Protecting democracy from abroad: democracy aid against attempts to circumvent presidential term limits

ABSTRACT The article addresses the question of whether international democracy aid helps to protect presidential term limits – a commonly accepted but increasingly challenged safeguard for democracy. According to our analysis, democracy aid is effective in countering attempts to circumvent term limits, thus, it contributed towards protecting democratic standards in African and Latin American countries between 1990 and 2014. Democracy aid helps to fend off term-limit circumventions, but it is not as effective in deterring presidents from trying to circumvent presidential term limits. Our analysis furthermore suggests that there is double the risk of an attempt to circumvent term limits in Latin American than in African states. Although our results confirm prior findings that “targeted aid” such as democracy aid makes a difference for maintaining democratic institutions, it challenges studies that argue democracy assistance has become “tame.” Our findings furthermore support previous indications that more refined theories on the effects of democracy aid in different phases of domestic processes are necessary, in particular in the face of global autocratization trends.


Introduction
Limiting the terms of heads of state is an institutional safeguard for democracy. These limits shall prevent abuse and the extension of executive power in political regimes, 1 especially in (semi-)presidential systems. 2 In the context of recent global autocratization trends, circumventing presidential term limits has become one part of the toolbox of pseudo-democrats who seek to extend their political power. 3 During the last decade, incumbents in Africa and Latin Americatwo regions with predominantly presidential systemsincreasingly attempted to extend their mandates beyond the term limits foreseen in the respective constitutions. In most cases, they opted for legal reforms such as constitutional changes or re-interpretations of legal norms to extend term limits. 4 African and Latin American presidents attempted to circumvent roughly every second instated term-limit rule between 1990 and 2014; they successfully circumvented 30 term limits and remained in power for a third (or more) term.
This article sets out to better understand this widespread practice of expanding executive powers and to explain if international democracy aid may counter term extensions. Despite the increase in literature on the driving forces of attempts to extend term limits in comparative politics, studies have acknowledged international factors but have hardly studied them in depth. Analysed factors span from the relevance of specific actor groups, such as the military or political parties, 5 to institutional arrangements such as government capacities or the relationship between the legislature and executive, 6 to political corruption. 7 Although support to promote democracy has become a standard element of the foreign policies of "Western" countries, 8 we hardly know anything about the effects of democracy aid on protecting democracies from autocratization trends. More specifically, empirical evidence on democracy aid's influence on the (un)successful outcomes of attempts to circumvent presidential term limits is still scarce and limited to country case studies 9 and one cross-national study of sub-Saharan Africa. 10 Our analysis addresses this gap by providing a comprehensive cross-national and inter-regional study by answering the following research question: Did democracy aid help to protect presidential term limits in Africa and Latin America between 1990 and2014? We approach this question by splitting it into two further questions, which do justice to the two typical steps of circumventions of presidential term limits. Each step has different implications for international actors' involvement, as we outline below.
(1) Does democracy aid help to deter incumbents from trying to circumvent term limits? (2) Does democracy aid help to fend off term-limit circumventions?
Theoretically, this article contributes towards refining the explanations of the role of democracy aid in the "tug-of-war" between democratic consolidation and autocratic backsliding by integrating two strands of literature. It draws, first, on International Relations theories of foreign aid and democracy promotion, which focus on international effects on domestic political processes. We combine this literature with the growing research on presidential term limits in developing regions. Conceptually, it introduces a temporal dimension by systematically distinguishing two steps in the sequence of attempts to circumvent term limits. This distinction allows us to identify the typical dynamics, which have different implications for the strategies applied by donors of democracy aid. Empirically, this contributes generalizable empirical evidence on democracy aid's influence on term limits because it provides the first quantitative analysis with a cross-regional sample. The analysis includes 49 attempts to extend presidential term limits between 1990 and 2014 in Africa and Latin America. It combines secondary data with primary data collected by the authors. Methodologically, this article applies a survival analysis, which is able to model the "lifespans" of term limits, and thus, the resistance of democratic institutions. This allows us to also account for censored and truncated data as well as the duration dependence of term limits.
In the remainder of this paper, we first introduce the key concepts and build hypotheses in part two. They are based on socialization and rational choice-based theories, which link democracy aid to the likeliness of success or failure of attempts to extend presidential term limits. In the third part of this article, we explore the advantages and limits of survival analysis to address the research question and introduce the dataset built for the empirical analysis. In the fourth part, we discuss the results. We find that democracy aid is effective in countering attempts to circumvent term limits, thus, it contributed towards protecting democratization in African and Latin American countries between 1990 and 2014. We find consistent results that democracy aid lowers the risk of there being a successful circumvention, whereas the effect of democracy aid on the risk that presidents might attempt to circumvent term limits is somewhat uncertain. The article concludes with a summary of the findings and an outlook on future research to enhance theory-building on democracy protection.

Theoretical framework and hypotheses: protecting term limits with democracy aid
Democracy aid as a targeted form of foreign aid has been proven to enhance the democratic quality of political regimes. 11 Whereas most scholars focused on supporting the establishment and consolidation of democratic institutions and behaviours, analyses paid less attention to protecting democracy from anti-democratic influences. 12 However, in light of global autocratization trends, it is relevant to know if past efforts to support democracy helped to counter anti-democratic forces. This empirical analysis contributes towards closing this research gap by analysing whether democracy aid can help to protect one of democracy's most relevant "safeguards"presidential term limits. Although "pseudo-democrats" carefully weight their probabilities of "overstaying," presidents frequently fail to extend their terms: 38% of circumvention attempts were unsuccessful between 2000 and 2018. 13 Applying a broad understanding of presidential term limits, we define them as "a constitutional restriction on the number of fixed termsconsecutive or otherwisethe head of state may serve." 14 Term limits are thus a democratic norm that serves to limit executive political power. Term-limit circumventions describe instances in which term limits are either extended or abolished. Conceptually, we differentiate two stages in the process of term-limit circumventions. First is the initial attempt to circumvent a presidential term limit, which is the moment an incumbent resolves and proceeds to extend or abolish a term limit. Second, the de facto outcome of an attempt to circumvent term limits can be successful or unsuccessful. Staying in office through either electoral or non-electoral means signifies a successful outcome, while leaving office implies the contrary for the incumbent. In such processes of circumvention, successful democracy aid should work towards maintaining term limits and electoral governmental change.
Democracy aid targets core political institutions and actors, especially elections and electoral commissions, pro-democratic civil society groups, parliaments, media organizations, the judiciary, and human rights commissions. 15 It supports institutional reforms such as electoral management or parliamentarian oversight to improve accountability. To foster and implement these reforms, it addresses the behavioural and attitudinal changes of officeholders by providing advice and trainings. 16 In addition, it supports pro-democratic actors such as activists, advocacy groups, social movements, and political leaders by providing trainings and organizational funding. Timing matters in democracy aid. It directs aid at gradual change processes such as institutional and legal reforms and their implementation as well as flexible assistance during critical junctures, such as instances of regime change and democratic transitions. Actor-centred approaches become more relevant during critical junctures, as change-makers are decisive for creating a new political order, for instance through social mobilization and constitution-building. 17 Although donors of democracy aid have established "toolboxes" to support democratization, they still lack explicit approaches to counter gradual autocratization processes. Regional organizations with democracy clauses are an exceptionthey are purported to serve as safeguards for democracy in their member states, particularly in Europe, Latin America, and Africa. 18 However, there are several reasons to suggest that the democracy aid of bilateral and multilateral donors contributes towards protecting presidential term limits, and hence democracy.
Democracy aid works along two mechanisms of political action that are usually complementary in practice. First, following the socialization-based logic of appropriateness, 19 democracy aid fosters learning processes, which contribute towards institutional reforms and behavioural and attitudinal changes through civic education, seminars, capacity-building, and persuasion by arguing. 20 Such activities can influence the direction of a circumvention attempt. Where democracy aid builds capacities to foster vertical and horizontal accountabilitysuch as investments in a strong and lively civil society or a knowledgeable and critical parliamentit prepares oppositional actors to counter anti-democratic reforms such as attempts to circumvent term limits. 21 Persuading the political establishment that supports an incumbent's attempt to circumvent the term limit is a first step towards discretely signalling a donor's disapproval of the intended institutional reform. However, arguing with a pseudo-democrat who has decided to stay in power at all costs is certainly not the most promising approach. Here, the second mechanism of democracy aid comes in. 22 A rational choice-based logic of consequences foresees that democracy aid goes along with costs and benefits for sitting governments and the political establishment. 23 According to this logic, target states receive financial and reputational rewards from donors if they foster democratization; or, in the opposing scenario, they are punished with financial conditionality or sanctions if they counter democratization or reverse democratic achievements such as term limits. Then why should pseudo-democrats accept democracy aid at all? Because they can hardly avoid receiving democracy aid completely since it is part of a broader political economy of aid. Incumbents of target countries rely on aid and consider the costs of non-democratic behaviour in their relationships with "Western" donors. Furthermore, donors often integrate democracy aid in their aid negotiations with sitting governments. 24 Such aid negotiations are important communication channels and often set standards for democratic conditionalities. The latter are enforced once an agreed democratic reform fails or if sitting governments counter democratization. 25 Although arguing and persuasion are "toothless," they can be important complements once conditionality is factored in.
In line with previous research, we acknowledge that the democracy level of the target regime as well as its political-historical relation with donors influences both strategic donor allocation as well as the effectiveness of democracy aid. 26 For instance, if there were no, or very weak, civic or parliamentarian opposition or institutional safeguards against a term-limit circumvention attempt, democracy aid would lack addressees. It is furthermore less likely that donors would fiercely support opposition actors or apply sanctions or conditionalities if a target state has strategic importance or aligns ideologically with the donor.
Conceptually, we take a procedural perspective and distinguish two steps in the process of circumvention attempts, namely the attempt (step 1) and the attempt's outcome (step 2), as depicted in Figure 1.
Our approach is a refinement of research by Dietrich and Wright (2013), which only focused on democracy aid's influence on actual term-limit circumventions. 27 However, distinguishing between circumvention attempts and successful circumventions offers additional insight into democracy aid effectiveness. It indicates the relevance of timing and whether democracy aid helps in preventing presidents from attempting circumventions at all, or whether it contributes towards protecting term limits as soon as presidents make their attempt.
Despite differences between the two steps described below, they have an important commonality. They are driven by a simple actor constellation, which consists of proand contra-reform groups. 28 Although motivations within each group might differ from case to case, each side's objectives are generalizable across cases. On the one hand, the contra-reform group objects to the reform of term limits and the incumbent's continuation in office. On the other hand, pro-reform actors support the incumbent's attempt to stay in power. Donors who provide democracy aid are part of the contra-reform group. Our analysis covers all sources of democracy aid, spanning from donor countries and international organizations to foundations (see also Section 3, "Method, data, and operationalization").

Attempts to circumvent presidential term limits and democracy aid
In a first step, incumbents announce their intentions to either run another time for office, notwithstanding other legal provisions, or to reform the term-limit provision. We identify these attempts by assessing whether incumbents take formal measures to prolong their rule, despite a constitutionally prescribed term limit; or by checking whether sufficient consensual information exists in the media and the research literature that an incumbent intends to do so. The latter becomes important only in a few cases in which incumbents face so much opposition that they stop their attempt early on. 29 Prior to any such official attempt at a term-limit reform, however, the incumbent will assess their chances of success. If the continuous democracy support provided in the past has bolstered democracy sufficientlyby, for instance, increasing the capacity and power of civil society, the media, the political opposition, the parliament, or the judiciarythe incumbent will already have experienced headwinds in prior years. They will therefore judge their chances for success to be slim and refrain from making any official attempt at reform. Furthermore, they might anticipate their weak or strong position vis-à-vis donors and their foreign policy interests. If donors have vested foreign policy interests that the incumbent can use as political leverage, such as regional security interests, the incumbent would judge it to be unlikely that donors will intervene. 30 If, however, donors have already intervened or exerted pressure in the name of democracy protection in the past, or ifby providing a lot of democracy aidthey have signalled that democracy is dear to them, they would draw the opposite conclusion.
This reasoning leads to our first hypothesis: Hypothesis 1: The more democracy aid a recipient country receives, the lower the risk that an incumbent will attempt to circumvent a term limit.

Outcomes of presidential term-limit circumventions and democracy aid
Donors of democracy aid face wider latitude in opposing a term extension more openly when the incumbent has kicked-off an institutional reform process officially and the manipulation of the term-limit provision has become transparent. They invest in democracy assistance to enhance the capacities of social movements and civil society organizations to influence the reform path and offer legal advice to shift the contents of institutional reforms in a democratic direction. They are more likely to apply a logic of consequences such as rhetorical condemnations and aid conditionalities than before. 31 When incumbents nonetheless succeed in enforcing a new provision, they stand officially for another term in office in national elections and run electoral campaigns. After this second step, the objective of the anti-reform group changes. Opposition groups ally to prevent the incumbent's electoral success. Belonging to the anti-reform group, donors of democracy aid change their strategies, too, and focus more on electoral support to foster clean and fair elections. This is an important contribution for legitimizing the electoral process, 32 particularly if the incumbent eventually fails. Yet, donors of democracy aid also need to take care not to help in legitimizing an incumbent who stays in power. Supporting the electoral campaigns of nonstate actors and the political opposition to sensitize voters to the idea that yet another term for the same incumbent is not legitimate democratically can flip the coin during the electoral process. For illustration, consider the case of Senegal. The incumbent president, Abdoulaye Wade, reaffirmed in 2011 an earlier announcement he had made to run a third time for president in the upcoming February 2012 elections, despite a twoterm limit. Since the constitutional court green-lighted his candidacy, Wade indeed was able to run for president but was eventually defeated. 33 Throughout this episode, donors increased support to the domestic civil society and media opposing his attempt. 34 For instance, in 2011 the United States funded the Senegalese civil society organization RADDHO (Rencontre Africaine Pour la Defense des Droits de l'Homme)a key player in the social mobilization against Wade's third termthrough the National Endowment for Democracy (NED) with the following goal: To promote a credible election process in February 2012, RADDHO will promote the constitution among ordinary Senegalese and generate constructive dialogue on its provisions. RADDHO will also encourage participation among eligible voters in the elections through a voter education campaign and election observation. The campaign will include eight conferences in high schools and universities, 42 radio programs, one debate to air on national television, and a capacity building workshop for supporters of the movement against constitutional changes to presidential term limits. 35 Donors complemented this with pressure on Wade to refrain from his course of action, both publicly as well as behind closed doors. This leads to our second hypothesis: Hypothesis 2: Democracy aid helps to protect democracy by reducing the risk that an incumbent can successfully circumvent a presidential term limit.

Method, data, and operationalization
Our main interest is in the resistance of democratic institutions, or more specifically, how long a term-limit rule can survive without being extended or abolished. Additionally, we want to know whether particular incumbents remain in office following their attempts to change the term limit. Hence, our unit of analysis consists of two elements the actual term-limit stipulation in the constitution and the extension of the incumbent's mandate. This novel approach adds value to previous studies, which use the incumbent presidents as the unit of analysis because it combines an institutional and behavioural perspective. We investigate the hypotheses by fitting a survival model to our data. 36 Survival models estimate the riskmore specifically the "hazard"of an event during the course of "survival" of the observed unit. The unit of analysis "fails to survive" when the failure event occurs. We are interested in circumventions as well as attempted circumventions of presidential term limits, which, in methodological terms, are the failure events in our data. Our independent variable, the hazard rate, reports the rate at which each of these failure events takes place. The hazard rate reports the risk that term limits, which have survived without circumvention or attempted circumvention so far, face in each additional year of their "lifespan." An important advantage of survival models is that they take "left-truncation" and "right-censoring" into account. Left-truncation occurs when the onset of a unit of analysis happens before the time period of analysis. Right-censoring, in contrast, describes when a unit of analysis survives until or beyond the end of the analysis period. In our case, all term limit-rules that were already instated pre-1990 are left-truncated; all term-limit rules that were not circumvented up to 2014 are right-censored. Another advantage is that we can stratify our data according to the sequence of the term limits instated after one another in one state, which allows for discriminating between sequentially related risks for circumvention (attempts).
The start of the analysis period, 1990, is a cut-off point for studying democracy aid. After the Cold War, the dynamics of international relations and, hence, the practice of providing foreign aid changed. 37 Democracy aid became a common element of official development assistance (ODA). 38 Whereas in most African states democratic transitions and the introduction of presidential term limits took place post-Cold War, many Latin American states already had term limits in place or (re-)introduced them during the 1980s. 39 Regarding the endpoint of the analysis, 2014, we face the limitation that data for our main explanatory variable (democracy aid) are only available until 2013. 40 We collected data on the term limits of 63 African and Latin American countries during our period of analysis, resulting in 1380 country-years as units of observation. Data were drawn from the Constitute Project, 41 which presents constitutions collected online and by media outlets. We coded 93 term-limit rules for the period 1990-2014. 42 Term limits that were circumvented and then reinstated do not count as one term limit, as they face different risks of (attempts of) circumvention. Hence, we operationalize them as separate units of analysis. In order to keep the risks between these repeated term-limit rules separate, we specify a conditional gap time Cox model: Where h(t) describes the hazard rate, k describes three strata of whether the term limit is the first, second, or third and above term limit in country j, and x represents a vector of explanatory variables. Hence, we stratify the model by term limit, and cluster by country. 43 We run two model specifications, each with a different dependent variable. In our first model, we estimate the hazard that an attempt to circumvent a term limit is made, regardless of its result. In our second model, we estimate the hazard that a term limit is actually successfully circumvented. Each time, the sample includes all term limits in African and Latin American presidential systems between 1990 and 2014. Hence, the sample for the second model includes not only term limits that presidents attempted to circumvent, but also term limits for which no circumvention attempt was made. In operationalizing the attempts of term-limit circumventions, our coding follows the criteria introduced by Versteeg et al. 44 They identify five term-limit circumvention strategies: . Amending the constitution, the most common strategy . Putting in place a new constitution without term limits . Using the judiciary to re-interpret a term limit . Delaying elections . Making a placeholder person the new president.
We use all except the last strategy in operationalizing term-limit circumventions, as in this strategy a new person takes over the presidency de jure.
To take account of different steps in term limit circumventions, we define a successful circumvention as one in which the incumbent actually enters the next presidential term. Hence, cases in which an incumbent has secured for himself the opportunity to run for president again but was eventually electorally defeated, as in Senegal in 2012, do not count as successful circumventions in our data. We observed 49 attempts to circumvent term limits between 1990 and 2014, of which 30 were successful.
Our interest is in the effect of democracy aid on the risk that presidents might attempt to circumvent term limits as well as on successful circumventions. We take data on democracy aid from AidData. 45 AidData provides the most comprehensive project-level data of foreign aid, and it is commonly used in research on democracy aid. 46 We operationalize democracy aid as a four-year moving average to smooth out year-to-year fluctuations. This also accounts for the long foreign aid disbursement schedules. Additionally, we adjust democracy aid by population to gauge its relative magnitude and use a squareroot transformation to account for large values (see Appendix A2 for information). 47 We include a set of covariates that we regard as a sufficient minimum to block any confounding between democracy aid and term-limit circumventions. Hence, we do not include covariates affecting term-limit circumventions that are otherwise unrelated to democracy aid. 48 First, to account for endogenous democratic development not caused by democracy aid, we include the recipient state's democracy level using V-Dem's Liberal Democracy Index. 49 Second, as donors allocate democracy aid strategically according to their foreign policy interests and the recipient's democracy level, democracy aid and term-limit circumventions could be associated through a backdoor pathway. Including no covariates that capture donors' strategic allocation of democracy aid would create endogeneity through omitted variable bias. 50 Recent research shows that democracy aid allocation, at least regarding the United States, depends on the extent of democratization in the recipient state, its strategic importance, and its ideological alignment with donors' foreign policy. 51 To account for the extent of democratization, we apply V-Dem's Liberal Democracy Index one year prior to the four-year moving average of democracy aid, hence it is lagged by five years. To proxy the recipient states' strategic importance for donors, we count the security alliances between the respective recipient state and its donors in each year in the Correlates of War data. 52 To measure the ideological alignment between the recipient and donor states, we include a covariate that reports the mean difference between the recipient and its donors' positions towards the liberal world order based on United Nations General Assembly voting. 53 Third, most democracy aid is part of ODA. Although prior analyses have not found a clear association between ODA and term-limit circumventions, 54 some nevertheless suggest a relation between the two. 55 Donors might use ODA commitments for political conditionality, undermining the political support that an incumbent attempting to circumvent a term limit has. Therefore, we include a square-root transformed fouryear moving average of non-democracy aid ODA per capita taken from AidData. 56 Fourth, we include a count of previously failed circumvention attempts. Through its stratification, our model takes into account that a term limit has a different risk if a previous term limit has been circumvented successfully. 57 To account for a similar effect of unsuccessful attempts, we include the number of previously failed attempts. In total, we count 21 failed attempts.
Finally, we include a regional dummy that reports zero for Africa and 1 for Latin America. This allows us to identify regional differences. It is practice to lag timevarying variables in survival analysis models to ensure that the explanatory variable's change actually happens before the failure event. 58 Therefore, all time-varying variables are lagged by one year.

Empirical analysis and interpretation of the effects of democracy aid on protecting term limits
This is the first analysis, to the authors' best knowledge, that distinguishes an explanation for mere attempts to circumvent term limits as well as successful circumventions. Overall, democracy aid helps in protecting term limits. This effect is stronger and statistically more significant for actual term-limit circumventions (model 2 in Table 1) than for whether incumbents make attempts (model 1 in Table 1). To probe these results, we carry out a number of robustness tests, which are presented in the Appendix. 59 In the following interpretation of the empirical results, we focus only on the influence of Democracy Aid and Region because, first, our model is constructed in such a way so as to only assess hypotheses on democracy aid, whereas the additional covariates serve to isolate the relationship between democracy aid and term-limit circumventions; and because, second, the regional dummy variable nonetheless indicates an interesting side finding that flags the need for additional research on interregional differences.
Focusing first on attempts to circumvent term limits (model 1), we find that democracy aid has a negative effect on the hazard for term-limit circumvention attempts. An increase in democracy aid by one standard deviation (i.e. an increase from no democracy aid at all to $2.50 per capita) is associated with a 20% decrease in the hazard rate. However, the standard error for this association is comparably large, so one cannot put too much confidence into this result. Although it suggests that democracy aid more often than not helps to protect democracy by reducing the risk of an incumbent's attempt to formally prolong their term (hypothesis 1), there is a need to analyse further the varying circumstances creating the uncertainty about the effectiveness of democracy protection in this regard.
With regard to the differences between Latin America and Africa, our model suggests that being a Latin American state raises the hazard for an attempt to circumvent term limits by a factor of 1.8, a side finding that we evaluate in more depth later. Turning to our second dependent variable, we find that democracy aid lowers the risk of actual successful term-limit circumventions (model 2). The negative effect of more democracy aid conditional on all other covariates in our model is statistically significant and substantial in size. In states where international actors spend on average $2.50 per capita in democracy aid during a four-year period, the risk that the incumbent de facto circumvents the presidential term limit in the fifth year is only about half as that of a state that received no democracy aid. This stronger effect of democracy aid on the outcome of term-limit circumventions is plausible because of the process' particularities once an attempt has officially been made (Figure 1, outcome). In such instances, donors of democracy aid, who oppose changes in the term-limit rule, face less ambiguity and have more direct entry points for their assistance. Anti-reform groups built alliances to contest the incumbent and the irregularities during institutional reform processes. Donors can align with these oppositional forces, support them, launch additional initiatives, and pressure the incumbent more legitimately.
With regard to the regional dummy variable, the risk for term-limit circumventions in Latin American states is twice as great as in African states when taking into account the effects of all other covariates included in our model. This sizeable effect is in line with the results of the first model investigating the hazard for attempts only, and it is also corroborated by robustness checks. Finding this difference between the two regions prompts the need for future exploration and analysis. The finding may reflect the antagonistic interplay between the historical evolution of rigid term-limit regimes and strong presidentialism in Latin America. Throughout our period of analysis, the proportion of Latin American states with term limits is quite high at about 90%, and thus it exceeds the proportion of African states by about 20 percentage points. This finding seems to be driven by the particular historical path dependence in Latin America. Presidential term limits played an important role much earlier than in Africa, where they were introduced mainly during the third wave of democratization at the end of the 1980s. Historical experiences with caudillismo in the 19th and the first half of the twentieth century led to the evolution of rigid term-limit regimes during the second half of the twentieth century up until the 1990s, when the pendulum swung back and the "constitutional adjustment" of term limits led to an extension of presidential tenures in many states. Tenure constraints on the executive were often deemed a means to remedy weak legislatures and judiciaries that provided only ineffective checks on executive power. 60 Up to the start of our period of analysis, the great majority of Latin American countries had quite rigid term-limit rules in place, often allowing for no immediate re-election. Starting in the early mid-1990s, many Latin American states then transitioned to regimes allowing one subsequent re-election. Prominent examples include, for instance, Peru (Fujimori in 1993), Brazil (Cardoso in 1996), and Colombia (Uribe in 2006. Some scholars therefore argue that these are examples of "constitutional adjustments" rather than presidential power-grabbing, as in these cases term limits rendered presidential tenures too short for effective policy-making, and extensions therefore presented a solution for "short-termism." 61 However, this does not mean that many Latin American states do not feature strong executives and "presidential hegemony." 62 Starting from the mid-1990s onwards, many presidents who extended their tenures in office by one term also attemptedor at least flirted witha third term later (e.g. Fujimori). An increase in the proportion of Latin American states that abolished term limits entirelystates where presidents can hence be re-elected without limitation since about 2009 (starting with Chavez in Venezuela)can be seen simultaneously as an indication and consequence of increases in presidential hegemony.
In sum, the empirical analysis shows that democracy aid helps in protecting term limits. Its influence on reducing the risk of attempts being made to circumvent term limits, however, is smaller than its influence on reducing the risk of an actual circumvention. The latter finding is statistically also more certain than the former. To showcase the practical implications for the effect of democracy aid on successful term-limit circumventions, we estimated the survivor functions of term limits for a plausible country case (Figure 2). This plausible country case represents an electoral democracy that receives a minimum amount of ODA and is politically rather close to its donors, but it is not allied with any of them and has not experienced an attempt to circumvent a term limit before. 63 We plot three different democracy aid scenarios in which the state receives (1) zero, (2) $1.50 per capita, and (3) $27 per capita in democracy aid (please see Appendix B for our choice of values). Defining the median "life expectancy" of term limits as the point in time until which for each curve 50% of the term limits still survive, we see that a small democracy aid amount of $1.50 extends the "life expectancy" by only a few years, whereas a large amount of $27 extends it by 23 years. Conditional on a medium to large amount of democracy aid, a much larger estimated proportion of term limits survives at any point in time compared to the other two functions. The shaded and overlapping confidence intervals caution against taking these estimates Note: Estimated survivor functions for term limits with 95% confidence intervals. Dependent variable is successful circumventions with no prior attempts. All covariates are held constant at the 25th percentile value except for Region, Failed bid, and No. of alliances (constant at zero), while Democracy aid varies as shown in the figure. To retrieve the log-log transformed confidence intervals, the model's standard errors were estimated using the observed information matrix instead of clustered sandwich estimation, which has implications for the independence of observations. Please see Appendix A for a presentation of the model as well as separately presented survival functions and confidence intervals. too literally. Due to the limited nature of our datawhich provides only 30 failures for the successful circumvention model, and 49 failures for the attempt modelmuch statistical uncertainty accompanies the estimated survivor functions. The analysis shows nonetheless that especially a medium to large amount of democracy aid can make quite a difference.

Conclusions
The article investigated whether democracy aid helps to protect presidential term limits. This question is relevant, as term limits are an important barrier to the personalization and aggrandizement of political power. Being one of the symptoms of global autocratization trends, attempts have been increasing in recent years, especially in Africa and Latin America. The answer to the question of whether democracy aid makes a difference in protecting term limits would not only help address a research gap, it would also provide relevant knowledge for informed political decision-making.
Overall, our results confirm prior studies that "targeted aid" such as democracy aid works and that its effects vary according to the phases of institutional reform processes. Democracy aid in 62 African and Latin American countries between 1990 and 2014 had positive effects and contributed towards protecting term limits according to our empirical analysis. Even moderate amounts of democracy aid can have a sizeable influence. More specifically, democracy aid substantially reduced the risk of incumbents circumventing term limits, whereas it had less effect on the attempts at circumvention. In other words, our findings suggest that democracy aid helps to fend off term-limit circumventions, whereas it is more uncertain that democracy aid helps to deter presidents from trying to circumvent term limits. This variety in the findings proves that our introduction of a novel conceptual distinction between the attempt to circumvent a term limit and the outcome of such an attempt allows for a more refined understanding of the effects of democracy aid on protecting term limits than previous studies.
What can explain this difference in the effectiveness of democracy aid in the two steps for attempting to circumvent term limits? First, the findings emphasize the temporal dimension of aid effectiveness, which has often been overlooked in the study of democracy promotion. Democracy aid aims at capacitating domestic actors and strengthening institutions that defend and protect democracy. Pro-democratic actors such as NGOs might have received democracy aid before an attempt to circumvent term limits was made. Explicit action against the term-limit circumventions of these actors takes place only after the incumbent has embarked on the attempt to evade or abolish the term limit. It is at that point when the additional capacities of these actors due to democracy aid become important. This implies that democracy aid may not necessarily have immediate results, but it may still make the difference in protecting democracy down the line. In addition, attempts to circumvent term limits cause strong reactions in pro-democratic actors, for instance when forming social movements with low levels of organization. In these situations, international actors provide ad hoc assistance and adjust their aid programmes. This combination of longer-term support that helps to create capacities already before an attempt has been made and the ad hoc support of international actors seems to be crucial in tense times of term-limit circumventions. Second, the type and range of mechanisms employed for the direct protection of democracy vary according to each step in the attempt to circumvent term limits. The greater the opposition against the attempt to circumvent term limits and the greater the indication of manipulation by the incumbent, the more international actors can intensify their direct and ad hoc action. For instance, once the rule or procedure is changed despite large-scale societal opposition, international actors are more likely to sanction the sitting government or incumbent.
Looking at the broader picture, our findings can inform theories of international democracy protection. We prove that democracy aid can repel attempts to erode democratic institutions effectively and that it makes a difference for pro-democratic political struggles. These results challenge recent findings that democracy assistance has become tame towards autocrats. 64 Bush concludes that transnational democracy assistance shies away from political struggles for more democracy. Based on our crossnational research, which we complemented with qualitative case studies published elsewhere, we argue that democracy aid is not as tame as suggested. However, we require more nuanced evidence for knowing under which conditions democracy aid effectively deflects autocratic trends. First, as other studies have suggested already, 65 we need to learn more about the effects of different types of mechanisms or "aid modalities" of democracy aid. Although we can confidentially conclude that democracy aid based on the logic of appropriateness supports pro-democratic forces, it is less certain how it interplays with measures based on the logic of consequences such as conditionalities. Given the scarce and only case study-based data on conditionality and sanctions during attempts to circumvent term limits, no systematic assessment of the interplay between instruments of the logic of consequences and instruments of the logic of appropriateness is possible at this point.
Second, it is necessary to identify typical patterns of historical contexts. Our analysis is the first one that took a cross-regional perspective in the study of democracy aid and its success in circumventing term limits. There is a greater risk for circumvention attempts of presidential term limits in Latin American than in African states. This indicates the need to compare the political evolutions of presidentialism and term limits more systematically. Against this finding, investigating cases as paired comparisons with two case pairs from Latin America could help to identify typical path dependencies. 66 Third, the conceptual foundations and findings of this analysis speak to a broader global phenomenon, namely the global autocratization trend. In this context, protecting democracy will become more important in the future. Knowing that democracy aid can support domestic pro-democratic actors is not only an encouraging signal for policymakers, but also an incentive to build more knowledge about the "dos" and "don'ts" of democracy protection. The conceptual distinction of the sequences of political reform processesas proposed in this analysisis a promising point of departure for further research on that matter. Autocratization processes unfold in certain sequences. Learning how democracy aid works best in each of these sequences is of great interest for this research and an important step towards protecting democracy worldwide.

Notes on contributors
Julia Leininger (PhD University of Heidelberg) heads the Department "Transformation of political (dis-)order" at the German Development Institute/Deutsches Institut für Entwicklungspolitik. Her research and evidence-based policy recommendations focus on democracy promotion and protection, social cohesion, African politics, and future scenarios for a transformation to sustainability. Daniel Nowack is a researcher in the Department "Transformation of political (dis-)order" at the German Development Institute/Deutsches Institut für Entwicklungspolitik. He is a doctoral candidate at the Institute for Political Science (IPW) at Heidelberg University. In addition to conducting research he provides research-based policy recommendations for German and international actors in development cooperation. His research focuses on democracy, democracy support and promotion, and the interplay among cultural change, civil society, and social movements.