially empowered by the event. However, it is also a broader one in the sense that it is restricted to neither specific events (in darker grey) nor particular regime types. Incumbent takeovers are thus important and observable events that significantly strengthen the relative power of the ruler – be it in democracies, hybrid regimes, or autocracies. This delimitation makes our concept of incumbent takeover distinct from related, but broader concepts of ‘democratic backsliding’, ‘autocratization’, or other related concepts used to analyse regime developments across the democracy–autocracy spectrum. Hence, the unified approach.

Coding procedures

To build a dataset on incumbent takeovers, we follow a four-step approach. First, we construct a ‘candidate dataset’ (cf. Powell & Thyne, 2011: 252) with existing indicators from 11 available sources. Next, we make sure that each case conforms to our conceptual understanding. We then supplement the resulting candidate dataset with our own coding of incumbent takeovers in the periods of 1972–2019 and 1918–45. In turn, we use the resulting dataset to identify the onset year of the takeover as well as the takeover spell. Thus, our data work follows a ‘wisdom of the crowds’ approach. As the onset and duration of incumbent takeovers can be very difficult to assess empirically, we deliberately compiled as much information as possible in an attempt to reduce uncertainty and bias. Next, we explain the coding procedure in detail.

Compiling the candidate dataset

A number of available indicators capture one or more of the different types of incumbent takeover described earlier. In compiling the candidate dataset, we have, to the best of our knowledge, drawn from an exhaustive list of available data that captures the phenomenon of interest. Table I presents the 11 indicators chosen for the candidate dataset, all of which have been widely used in comparative empirical research. The different indicators vary in scope in terms of countries covered, the time-period coded, and the number of positive cases of incumbent takeover identified. What is crucial for our purpose is that the indicators in combination cover different regime types across the world and that they provide us with an opportunity to leverage, in most cases, more than one assessment per leader. This gives us a larger amount of information in evaluating when a takeover onset takes place and whether it takes place over a prolonged period. None of the existing indicators offer such distinctions.

Table I reveals that the majority of indicators – but not all – capture three ‘typical’ types of events initiated by the sitting incumbent: illegal closure of the legislature; the extension, removal, or avoidance of term limits; and the abolishment of elections. Some indicators, such as the one from Hyde & Marinov’s data (2012), only code one of these event types. Others, such as the indicators from the PIPE (Przeworski et al., 2013) and the V-Dem dataset (Coppedge et al., 2019), apply a less restrictive conceptual framework and thus cover more empirical referents. Apart from these three main categories, a number of indicators code events based on less explicit criteria. For example, Powell & Thyne (2011) do not define what they understand by ‘autogolpes’, nor do Marshall & Marshall (2014: 3) when they refer to ‘auto-coups’ as ‘subversion of the constitutional order.’ As we explain later, we thoroughly check whether all included cases do indeed fit our understanding of incumbent takeover.

We should note that many indicators primarily code cases of takeovers occurring from within more or less democratic settings. While some indicators attempt to include events in all regime types, they are clearly less consistent when it comes to coverage of events in autocracies. For instance, V-Dem’s (Coppedge et al., 2019) coding of ‘self coups’ includes the coronation of Bokassa in the Central African Republic but leaves out many similar events in other autocracies, where incumbents scrap the existing constitution or change informal rules to their own benefit. The indicators on tenure extensions from Baturo (2014, 2019) and Hyde & Marinov (2012) – because of their observability – offer consistent coverage of both democracies and autocracies. To address the ‘autocracy’ gap we – in addition to accounting for tenure changes – also leverage the codings from Geddes, Wright & Frantz (2014).⁶ In the candidate dataset, we initially include all years in power as protracted takeovers, during which incumbents start out as personalist (or personalist hybrids) or become personalist (or personalist hybrids). We do so based on the assumption that a takeover event must have happened in order for the ruler’s regime to be (re-categorized as such. For the years from 1918 to 1945, not covered by Geddes, Wright & Frantz (2014), we perform supplemental coding of our own (see next section).

Coding additional cases

As Figure 1 shows at the bottom, we supplement the candidate dataset with additional new coding of our own. First, because we are particularly concerned with not omitting any of the most recent takeovers, we code for the period 1972–2019, making extensive use of Freedom House data and the accompanying annual Country Reports. We identified the countries that were down-ranked from either ‘free’ to ‘partly free’ (70 cases) or from ‘free’ or ‘partly free’ to ‘not free’ (130 cases) (for a similar approach, see Svolik 2019: 20–23). Subsequently, we read relevant annual reports for these 200 instances and assigned aggregate categories (Figure 2 reports the frequencies) for general reasons behind downgrading. For additional details on the coding procedure and a full case list, see the Online appendix.

Overall, we identify six types of downgrades. Figure 2 shows that a significant increase in the concentration of power, what we treat as ‘incumbent takeover’, happens in connection with 10 ‘partly’ and 32 ‘non free’ transitions. Also noteworthy, alongside coups d’état, incumbent takeover is the most common reason behind

Table I.

Overview of data sources

Source	Concept	Coverage (years)	Coverage (regime types)	N (candidates/events included)	Illegal closure of legislature	Tenure (extension, removal or avoidance)	Abolishing elections or changing election rules	Miscellaneous
Available indicators
Przeworski et al., 2000	‘Transitions to dictatorships by incumbents (autogolpes)’	1950–90	Democracies	22/19	x		x	x
Cheibub, Gandhi & Vreeland, 2010	‘Consolidation of incumbent advantage’	1946–2008	Democracies	40/35	x	x		x
Powell & Thyne, 2011	‘Seizure of power by chief executive (autogolpes)’	1950–2010	All regimes	52/47	x			x
Haggard & Kaufman, 2012	‘Elite reaction or intra-elite reversion’	1980–2000	Democracies	23/8				x
Hyde & Marinov, 2012	‘Tenure extension’	1946–2012	All regimes	83/13		x
Przeworski, 2013	‘Transitions to dictatorships by incumbents (autogolpes)’	1789–2008*	Democracies	85/74	x	x	x
Marshall & Marshall, 2014	‘Auto-coups’	1946–2013	Democracies	33/31	x			x
Svolik, 2015	‘Incumbent takeover’	1919–2008*	Democracies	23/22	x		x	x
Baturo, 2019	‘Tenure extension’	1945–2018	All regimes	130/125		x	x
Geddes, Wright & Frantz, 2014	‘Personal or personal hybrid regime’	1946–2010	Autocracies	139/122**				x
Coppedge et al., 2019	‘Self-coup’	1919–2019*	All regimes	82/78	x	x	x	x
Authors’ coding
Freedom House	‘Incumbent takeover’	1972–2019	Democracies & hybrid regimes	200/42	x	x		x
Boix, Miller & Rosato, 2013; Djuve, Knutsen & Wig, 2020	‘Incumbent takeover’	1918–45	Autocracies	162/24	x	x	x	x
Geddes, Wright & Frantz, 2014	‘Incumbent takeover’	1946–2010	Autocracies	122/122**	x	x	x	x

* These datasets all go back to 1789. We include only the period from 1919 onwards.

** The included GWF (Geddes, Wright & Frantz, 2014) cases are all hand-coded to determine the onset year of the incumbent takeover.

Figure 2.

Authors’ coding of reasons for Freedom House downgrades

transitions to ‘unfree’ status. It is also a very common reason for transitions from ‘free’ to ‘partly free’ (14%). For all 42 cases we code as incumbent takeovers, Freedom House refers explicitly to the fact that a significant accumulation of executive power has occurred.⁷

We also supplement with our own coding of takeovers in non-democracies in the 1918–45 period as these seem less well represented by the indicators we rely on. First, we use the democracy-autocracy measure from Boix, Miller & Rosato (2013) to identify all authoritarian country-years in the period, as well as the historical regime dataset from Djuve, Knutsen & Wig (2020) to identify the relevant regimes’ spells and their leaders. Of the 190 cases identified, we found that 28 already appear in our dataset. For the remaining 162 cases, we used various secondary sources to determine whether incumbent takeovers happen and when. In total, we identified and included 24 additional cases of incumbent takeover. A list of these takeovers as well as coding details are available in the Online appendix.

Cleaning the candidate dataset

Altogether, the candidate dataset includes 779 events – 713 from the 11 existing indicators, 42 from our coding of Freedom House reports, and 24 from our coding of autocracies in the 1918–45 period. Next, we account for possible duplicates and different types of false positives. Table I reports both the number of events included from each indicator in the candidate dataset and the number that remains after our cleaning process.

As a first step, we remove all duplicates – that is, cases, where different sources assign an incumbent takeover to the same leader in the same year. In total, we found 156 pure duplicates. We also excluded 56 NELDA years that categorize term limit events that we already have from Baturo (2019) but follow a different rule of operationalization (see Online appendix for details). Thus bringing the candidate list of events down to 566 events.

Next, we clean the dataset for false positives. That is, we comb through all remaining candidate events from each of the 11 indicators to make sure that the included events conform to our conceptual understanding of an incumbent takeover. Altogether, we identified 71 false positives. Many are found that represent coups or coup-like events, in which elites or other actors challenge the incumbent and not the other way around. This includes coups in Pakistan in 1999 and Thailand in 1991, the ousting of the king in Jordan by parliament in 1951, the vice-president deposing the president in Honduras in 1954, and the USSR-instigated seizures of power in Eastern Europe, among others. We also exclude a number of cases that more closely resemble external interference or externally orchestrated takeovers such as Malta in 1930, 1936, and 1958, when direct British rule was imposed at a time when the country was not independent. Finally, we remove events such as when term limits are imposed, not removed, and no grandfathering clause is established (e.g. Ecuador 1979, or Haiti 1988), or when term-limit extensions are only short term and the incumbent steps down immediately thereafter (e.g. Brazil 1988, or Comoros 2009).

In addition, we remove false positives among the included interval estimates from Cheibub, Gandhi & Vreeland (2010) and Geddes, Wright & Frantz (2014) (henceforth GWF). For example, we identify 17 cases from GWF that do not conform to our understanding of incumbent takeover. Most likely, these leaders have been classified as personalist simply as a sort of residual solution. Examples include Andry Rajoelina in Madagascar and Rahmon Nabiyev in Tajikistan. We also identified false positives among the group where leaders are second (or third) in line of a personalist regime but seemingly do not accomplish a takeover. Rather, they merely ‘inherit’ personalism in the same regime spell (for more details, see Online appendix).

Coding the onset and end year of incumbent takeovers

The resulting event dataset comprises 495 observations of takeovers carried out by 279 political

Figure 3.

Examples of cases with both interval and point estimates

leaders in 132 countries across a 100-year period. As the final step, we use these data to determine a single onset year and, if applicable, an end year of the takeover spell for each of the 279 rulers in our dataset. Some of the indicators we used to build this dataset code discrete events in the form of particular years, while others – Geddes, Wright & Frantz (2014) and Cheibub, Gandhi & Vreeland (2010) – offer intervals. This means that for some rulers, we have only one point estimate for a takeover event, and for others, we have a number of point estimates, a combination of point estimates and intervals, or only an interval.

On average, excluding leaders for which only one takeover observation from one source is available, the dating of takeovers varies by just over three years. This disagreement is not surprising as the indicators we use focus on particular variants of incumbent takeovers, and among those that code similar types, conceptual and operational differences often exist. It is a modelling decision whether to regard such cases as one event, different takeover events, or as cases of protracted takeovers spanning several years. It also tells us about the degree of uncertainty as to when a particular takeover has occurred. Thus, we retain the information provided by each indicator in our dataset, and include the range determined by different indicators, to allow researchers to pursue different approaches. However, as we will explain, we also make it possible for researchers to analyse such cases relying on one takeover event.

To illustrate the complexity, Figure 3 presents examples of rulers for which we have both different discrete point estimates and one or two interval estimates. Fujimori of Peru represents a consensus case, for which most indicators agree that the main takeover occurred in 1992 – a classic case of an autogolpe. The cases of Ghana’s Nkrumah and Russia’s Putin represent the exact opposite: significant disagreement between indicators. In the former example, Nkrumah did indeed carry out several takeover-like actions starting in 1958 with the abrogation of the rule of law and ending in 1964 with the adoption of the lifetime presidency. In Russia, no manifest takeover events happened (Baturo & Elkink, 2016). Instead, Putin’s authority building has been a genuinely prolonged process, where a longer interval informs us about the incremental nature of the takeover. Finally, the example of Barre of Somalia represents some of the more closed autocracies, for which, apart from easily observable term-limit changes, we

Figure 4.

Incumbent takeovers (total number of onsets) across the world 1918–2019

simply lack reliable information on elite struggles and the ruler’s acquisition of personal power. Thus it is clear that apart from cases like Fujimori, careful individual judgement is needed to determine the appropriate takeover onset year and interval. Next we briefly explain how we do that.

For rulers with only one point estimate and no interval estimates available, we code the year of that point estimate as the onset and treat it only as a one-year discrete event. For rulers with point estimates available for different years, we treat the first of these estimates as the onset year and the last as the end year of the takeover event. Take the example of Aleksandr Lukashenka’s takeover in Belarus. Following our rules, it is coded to start in 1995, when he significantly expanded presidential powers, and it ends in 2004, when he, by means of a referendum, removed presidential term limits.

Lastly, for rulers with interval estimates available, we followed a two-step approach. First, we checked whether the first year of the interval or another year should serve as an estimate of the first year of the takeover. For some cases, the first year of the interval is equal to the first year of the ruler’s tenure. For others, the two differ. For example, Hugo Chávez in Venezuela assumes power in 1999 but is not coded by the GWF as personalist until 2005. For cases like these, we treat the onset year of personalism as a potential takeover event on par with the other events found among included indicators. For the 110 rulers, for which the GWF dataset codes the regime as personalist or personalist hybrid from the beginning of their tenure, we hand-coded whether the entry year or a later year should serve as the onset year of the incumbent takeover. In total, we re-code onset years for 61 of the 110 cases (see Online appendix for details). For the remaining 49 rulers, we found the entry year to be the best estimate. In the Online appendix, we provide an overview of the onset and the duration of all 279 takeovers in the dataset.

Empirical patterns

What can the new data tell us about incumbent takeovers across the world and over time? Figure 4 provides a spatial overview. As expected, those countries that have seen the most takeovers are located in regions we know are, and have been, particularly prone to political instability in general. That is, most takeovers take place in Africa, followed by Latin America and the Caribbean. Bolivia tops the list with nine events, with Ecuador, Panama, Pakistan, and Haiti following with six takeovers each. Only a handful of the oldest and most established democracies in the world have so far

Figure 5.

Time from entry into office to onset of incumbent takeovers (years). Percent of leaders per regime category and the year of their first/only takeover. More democratic are leaders with positive Polity2 scores at the time of entry into office, including nine leaders of transitional regimes with Polity2 scores of 0; non-democratic otherwise

been immune. One exception to this rule is France, which is recorded to have had two takeover events: in 1940 and 1958.

In Figure 5, we illustrate differences in how quickly rulers commit takeovers after taking office. Most takeovers (60%) happen within the first three years after the leader has entered office. The difference between more democratic and more autocratic leaders in this regard is not stark. That said, during their first two years in power more authoritarian leaders do move faster than their democratic peers (45% against 28%). As years in office pass, we see fewer and fewer incumbent takeovers, which occur, predominantly, when rulers face, and challenge, two-term limits.

Figure 6 plots the number of incumbent takeover onsets per year throughout the whole period. On average, we see around four incumbent takeovers per year. Peak periods include the second reverse wave, with 43 onsets in the 1960s, but also in the liberal yet tumultuous 1990s when incumbent takeovers happened 39 times. The worst takeover years are 1960 and 1971, with nine events occurring in each. The figure also shows differences between takeovers that happen from within more and less democratic countries. As an incumbent takeover will color the regime rankings for that year, we divide leaders based on their status at the time of entry into office. Those who enter office with a positive Polity2 score we treat as more democratic, and the

Figure 6.

Incumbent takeovers (total number of onsets per year) over time 1918–2019. LOWESS smooth lines and raw counts (in grey) of the first/only takeovers in a given year. More democratic are leaders with positive Polity2 scores at the time of entry into office, including nine leaders of transitional regimes with Polity2 scores of 0; non-democratic otherwise

remainder as non-democratic.⁸ Not surprisingly, we find that non-democratic takeovers (172 leaders) by far outnumber takeovers that happen in more democratic countries (107 leaders). Moreover, they have predominated in almost all decades of the time period covered.

However, since about 1980, ‘democratic’ takeovers have been on the rise in absolute numbers, and in the most recent decades even outnumber non-democratic cases. Nonetheless, the increase is marginal, and as we show in the Online appendix, these democratic takeovers happen only in electoral and minimalist democracies. Moreover, if we account for the fact that the number of independent countries increases in the international system over time and the absolute number of autocracies and democracies also differs accordingly, the relative frequency of incumbent takeovers is actually in decline. Interestingly, this downward trend is driven by declines in takeovers in autocracies, while the share in more democratic countries has been relatively constant in the last few decades (for details, see the Online appendix). Thus, our data lend only limited support to the argument that incumbent-driven subversions of democracy are on the rise (Svolik, 2015, 2019: 61; Bermeo, 2016). One explanation for this surprising finding, perhaps, is that we employ a more demanding threshold for what we treat as incumbent takeovers. Thus, it may be that executives across the globe are concentrating more power in their own hands in these years, but that this power expansion is just too gradual, not manifested in any observable institutional or behavioral indicators, to be included in our data. Though worrying in its own right, our results show that the situation is perhaps less alarming and might still be reversed.

Conclusion

Many leaders across the world, such as Chávez in Venezuela, Putin in Russia, and Orbán in Hungary, have taken over their previously democratizing regimes and concentrated political power in their own hands. Likewise, rulers of institutionalized non-democratic regimes, such as Xi Jinping of China or Hun Sen of Cambodia, have increased their personal influence and taken over collective leadership structures. While important differences between these processes exist, we centred on the commonalities of incumbent power concentration and presented a new, comprehensive dataset of incumbent takeovers across regime types.

The dataset allows researchers to investigate a wealth of important questions such as: Which rulers make incumbent takeovers? And when are they most likely to do so? This is not only important for understanding when democracies are most at risk but also for more accurate forecasting of when and where non-democracies are most prone to degenerate into arbitrary one-man rule, which we know to be the regimes most prone to conduct an aggressive foreign policy, including initiating interstate disputes and military conflicts (Peceny, Beer & Sanchez-Terry, 2002; Way & Weeks, 2014). Available datasets on democratic backsliding and self-coups do not provide the same consistency in coding and comprehensiveness in coverage across both time and regime types as we do. Though recent empirical research on personalization from Geddes, Wright & Frantz (2018) allows us to compare the levels of personalist rule, it remains difficult to identify specific transitions to personal rulership and to specify how long such transitions take. We thus have something to offer both to scholars working exclusively on democracies or autocracies, but most importantly, our dataset offers opportunities to investigate systematically whether similar or different factors explain takeovers across political regime types.

This dataset will also facilitate systematic analyses of the effects of incumbent takeovers on intrastate political and conflict-related outcomes. Case examples from around the world show that incumbent takeover is a phenomenon crucial for understanding political regime developments and their consequences. With our data, political scientists can now – in a more systematic way – study whether incumbent takeovers serve as potential triggers of coups and mass demonstrations, precursors for increases in the use of state repression, or as short-term causes for the outbreak of civil wars. Again, the benefit of the incumbent takeover dataset is that such analyses can both be carried out for democracies and autocracies in isolation but also be investigated across regime types. This is an important step forward.

Footnotes

Acknowledgments

We thank the journal editor, anonymous reviewers, as well as the participants at a June 2020 workshop on ‘Political regimes and their correlates’ at the Department of Political Science at Aarhus University for valuable feedback.

Replication data

The dataset, codebook, and do-files for the empirical analysis in this article, along with the Online appendix, are available at . All analyses were conducted using Stata 12.1 and R 3.5.2.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Jakob Tolstrup

Notes

References

Baturo

Alexander

(2014) Democracy, Dictatorship, and Term Limits. Ann Arbor, MI: University of Michigan Press.

Baturo

Alexander

(2019) Continuismo in comparison: Avoidance, extension, and removal of presidential term limits. In: Baturo

Alexander

Elgie

Robert

(eds) The Politics of Presidential Term Limits. Oxford: Oxford University Press, 75–102.

Baturo

Alexander

Elkink

Johan A

(2016) Dynamics of regime personalization and patron–client networks in Russia, 1999–2014. Post-Soviet Affairs 32(1): 75–98.

Bermeo

Nancy

(2016) On democratic backsliding. Journal of Democracy 27(1): 5–19.

Boix

Carles

Miller

Michael K

Rosato

Sebastian

(2013) A complete data set of political regimes, 1800–2007. Comparative Political Studies 46(12): 1523–1554.

Cameron

Maxwell A

(1998) Latin American autogolpes: Dangerous undertows in the third wave of democratisation. Third World Quarterly 19(2): 219–239.

Cassani

Andrea

Tomini

Luca

(2018) Reversing regimes and concepts: From democratization to autocratization. European Political Science 19(3) DOI:10.1057/s41304-018-0168-5.

Cheibub

José Antonio

Gandhi

Jennifer

Vreeland

James Raymond

(2010) Democracy and dictatorship revisited. Public Choice 143(1–2): 67–101.

Coppedge

Michael

Gerring

John

Knutsen

Carl Henrik

Lindberg

Staffan I

Teorell

Jan

Altman

David

Bernhard

Michael

Steven Fish

Glynn

Adam

Hicken

Allen

Lührmann

Anna

Marquardt

Kyle L

McMann

Kelly

Paxton

Pamela

Pemstein

Daniel

Seim

Brigitte

Sigman

Rachel

Skaaning

Svend-Erik

Staton

Jeffrey

Wilson

Steven

Cornell

Agnes

Gastaldi

Lisa

Gjerløw

Haakon

Ilchenko

Nina

Krusell

Joshua

Maxwell

Laura

Mechkova

Valeriya

Medzihorsky

Juraj

Pernes

Josefine

Römer

Johannes von

Stepanova

Natalia

Sundström

Aksel

Tzelgov

Eitan

Wang

Yi-ting

Wig

Tore

Ziblatt

Daniel

(2019) V-Dem [Country-Year/Country-Date] Dataset v9. edited by Varieties of Democracy (V-Dem) Project (https://doi.org/10.23696/vdemcy19).

10.

Djuve

Vilde Lunnan

Knutsen

Carl Henrik

Wig

Tore

(2020) Patterns of regime breakdown since the French Revolution. Comparative Political Studies 53(6): 923–958.

11.

Geddes

Barbara

Wright

Joseph

Frantz

Erica

(2014) Autocratic breakdown and regime transitions: A new data set. Perspectives on Politics 12(2): 313–331.

12.

Geddes

Barbara

Wright

Joseph

Frantz

Erica

(2018) How Dictatorships Work: Power, Personalization, and Collapse. New York: Cambridge University Press.

13.

Haggard

Stephan

Kaufman

Robert R

(2012) Inequality and regime change: Democratic transitions and the stability of democratic rule. American Political Science Review 106(3): 495–516.

14.

Hegre

Håvard

Ellingsen

Tanja

Gates

Scott

Gleditsch

Nils Petter

(2001) Toward a democratic civil peace? Democracy, political change, and civil war, 1816–1992. American Political Science Review 95(1): 33–48.

15.

Hyde

Susan D

Marinov

Nikolay

(2012) Which elections can be lost? Political Analysis 20(2): 191–210.

16.

Kendall-Taylor

Andrea

Frantz

Erica

Wright

Joseph

(2017) The global rise of personalized politics: It’s not just dictators anymore. Washington Quarterly 40(1): 7–19.

17.

Lührmann

Anna

Lindberg

Staffan I

(2019) A third wave of autocratization is here: What is new about it? Democratization 26(7): 1095–1113.

18.

Marshall

Monty G

Marshall

Donna Ramsey

(2014) Coup d’état Events, 1946–2013. Center for Systemic Peace (https://www.systemicpeace.org/inscr/CSPCoupsCodebook2015.pdf).

19.

Peceny

Mark

Beer

Caroline C

Sanchez-Terry

Shannon

(2002) Dictatorial peace? American Political Science Review 96(1): 15–26.

20.

Powell

Jonathan M

Thyne

Clayton L

(2011) Global instances of coups from 1950 to 2010: A new dataset. Journal of Peace Research 48(2): 249–259.

21.

Przeworski

Adam

Michael Alvarez

Alvarez

Michael E

Cheibub

Jose Antonio

Limongi

Fernando

(2000) Democracy and Development: Political Institutions and Well-being in the World, 1950–1990. Vol. 3. Cambridge: Cambridge University Press.

22.

Przeworski

Adam

Newman

Lindsay S

Park

Queralt

Didac

Rivero

Gonzalo

Shin

Kong Joo

(2013) Political Institutions and Political Events (PIPE) Data Set. Department of Politics, New York University.

23.

Svolik

Milan W

(2012) The Politics of Authoritarian Rule. Cambridge: Cambridge University Press.

24.

Svolik

Milan W

(2015) Which democracies will last? Coups, incumbent takeovers, and the dynamic of democratic consolidation. British Journal of Political Science 45(4): 715–738.

25.

Svolik

Milan W

(2019) Polarization versus democracy. Journal of Democracy 30(3): 20–32.

26.

Tolstrup

Jakob

Souleimanov

Emil

(2021) Centre–province relations in personalist autocracies: Explaining the emergence and survival of powerful provincial leaders. Government and Opposition: 1–21 (https://doi.org/10.1017/gov.2021.13).

27.

Waldner

David

Lust

Ellen

(2018) Unwelcome change: Coming to terms with democratic backsliding. Annual Review of Political Science 21: 93–113.

28.

Way

Christopher

Weeks

Jessica LP

(2014) Making it personal: Regime type and nuclear proliferation. American Journal of Political Science 58(3): 705–719.