Render unto Caesar Just a Little Bit Longer: The Relationship between Constitutional Reforms and Executive Survival

Abstract

Since the third wave of democracy, term limits have become a popular fixture of most constitutions intended to constrain the executive. Yet, recent constitutional reforms around the world show that presidents seeking re-election sometimes overturn the entire constitutional order to extend their power. What is the impact of these constitutional manipulations on the longevity of the executive in office? Using survival analysis of all political leaders and national constitutions from 1875 to 2015, this article demonstrates, for the first time, that when ‘authoritarian-aspiring’ presidents remove constitutional term limits, they increase their stay in office by more than 40%. Our findings contrast with a widely held position in the comparative authoritarian literature suggesting that dictators survive longer under institutional constraints. On the contrary, we argue that by removing constitutional barriers, rulers consolidate more power at the expense of their most ambitious allies and can stay in power longer.

In a letter to James Madison, Thomas Jefferson famously expressed his disappointment that presidential term limits were absent from the American Constitution and suggested that it was one of the major defects of the document (Jefferson 1787/2006). Indeed, Jefferson believed that constitutional constraints on the executive would prevent tyranny. Concerns about unrestrained executives have become even more urgent today, as executives in democracies and non-democracies alike, from Hungary to Venezuela, and Egypt to Uganda, have used constitutional changes to increase their own power or that of their parties and to ensure longevity in office. The question remains, however, whether these constitutional reforms actually increase the longevity of executives in office.

Recent studies show that constitutional reforms are among the most common strategies authoritarian leaders utilize to stay in power (Baturo 2014; Bunce and Wolchik 2010). The extant literature has advanced our knowledge about the circumstances under which political leaders are more likely to remove constitutional barriers in order to stay in power longer (see, for example, Baturo 2014; Bunce and Wolchik 2010; McKie 2017). These authors depict how wily authoritarians seek to change the rules to suit their self-interests of staying in power. Although these contributions have raised this important question for various cases, they have failed to examine systematically whether this ‘constitutional change’ strategy does indeed help authoritarian leaders prolong their tenure. Employing survival analysis, we test this relationship worldwide – for the first time – and show that constitutional alteration has a considerable positive impact on survival of authoritarians.

Building on several available databases, we develop an original data set of all leaders of independent states from 1875 to 2015 as well as all constitutional changes during their tenure. We study all world leaders rather than only authoritarians to account for those who come to power through democratic institutions but gradually undermine those institutions to remain in power. We show that when authoritarians undertake multiple constitutional reforms, their tenure in office is extended for an average of 10 years, whereas democrats who reform constitutions more than once remain in office an additional four years.¹ In other words, authoritarians who reform their constitutions last 2.5 times longer in office than constitution-reforming democrats.² More specifically, the results show that removing constitutional term limits reduces their risk of removal from office by over 40%. Our survival analysis also shows some support for the effect of power consolidation, regime type and government system on political leaders' survival in office.

The findings suggest that authoritarian leaders survive longer when they remove institutional constraints – in particular, constitutional term limits. This view contrasts with a widely held position in the comparative authoritarian literature suggesting that dictators survive longer under institutional constraints (de Mesquita et al. 2003; Gandhi and Przeworski 2007; Wintrobe 1998). These institutional constraints, however, facilitate the incumbent's survival mostly in ‘contested’ electoral authoritarian regimes and only if they mitigate the credible commitment problem (Magaloni 2008; Svolik 2012). Instead, we argue that authoritarian leaders are better off removing such institutional constraints. Moreover, our analysis shows that the impact of constitutional term-limit alteration does not depend on the degree to which the incumbent consolidates power prior to changing the constitution. That is, the effect of removing constitutional barriers such as term limits is not contingent upon the degree to which leaders amass power and consolidate their positions (‘personalizing’ their rule).

Finally, the results from this analysis show that beyond attempts at window-dressing, ‘drafting from above’ (Eisenstadt et al. 2017) is sometimes motivated by re-election-seeking presidents who are willing to overturn the entire constitutional order so they can stay in power longer.³ This claim furthers extant research by demonstrating concrete reasons for leaders to expend considerable resources to write and implement a new constitution undermining democratic elections and extending their own eligibility to run for additional terms, indicating that constitutional reform may be adverse for democracies more often than it is beneficial.

Indeed, constitutional reforms are critical mechanisms for guarding against rigid judicial review which can undermine the will of the majority. Jefferson would have been delighted to know, for instance, that the American Constitution was eventually amended in the mid-20th century to institutionalize term limits. However, as our empirical analysis shows, the same mechanisms for constitutional change can also undermine democratic norms when they are used for undemocratic purposes (Dixon and Landau 2015). This dilemma has spurred a debate about the legality of constitutional changes and amendments and the possible threats to democracy that might result from ‘unconstitutional’ constitutional changes (see, for example, Albert 2009; Dixon 2011; Dixon and Landau 2015). The same fears shroud wholesale constitutional changes at moments of refounding the social contract, when ‘authoritarian-aspiring’ leaders in both democratic and non-democratic settings ‘reform’ their constitutions solely to extend their rule.

This article proceeds as follows. In the next section, we elaborate on the role of constitutional change in authoritarians' ‘menu of manipulation’ (Schedler 2002) and the circumstances under which constitutions are changed. Next, we posit our hypothesis that altering constitutional term limits increases the incumbent's longevity in office. We then present descriptive data for survival rates and statistically test the relationship between constitutional change and survival in office. Finally, we conclude by discussing and interpreting the results.

Constitutional change and authoritarian survival

The survival of authoritarian incumbents is a function of several factors. For example, semi-authoritarian or hybrid regimes are believed to be less durable than closed authoritarian ones (Brownlee 2009; Howard and Roessler 2006; Knutsen and Nygard 2012). Although pseudo-democratic institutions such as elections and legislatures can be successfully manipulated to legitimize the incumbent's rule and to co-opt the opposition in hybrid regimes (Boix and Svolik 2013; Gandhi 2008), they can also add uncertainty to elections that could result in an authoritarian's removal from power and the onset of regime liberalization (Bunce and Wolchik 2010; Levitsky and Way 2002).

Authoritarian incumbents, whether in hybrid or closed authoritarian regimes, deploy a series of strategies to ensure their survival and to ward off potential challenges to their rule, whether from the opposition or the masses. Authoritarian rulers in hybrid regimes use a range of strategies, including vote engineering, manipulation of the media and harassment of the opposition to influence electoral outcomes and ensure their survival (Eisenstadt 2004; Hale 2011; Krastev 2006; Schedler 2002). A combination of economic incentives, ideological appeal and indoctrination, as well as threat of coercion, are also used to keep challengers at bay and to ensure popular support for the regime (Magaloni 2006). In particular, autocrats possess two principal instruments – repression and co-optation – to survive in power (de Mesquita et al. 2003; Wintrobe 1998).

Authoritarian regimes with large oil revenues and other sources of rent can co-opt the ruling circle through private transfers and are less likely to invest in creating binding institutions (Wright 2008). However, when authoritarian regimes face economic difficulties in the absence of rent sources and are challenged by a strong opposition, they are more likely to invest in binding institutions to co-opt the opposition and their inner circle through policy concessions (Smith 2005).

Authoritarians manipulate these political institutions to extend their rule and ensure survival by ‘absorbing, channeling, dampening, deflecting, or dispersing oppositional energies’ (Schedler 2013: 73–74). To ensure survival and prevent threats to their power, authoritarian rulers establish institutions such as legislatures and parties to manipulate and co-opt their allies and opponents (Wright and Escriba-Folch 2012). Specifically, legislatures are used to offer a venue for incorporating larger segments of society, including political opponents and rivals (Eisenstadt 2004; Gandhi 2008; Gandhi and Przeworski 2007).

Institutional constraints, however, can enhance authoritarians' survival only if they alleviate the commitment problem (Svolik 2012). That is, they should allow the dictator's ‘loyal friends’ to credibly detect and then punish the dictator's opportunism (Magaloni 2008; Svolik 2012). In other words, mechanisms are in place, formally and informally, to ensure that the authoritarian leader will safeguard the interests of the elites who bolster the regime or be accountable – within his or her narrow constituency – to sanctions (North and Weingast 1989).⁴ Since authoritarian regimes differ from democracies in the origins of their leaders and ruling coalitions, they differ in how they behave, how (or whether) they resolve this commitment problem, and how stable they are (Geddes 1999; Geddes et al. 2014). For example, studies suggest that single-party regimes and monarchies generally tend to be politically more stable and durable compared to military and personalist regimes, because they better succeed in resolving the commitment dilemma (Brownlee 2009; Geddes 1999).

Besides ruling political parties and elections, constitutions are also among the political institutions which autocrats can use to signal credible commitment to their inner circle and potential rivals (see, for example, Svolik 2012; Weingast 1993). As Tom Ginsburg and Alberto Simpser argue (2014: 2), ‘authoritarian constitutions can help oligarchic actors to work together by establishing focal points, procedures, and institutions, thereby addressing problems of coordination and problems of commitment’. By limiting executive tenure to two consecutive five-year terms, among other reforms, the revised Chinese Constitution under Den Xiaoping, for instance, effectively alleviated the commitment problem in authoritarian power-sharing and contributed to the post-Mao regime's stability (Svolik 2012: 85–86). Yet, autocrats often change the constitutional order to extend their longevity in power, which can enhance rather than alleviate the commitment problem. In such cases, authoritarian durability is not a function of institutional constraints such as constitutional term limits. On the contrary, we argue that removing such institutional constraints can lead to more longevity in power. By removing constitutional barriers, rulers consolidate more power at the expense of their most ambitious allies and can stay in power longer.⁵

Indeed, most dictators carry out complete constitutional changes or extensive amendments to ensure their interests are secured and their survival in office cannot be challenged. One of the most common constitutional features regularly under assault by authoritarians is executive term limits. Since the third wave of democracy, term limits have become a popular fixture of most constitutions intended to constrain the power of the executive (Ginsburg et al. 2011). As such, amending and modifying them might signal non-democratic intentions to an international audience or domestic allies and rivals. Nonetheless, amending or changing the constitution to remove term limits altogether, or to extend their tenure through institutional design changes, is common across non-democracies (Bunce and Wolchik 2010). Some studies suggest that between 20% and 30% of presidents extend their term limits in some way (Baturo 2014; Ginsburg et al. 2011). In a recent study tracing the constitutional strategies of 234 incumbents since 2000, Mila Versteeg et al. (2020) find that two-thirds of attempts to overstay in office involved constitutional amendments, usually by removing term limits.⁶

Our descriptive analysis shows that removing the constitutional term-limit barrier has a significant effect on the incumbent's survival in power. As Figure 1 shows, incumbents who do not remove executive term limits when changing their constitutions last on average only about two years in office. By contrast, those who lift the term limits while changing their constitutions survive about 10 years. In other words, removing constitutional term limits quintuples the incumbent's longevity in office, presenting authoritarian incumbents with a very efficient and tempting choice.

Figure 1. Term-Limits Removal and Leader's Survival Rate

Several factors, however, determine when these presidents opt to change their constitutions to extend their term limits. In a comprehensive study, Alexander Baturo (2014) suggests that both political institutions and personal interests determine whether presidents would circumvent term limits and extend their rule. If presidents found or control their party and dominate and control the legislature and the judiciary, they can readily circumvent constitutional term limits. However, personal gains and monetary interests are also an important factor explaining manipulation of institutions and term limits. Specifically, if the amount of rent-seeking from the office of the president is substantial, it will be more difficult for the incumbent to leave office; hence the increase in the likelihood of removing term limits.

In other important research, Kristin McKie (2017) argues that constitutional term limits on the executive were not imposed in numerous African countries where one unified party controlled the constitutional reform process and perceived that it would win future elections. By contrast, in countries where the likelihood of electoral victory is less certain, incumbents impose term limits as an electoral insurance mechanism.

Although authoritarians do not shy away from violating states' institutions for personal gain, blatantly removing barriers to their tenure can delegitimize their rule in the eyes of both domestic and international audiences. Hence, introducing a myriad of constitutional changes can serve to mask the key specific change extending the incumbent's tenure. For instance, Alexander Baturo and Robert Elgie (2018) show that after the Cold War the number of bicameral dictatorships increased significantly and a closer examination of constitutional changes in those polities demonstrates that constitutional legislative reforms occurred in tandem with modifications to executive tenure restrictions such as term limits. Hence, increasingly, authoritarians use wholesale constitutional reform projects to draw attention away from the non-democratic changes they implement to extend their tenure (Baturo and Elgie 2018, 2019).

In sum, authoritarian incumbents use a wide array of strategies from policy concessions and rent distribution (Gandhi and Przeworski 2007) to electoral manipulation (Hale 2011; Levitsky and Way 2010) and constitutional changes (Ginsburg et al. 2011; Negretto 2012) to ensure their survival in office. Often, informal strategies of authoritarian survival, such as compromising on policy and broadening their search for allies, are pursued in tandem with formal strategies to stay in power, often through constitutional reform. Indeed, we further Nathan Brown's (2008) argument that ‘constitutional politics’ is really ‘normal politics’ as politicians seek to manipulate the public will for private gain. And in the realm of term-limit extensions, authoritarians do this with particular frequency and effectiveness. However, while this ‘menu of manipulation’ – prominently featuring constitutions – has been readily discussed, the degree of effectiveness of constitutional reform on authoritarian survival as perhaps the leading ‘menu’ strategy, has never been quantified. In this study, we offer first-hand empirical evidence showing the circumstances under which constitutional manipulations lead to extended authoritarian survival in office.

Hypothesis

Descriptive analysis (Figure 1) suggests a relationship between removing institutional barriers, particularly altering constitutional term limits, and authoritarian incumbents' longevity in office. The question that remains unanswered, however, is whether these constitutional manipulations do in fact increase the executive's longevity in office, or whether the incumbent's survival in office is a function of other factors such as power consolidation. Building on previous studies, we propose the following hypothesis:

Hypothesis: Altering constitutional term limits increases the incumbent's longevity in office.

As Valerie Bunce and Sharon Wolchik (2010) argue, constitutional change is one of the common strategies authoritarian regimes use to stay in power. Indeed, one of the first and most important steps for authoritarian-aspiring leaders to consolidate power is to remove any constitutional barriers to their rule. Altering constitutional term limits is the most straightforward way to achieve this end. Thus, our hypothesis predicts that removing constitutional term limits should extend the leaders' longevity in office.

Data and descriptive analysis

To test this hypothesis, we develop an original data set from several available databases. We use Hein Goemans et al.'s (2016) Archigos data set to record all political leaders of independent states from 1875 to 2015. Archigos records the dates of entry and exit of over 3,400 executives. Using Zachary Elkins, Tom Ginsburg and James Melton's (2014) Comparative Constitutional Project (CCP), we then code all constitutional changes made under these incumbents. The CCP data set quantifies the content of constitutions in all independent states from 1789 to 2013. Using the CCP data set, we code the number of constitutional change events for each incumbent (including new constitutions, amendments and interim constitutions). Next, we code whether those constitutional changes altered the executive term limits and, if they did, whether those changes removed or installed term limits.

The unit of analysis and observation in this study is incumbent political leaders of independent states from 1875 to 2015. Overall, we have 3,409 observations. Considering that the observations are individuals, our explanatory variables are not time varying. For those variables which naturally change over time, such as democracy score, we either use aggregated measures (i.e. means) or when more appropriate use the estimate at a particular point in the incumbent's tenure (e.g. level of power consolidation at the time of removing term limit). Below, we describe how we estimate each of these variables.

The outcome of interest in this study is the incumbent's risk of removal from office. We estimate the incumbent's survival in years using the Archigos data set, which defines the incumbent executive as the de facto political leader of the country, which might not be the same as the nominal political leader. This is relevant in countries with more than one titular head, such as in Russia from 2008 to 2012, where the de facto leader was Vladimir Putin and not President Dmitry Medvedev. We use all world leaders rather than only authoritarians to account for many ‘authoritarian-aspiring’ executives who ascend to power through democratic elections but consolidate power and turn the regime into an authoritarian one. Our data analysis shows that while 31% of constitutional term-limit alterations take place in authoritarian regimes, 23% of these alterations take place under democratic leaders.⁷ Moreover, among those leaders who specifically removed executive term limits from the constitution, 38% were authoritarians when they came to power, 14% were democrats and 48% were in hybrid regimes when entering power. Thus, it is imperative to include all leaders regardless of their regime type in our analysis. Nonetheless, we also separately estimate our main model for non-democratic and democratic leaders. The main predictors in our study are constitutional change and power consolidation, while we also control for several other factors, including system of government (presidential vs non-presidential) and regime type (authoritarian, hybrid, democracy).

Constitutional manipulations

For constitutional change, we create three measures using the CCP data set. First, ‘Constitutional Change Events’ is an ordered categorical variable (coded 0 if there is no constitutional change during the tenure of the political leader, 1 if there is one constitutional change, and 2 if there is more than one constitutional change event). Next, we create a binary variable, ‘Term-Limit Alteration’, if the ‘Constitutional Change Events’ is not 0 and if the constitutional change entails any alteration to executive term limits. This is to account for any manipulation of constitutional term limits which does not necessarily include lifting the term limit but is nonetheless aimed at extending the incumbent's longevity in power. For example, changing constitutional term limits from two terms to two consecutive terms does not entail term-limit removal, but allows the incumbent to stay in power by having a handpicked ‘caretaker’ to assume power nominally for four years, before the autocrat's formal comeback (e.g. Russia 2008). Lastly, we create three binary variables, including ‘Term-Limit Removal’, ‘Term-Limit Installation’ and ‘No Term-Limit Change’, to code the direction of term-limit alterations.⁸ We use all three binary variables together in our estimated models because these categories are not mutually exclusive. A case in point is President Abdelaziz Bouteflika, who changed the Algerian constitution three times during his 20-year rule. In 2002, he amended the constitution, without changing the term limits. In the 2006 amendments, however, he removed the term limits. And in 2016, he reinstalled the term limits but allowed himself to finish his term and run for another term (Maboudi 2019). As a result, Bouteflika is coded 1 for all three binary variables.

Our data show that a clear positive relationship exists between limiting terms and levels of democracy. As Figure 2 shows, the average Polity IV democracy score for countries where the constitution institutionalizes executive term limits is 6, on a scale of −10 (indicating full dictatorships) to 10 (indicating full democracies). The average democracy score is 2 when the constitution limits the executive to two consecutive term limits, which allows political leaders to run for more terms, but only after taking one term break. And if the constitution does not constrain the executive with term limits, the average democracy score is about −3.

Figure 2. Executive Term Limits and Level of Democracy

Besides the clear-cut categories shown in Figure 2, there are other forms of constitutional manipulation which are intended to increase the longevity of leaders. And while it is easier to code lifting or adding executive term limits in constitutions, it is much more difficult to code these other forms of constitutional change that are purposely aimed at extending a leader's tenure without necessarily abolishing term limits. For example, in 2019, the Egyptian parliament approved a bill to amend the constitution and to change the length of presidential terms from four to six years with two-term limits. This change would allow the current president Abdel Fattah el-Sisi potentially to remain in office until 2030, without abolishing the term limit. Similarly, Turkey in 2017 undertook a constitutional change from semi-presidential to presidential, allowing Recep Tayyip Erdoğan to remain in office until 2029 without lifting the term limit. Yet another recent example of extending executive tenure involves Uganda's Yoweri Kaguta Museveni, who amended the constitution in 2019 to remove the age limit of presidential candidates to keep him from ‘ageing out’.

From an empirical perspective, these authoritarian innovations pose a difficulty in conceptualizing and operationalizing constitutional changes that extend the incumbents' tenure in office. We account for these constitutional manipulations partly with our broad ‘Term-Limit Alteration’ variable, which captures cases of term-limit removal and term-limit manipulations (e.g. replacing a two-term limit with a limit on two consecutive terms). However, we acknowledge that there are other forms of constitutional manipulation allowing incumbents to extend their tenure in office, which do not involve term limits and we cannot code. For example, since not all extensions of term lengths or age limits allow the current leader to stay longer in office, we cannot code these instances as term-limit alterations.

Our study focuses specifically on altering executive term limits, which is the most direct, common and traditional form of constitutional manipulation in presidential regimes. We argue that ‘term-limit tampering’ is a leading indicator (and probably the most important such indicator) of institutional manipulation by authoritarians, but far from the only one. Still, we claim that if we can draw such strong patterns from this indicator alone (without even quantifying other means from Andreas Schedler's ‘menu of manipulation’), the patterns we identify would only be stronger if, as would be logical, other means of authoritarian perpetuation could also be coded.

Power consolidation

It can be argued that authoritarian leaders with a stable and strong base of support are able to amend the constitution more easily and stay longer in power. Indeed, high popularity of sitting presidents or power consolidation by presidents may be necessary preconditions for their attempts to change constitutions or use other strategies to stay in office beyond constitutional limits. In this sense, access to tools of constitutional change may not cause greater duration but will instead simply demonstrate the strength of the leader, which itself can be the cause of greater time in office. As such, we include different estimates of power consolidation in our model.

We use three proxy variables to measure power consolidation of the incumbent. Our first and most conventional proxy for power consolidation is the extent to which the incumbent is an established (as opposed to contested) autocrat. The degree to which autocratic leaders personalize their regimes and eliminate their rivals provides a valuable insight into power consolidation. For example, authoritarian leaders who have successfully personalized their regimes and concentrated power are less vulnerable to insider challenges, including coups (Grundholm 2020). As Milan Svolik (2012: 6) also argues, ‘established autocrats have acquired so much power that they can no longer be credibly threatened by their allies – they have effectively monopolized power’. By contrast, in contested autocracies ‘the allies are capable of using the threat of a rebellion to check the dictator's opportunism’ (Svolik 2012: 6). As a result, when established autocrats leave office, it is either through foreign occupation (e.g. Saddam Hussein), popular uprising (e.g. Muammar Qaddafi), or natural death (e.g. Joseph Stalin), none of which was at the hands of the ruler's inner circle (Svolik 2012: 7).

Following Svolik (2012), we define contested autocrats as those incumbents who have not consolidated enough power and as such can be removed by their inner circle. Thus, to operationalize power consolidation, we first construct a ‘Contested Leader’ variable based on three binary variables from the Shock Database (Miller 2021). The binary variables that we use measure whether the leader was removed from power through a civil war, assassination or a coup (all of which are indicators of an insiders’ rebellion). We then aggregate these variables and create the binary ‘Contested Leader’ variable, which is coded 1 if the leader was removed from office by any of these methods and 0 if not. While this is not a direct measure of power consolidation or contestation, removal from power by the ruler's inner circle is a clear and classic indicator of contested power distribution which is more likely to be seen in authoritarian regimes than in democratic regimes.⁹

Next, we use the age of the incumbent at the time of assuming power as the second proxy variable for power consolidation. In our data set, the average de facto leaders assumed power when they were 54 years old, with a standard deviation of 11 years. The youngest incumbent came to power at age 11 and the oldest leader was 88 years old upon ascending to power. It is indeed more difficult for very young and very old incumbents to consolidate power, eliminate rivals, change the constitution and stay longer in office. Thus, we created three age categories (11–35, 36–60 and 61–88), with the first age group as our base category. We expect that middle-aged incumbents have more opportunity to consolidate power and enjoy its benefits for their longevity in office.

Our third proxy for power consolidation is the timing of constitutional change. As Henry Bienen and Nicolas van de Walle (1991) show, time already in office is an important factor explaining the survival rate of dictators, indicating that, as the length of time a leader is in office increases, they are more likely to stay in power. We use the Archigos (2016) and CCP (2014) databases to count the time (in years) the leader was in power at the moment of constitutional change.

The ‘Timing of Constitutional Change’ variable is estimated at the time the executive decides to change the constitution. Since leaders who do not change their constitutions will be dropped, using this variable results in a significant number of missing observations; hence, we use this variable only as a robustness check. We expect that the longer the incumbent is in power before he or she alters the constitution, the more likely they are to survive longer in office. We acknowledge that none of these three proxies is an ideal estimate for power consolidation. We believe, however, that together these variables indicate degrees of power consolidation.

Control variables

As discussed earlier, leaders' decisions to alter their constitutional term limits as well as their survival rates are determined by several factors, including government systems and regime types, which we control for. As recent examples from Russia (2020), Egypt (2019) and Turkey (2017) show, autocratic leaders can manipulate constitutional rules in ways other than openly removing term limits to prolong their power. Consequently, we expect that as the number of constitutional changes in authoritarian and presidential regimes increases, the leader's survival in office increases too. We do not expect the same effect for hybrid regimes because when the opposition is strong (as in many hybrid regimes), constitutional reforms mostly function as the incumbent's outreach to the opposition for a requisite sharing of power, rather than as a tool for hoarding power and increasing survival in office. This is one explanation for the extant literature's finding that semi-authoritarian or hybrid regimes are less durable than closed authoritarian regimes (Brownlee 2009; Howard and Roessler 2006; Knutsen and Nygard 2012).

Previous studies also indicate that presidential systems are more vulnerable to constitutional manipulations when presidents control both the legislature and the judiciary or when they have personal interests (Baturo 2014), or when party fragmentation and shared control of constitution-making processes create uncertainty for future electoral wins by the incumbent party (McKie 2017). As such, we include two measures for government system (presidentialism vs non-presidentialism) and regime type (autocracy vs democracy) in our analysis.

We measure government systems using Cesi Cruz, Philip Keefer and Carlos Scartascini's (2018) Database of Political Institutions (DPI). Our binary ‘Government System’ variable is coded 1 (for presidential systems) when a de facto leader assumes power as a president or changes his or her title to president while in power. Subsequently, all non-presidential government systems are coded 0. For regime types, we use the Polity IV data set (Marshall and Jaggers 2013) to create three binary variables for ‘Democracies’ (+6 to +10 on Polity score), ‘Hybrids’ (+5 to −5 on Polity score), and ‘Authoritarians’ (−6 to −10 on Polity score). Since a leader's democracy score may vary during their tenure, rather than democracy score at entry or exit years, we use the mean of the Polity score over their tenure to create the regime type categories.¹⁰ Furthermore, since constitutional reforms for the purpose of extending longevity in office are more likely in authoritarian and hybrid regimes, ‘Democracies’ is taken as the base category in the estimated models.

We also include several other control variables accounting for domestic conflict, economic growth and regions. We control for domestic conflict because embattled incumbents are more prone to both constitutional changes and removal from power. The ‘Conflict’ variable is based on the weighted conflict index from Arthur Banks and Kenneth Wilson's (2016) Cross-National Time-Series Data Archive. This index estimates all major social, ethnic and civil conflicts, including assassinations, strikes, guerrilla warfare, government crises, purges, riots, revolutions and anti-government demonstrations. To account for the significance and different magnitude of each type of political conflict, Banks and Wilson (2016) weight them differently.¹¹

Since economic development fortifies authoritarian leaders, while economic crisis makes them vulnerable, we control for economic growth. The ‘Economic Growth’ variable is based on the GDP annual growth rate statistics from the World Bank's World Development Indicators (2013), and it measures the average growth rate during the tenure of the incumbent. In our sample, the variable ranges from an average of −62 to 59%. Finally, since evidence from several case studies indicates that the lifting of constitutional term limits is a regional phenomenon which is more often observed in regions such as sub-Saharan Africa and Latin America, we use dummy variables to control for world regions, with ‘West’ (including Western Europe, the USA, Canada, Australia and New Zealand) as the base category in the estimated models.

Survival analysis of executives

To test our hypothesis, we estimate a Cox proportional hazard survival model below, using the specified covariates. The equation for our base statistical model is

(1)$${\rm \lambda }_{\rm i}( {\rm t} ) = {\rm \lambda }_0( {\rm t} ) {\rm \;}e^{\beta _1Constitutional\;\;Change_i + \beta _2Power\;Consolidation_i + \beta _3Z_i}$$

where λ _i(t) is the hazard function at time (t). The λ₀(t) is the baseline hazard function at time (t) and corresponds to the probability of removal from office when all the explanatory variables are 0. As such, λ₀(t) is analogous to the intercept β ₀ in multiple regression, with an important difference: unlike intercepts in multiple linear regression which remain constant, here they change with time. The $e^\beta$ in this equation is the exponential of regression coefficient β and is the relative risk of removal from office for X = 1 relative to X = 0 for a binary explanatory variable. Finally, Z_i is a vector of covariates controlling for government system, regime type, levels of conflict, economic growth and regions.

The estimated Cox proportional hazard survival models are reported in Tables 1–3. The first column in these tables shows the coefficients and the second column shows the estimates of the hazard ratios. Negative coefficients indicate that increase in the predictor decreases the risk of removal from office. In other words, a negative coefficient shows longer executive tenure. Hazard ratios larger than 1.0 should be interpreted as increased odds of removal from office and values smaller than 1.0 as reduced odds. For example, a hazard ratio of 0.5 indicates that a one-unit increase in the explanatory variable halves the risk of removal from office, and a ratio of 2.0 indicates that a one-unit increase in the predictor doubles the risk of removal from office.

Table 1. Survival Analysis of the Executive

Notes: Robust standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 2. Survival Analysis of the Executive (The Impact of Term-Limit Removal)

Notes: Robust standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 3. Survival Analysis of the Executive (with Interaction Term)

Notes: Robust standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 1 shows the results of our base model which estimates the impact of constitutional changes (number of constitutional events and the term-limit alteration variables) and power consolidation (only Contested Leader and Entry Age variables) on the incumbent's survival in office. The results show that while constitutional change significantly reduces the risk of removal from office, power consolidation has a mixed effect. More specifically, the estimated model shows that altering constitutional term limits reduces the risk of removal from office by about 20%. Moreover, the results show that if the incumbent changes the constitution at least once, the risk of removal from office decreases by over 40%. If the incumbent changes the constitution more than one time, the risk of removal from office drops by almost one-third. The results, however, show that being a Contested Leader has no statistically significant relationship with the risk of removal from power. And compared to incumbents who come to power at an early age (under 35 years old), older incumbents (those who come to power at 61 years of age or older) have a higher risk of removal from office. The middle age group, however, is not statistically significant; that is, compared to the youngest group of rulers, the middle age group is not necessarily better positioned to survive longer in power.

Table 1 also shows that presidential systems have a 20% lower risk of removal and that, compared to democracies, authoritarian regimes have over 40% less risk of removal. The relationship between hybrid regimes and survival in office, however, is not statistically significant. The results also show that conflict increases the risk of removal from office, but the magnitude of this relationship is very small. Meanwhile GDP growth reduces such risk, indicating that leaders with better economic performance survive longer in office. Among the regional dummies, the results show that compared to the West, Eastern European and Latin American leaders face more uncertainty in office, while Middle Eastern and North African leaders are more stable, alluding to the contested nature of politics in the former regions compared to the Middle East and North Africa, where authoritarians are robust and remain mainly uncontested (see Bellin 2004).

Next, we use the three binary variables accounting for the content of term-limit alteration as the main variables for constitutional change. The results in Table 2 show that when incumbents specifically remove constitutional term limits, they reduce their risk of removal from office by over 40%. After adding the three binary measures of term-limit changes, the significance and direction for all other predictors and control variables remain very similar to Table 1, except for conflict variable which is no more statistically significant.

While our results show strong evidence for the effect of constitutional alterations on longevity in power, there is mixed evidence for the effect of power consolidation on survival. It could, however, be argued that the impact of constitutional change on survival in office is contingent upon power consolidation. That is, when authoritarians have consolidated their power, they are more likely to increase their survival through term-limit alteration.

To account for this potential contingency, we interacted the Term-Limit Alteration variable with the Contested Leader covariate. Normatively, we should expect that when contested leaders (those who fail to consolidate power beforehand) alter constitutional term limits, their risk of removal from office would increase. As Table 3 shows, while the interaction covariate is in the predicted direction, it is not statistically significant, meaning that contested leaders who alter their term limits do not necessarily have a higher risk of removal from office. In other words, the impact of constitutional manipulations on leaders' survival does not depend on the degree to which they consolidated or failed to consolidate power.

In sum, our findings provide empirical evidence supporting our hypothesis; that is, altering constitutional term limits increases the incumbent's longevity in office. The results show that constitutional changes are very important tools for rulers across different polities to avoid the risk of removal from office. The survival analysis, however, shows that power consolidation is not a strong predictor of longevity in office, and that the impact of constitutional manipulation on survival is not contingent upon power consolidation. Furthermore, the results indicate that not all political leaders benefit equally from constitutional reforms intended to increase their longevity in power. Our findings particularly show that authoritarian presidents benefit the most from periodic constitutional reforms and altering constitutional term limits. Taken together, the findings provide empirical support for our hypothesis, indicating that constitutional manipulations, especially by autocratic presidents, have significant implications for the survival of incumbents and for prospects of democracy in those nations.

Robustness checks

The empirical findings are robust and stand several robustness checks which are reported in the Online Appendix. Using our base model (Table 1) as a point of reference, we first ran the model without any control variables to ensure that the control variables are not driving the results. As Table A1 in the Online Appendix shows, the results for constitutional alteration do not change in our restricted model. That is, even without including control variables, constitutional alterations remain a statistically significant predictor of executives' longevity in office.

Next, we included the Timing of Constitutional Change variable and ran our base model. As discussed earlier, adding this variable drops almost 50% of our observations (i.e. leaders who did not change their constitutions). Nonetheless, after including the variable in our model, the direction and significance of most covariates remain very similar to the base model except for Contested and Government System variables, which are no longer significant, and Hybrid Regimes variable, which is now statistically significant. As Table A2 in the Online Appendix also shows, Timing of Constitutional Change has a negative and statistically significant correlation with the risk of removal from office. In other words, the longer the incumbents wait to change their constitutions, the more likely they survive longer in office.

In Table A3 in the Online Appendix, we substituted our three proxies for power consolidation with V-Dem's Presidentialism Index at the time of constitutional change. The Presidentialism Index should not be confused with presidential systems, for which we also control. V-Dem defines the Presidentialism Index as the extent to which a regime is characterized by presidentialism: that is, the ‘systemic concentration of political power in the hands of one individual who resists delegating all but the most trivial decision making tasks’ (Bratton and Van de Walle 1997: 63). Like Timing of Constitutional Change, using this variable comes with the caveat of losing observations, since both variables are estimated only for cases with constitutional change. As Table A3 shows, Presidentialism Index is statistically correlated with lower risk of removal from office. That is, those leaders who systematically concentrated political powers in their own hands survived longer in office. Using Presidentialism Index, however, does not significantly change the results of other covariates, including constitutional manipulations.

As a robustness check, we also interacted entry age of leaders with constitutional change. Table A4 in the Online Appendix shows that none of the interactions is statistically significant, confirming our initial finding that the impact of constitutional manipulations on survival is not contingent upon power consolidation. It is also important to note that when we include entry age interaction terms in our model, Term-Limit Alteration loses its significance, but Constitutional Change Event remains statistically significant. We find a similar effect when we interact Presidentialism Index with Term-Limit Alteration (Online Appendix, Table A5).

As discussed above, we also divided the data set into democratic and non-democratic (including hybrid) regimes to examine the effects of our predictors for each of these regime types. As the results in Table A6 in the Online Appendix show, while the relationship between term-limit alteration and survival in office remains statistically significant for non-democratic regimes, it is not statistically significant for democracies.

We also controlled for Barbara Geddes, Joseph Wright and Erica Frantz's (2014) regime types, setting democratic regimes as the base category. Table A7 in the Online Appendix shows that when we control for Geddes et al.'s regime types, the results do not change for our main predictors. Most importantly, constitutional alterations still increase the survival rate of executives. The results also show that, as Geddes and her colleagues predict, compared to heads of executive in democracies, incumbents in monarchies and party regimes have a higher survival rate. However, our results show that personalist leaders have a lower survival rate than democratic leaders, which is counterintuitive. A potential explanation for this surprising finding is the sample size, as there are 762 democracies in our data set compared to 72 personalist regimes.¹² Lastly, military incumbents have a higher risk of removal from office (lower survival rate) compared to democratic leaders, but this relationship is not statistically significant.

We also ran restricted models for each of Geddes et al.'s main autocratic regime types (personalist, party and military) to see whether the effects of altering term limits vary across these regime types. As Tables A8–A10 in the Online Appendix show, altering term limits more than halves the risk of removal from office in party regimes, but almost triples that risk in military regimes, and does not have any statistically significant impact in personalist regimes.

Lastly, it might be argued that since presidential term limits are one of the indicators in Polity IV scoring, Polity might not be a valid measure for democracy–autocracy regime types in our study. We should note, however, that as the presence of term limits is only one indicator in the Polity IV, and as we do not use the Polity IV data as a time-varying variable, there probably is not substantial bias. Nonetheless, we recoded Geddes et al.'s (2014) regime types to create a binary variable, ‘Autocracy’ (coded 0 for democracies and 1 for all types of autocracies), and ran our base model (Table 1) with this variable instead. As Table A11 in the Online Appendix shows, while the Autocracy variable is in the predicted direction (i.e. autocratic leaders have higher survival rates), the relationship is not statistically significant. More importantly, the results show that using a different regime type variable does not change the results for our main predictors.

In sum, the statistical analysis in this study demonstrates that constitutional changes in non-democracies are more likely to increase the survival of incumbents compared to democracies. And not surprisingly, authoritarian leaders are aware of the rewards that constitutional changes offer. Although there are cases where such undemocratic aspirations can result in political crisis and even ouster of the incumbent (e.g. Honduras 2009, Bolivia 2019) or international backlash (Hungary 2013), constitutional changes intended to create executive continuity most often endanger democracy or democratic prospects (Bunce and Wolchik 2010).

When constitutional reform is normatively bad

In his swearing-in ceremony address in 1986, Ugandan president Yoweri Museveni asserted that ‘the problem in Africa in general and in Uganda in particular is not the people but leaders who want to overstay in power’ (Tangri 2006). Ironically, after 34 years of ruling Uganda, the same president removed the last legal barrier (the constitutional age limit) to his presidency-for-life in 2018. Uganda's septuagenarian Museveni is among a long list of rulers who have attempted to extend their tenure through constitutional manipulation, although some were less shameless in their self-dealing.

Our analysis shows that authoritarians who reform constitutions, especially those who execute multiple reforms, stay in office 2.5 times longer than democrats. This is an unmistakable and stark finding. These reforms were intended to increase authoritarians' discretionary power, and their stay in power, by disabling career-ending term limits. Such survival strategies are not surprising to scholars who assume politicians seek to stay in power, especially considering the vast institutional manipulation toolkits of authoritarians, including electoral and legislative manipulations that are used to eliminate their opponents and to extend their rule (Gandhi 2008; Hale 2011).

Our findings, however, run against a widely held proposition in the comparative authoritarianism literature which contends that institutional constraints in general help the survival of authoritarian leaders by signalling credible commitment to the rulers' allies. As this literature suggests, institutional manipulations including constitutional reform projects intended to remove executive constraints are more costly in contested authoritarian regimes where the ruling elites can potentially punish fellow autocrats who seek to change the rules of the game and the balance of power for their own personal benefit (see Svolik 2012). By contrast, our analysis demonstrates that – regardless of levels of power consolidation and contestation – removal of constitutional term-limit barriers often helps autocrats survive longer in power.

Among our most important findings are that while ‘macro’-level distinctions such as authoritarianism, presidential systems and constitutional term-limit removals were significant, level of contestation of the leader was only marginally significant in some models, meaning that the second-order discussion about whether authoritarian incumbents sought to institutionalize rule or de-institutionalize it (i.e. the debate between those arguing for contested authoritarians vs those assuming consolidated authoritarians) was less important. The overwhelming story of this analysis is that of de-institutionalization through the reduction of term limits. In other words, authoritarians maximize power and enshrine it in constitutional reforms, so it appears systematic rather than arbitrary, regardless of whether this de-institutionalizes authoritarian regimes (under a legalistic guise).

Our findings also highlight the potential dangers of constitutional reforms in both democratic and non-democratic settings, as they can undermine executive constraints and perpetuate autocratic rules. While further research is needed to establish which portion of constitutional reforms are driven at least in part by incumbent executive self-interest, and which are driven by other causes, we believe constitutional revisions are increasingly driven by executive self-interest rather than collective interests.

Constitutions, once held in great public esteem as the blueprints of a peoples' collective will, have been reimagined as the mundane fingerprints of one person's self-interest. Wily executives are not the only ones steering constitutional reforms, which may originate in legislative bodies or constituent assemblies and include provisions governing the broadest range of issues. But we suspect, subject to further research, that incumbent executives, like Bouteflika, Erdoğan, Menem, Museveni, Putin and scores of others, author or at least strongly advocate for these provisions. The question that remains is how might we incentivize politicians to stop deferring the limits to their terms and instead be forced by the citizenry to commit credibly to constitutional reforms to help address societal injustices and inequalities, rather than trying to stay just a little bit longer, as they seemed to do more frequently in halcyon days?