Presidential Powers in Latin America Beyond Constitutions

been fewer attempts of conceptualizing and measuring these powers in combination with partisan sources and informal practices which also may be used by presidents to exert influence in Latin America. The article advances from previous attempts that have focused on constitutional powers in order to develop indices able to capture variation in presidential powers between one president to another in a specific country and within the same presidency, even if constitutional powers remain the same. The results also show that the factor that leads to the concentration of powers by the executive in Latin America is partisan powers. This finding contradicts an influential understanding in the literature that presidencies with more constitutional powers are more problematic for sustaining democracy.

Despite the relationship identified by the authors between the constitutional and partisan powers, Shugart and Mainwaring propose a typology of presidential powers in Latin America limited to constitutional guarantees of influence over the legislative process. If the constitutional powers of Latin American presidents have been studied extensively (García Montero 2009;Mainwaring & Shugart 2002a;Negretto 2013;Payne et al. 2003), there have been fewer attempts of conceptualizing and measuring these powers in combination with the partisan sources and informal practices which also may be used by presidents to exert influence in Latin America.
The article contributes both by advancing from previous indices that focus on constitutional powers and by showing that the factor that leads to the concentration of powers by the executive in Latin America is partisan powers. The indices developed here are able to capture variation in presidential powers between one president to another in a specific country and within the same presidency, even if constitutional powers remain the same. Indices restricted to these powers, such as the ones proposed by Negretto (2013) for Latin America, would not be able to do the same.
The results also show that the presidencies of Nicaragua, Bolivia, Ecuador, Uruguay, and Colombia were potentially dominant. However, this potential led to a concentration of powers by the executive only in Nicaragua and Bolivia, who possessed presidents with the most partisan powers among the 17 countries analyzed here.
The role of partisan powers contradicts Shugart and Carey's (1992) claim that presidencies with more constitutional powers are more problematic for the survival of democracy. It is also consistent with recent findings from Pérez-Liñán, Schmidt, and Vairo (2019) on the effects that the concentration of power by the executive has on democratic instability in Latin America. The problems in this sense identified by the literature do not refer to the presidencies with the highest values for constitutional powers in the indices developed here (from Chile and Brazil), but to three of the five (from Nicaragua, Bolivia, and Venezuela) with the highest indices of partisan powers.
The article is composed of five additional sections. The next item deepens the theoretical discussion and relates it to the arguments developed here. After that, the methodological procedures are explained. In the following three sections, the results and the conclusions of the analysis are presented and discussed.

Theoretical Discussion
Presidential power is defined in the literature in terms of the dimensions in which it can be exercised (Negretto 2013;Shugart & Carey 1992). There are also more specific views (Biglaiser & DeRouen 2004;Hicken & Stoll 2008), that understand it as the degree to which power is concentrated in the executive at the national level of government. For Shugart and Carey, presidential power is expressed in constitutional terms, divided into two basic dimensions, one that refers to power over legislation and the other that encompasses non-legislative powers, including authority over the cabinet and calling early elections to congress. Negretto (2013) expands the possibilities of exercising presidential power and understands it as a multidimensional concept, encompassing the presidents' authority in the exercise of their functions as party leader, head of state, head of government, and legislator. To operationalize this multidimensionality, however, the author follows the proposal of Shugart and Carey (1992) and restricts himself to the constitutional limits. According to Negretto, the main dimensions of presidential power that are regulated by the constitution refer to, on the one hand, nominating and removing ministers and authorities from the executive and judiciary branches and, on the other, taking part and exerting influence on law and policy formulation. These dimensions are, respectively, governmental powers and legislative powers. Shugart and Mainwaring (2002) had already drawn attention to the importance of the party dimension, claiming that the presidents' capacity to influence law and policy formulation in Latin America depends on the relationship between their constitutional and partisan powers. However, this understanding was little operationalized, even by the authors themselves. At the same time, informal practices that may grant or withdraw power from presidents also need to be considered for the Latin American countries, as the literature has pointed out (Hartlyn 1998;Helmke & Levitsky 2006;O'Donnell 1994O'Donnell , 1996Siavelis 2006).
Since Shugart and Carey (1992) proposed a way to measure presidential powers considering the legislative and non-legislative capacities granted by the Constitution, several authors have explored that and further suggestions to offer alternative forms of measurement. Whether they be for a greater number of cases (Doyle & Elgie 2016;Johannsen 2003) or specific regions, such as Latin America (Corrales 2013;Negretto 2013), parts of Europe (Hellman 1996;Metcalf 2000), or countries from the former Soviet Union (Frye 1997), these options are also based on constitutional prerogatives. Shugart and Carey (1992) measure presidential power by adding the scores given to a number of legislative and non-legislative powers. A limitation of their index, according to Negretto (2013), is the understanding that each legal prerogative evaluated contributes equally to a president's power. That means, for instance, that having veto power would be the same as having decree power. With the aggregation of the scores, they also do not take into account that the effects may be interactive. In other words, a specific combination of legal instruments may contribute to a president's overall power (Cheibub 2007;Negretto 2013).
There are studies (Carey & Shugart 1998;Negretto 2004) that show, for instance, that agenda-setting and veto powers have interactive effects. Thus, a president with moderate agenda-setting and veto powers would have a larger impact on the formulation of law and policies than one with strong power in one of those categories and no power in the other. An index estimated by aggregation could not capture these details, since the addition of the two scores at the mid-point of the scale would be equal to the sum of a maximum and a minimum score. Fortin (2013) goes further and evaluates that presidential power indices have limited validity, for reasons such as the impossibility of indicators in measuring the set of powers associated with the presidents or, on the other hand, in generating separate measures of legislative and non-legislative powers. Another problem would be the assumption that equal scores mean the same effects for a causal analysis, which could not be the case since they are the result of different combinations of power.
Despite agreeing with Fortin's evaluation, Doyle and Elgie (2016), who compile existing indices of presidential powers and submit them to procedures to increase their reliability, consider that many concepts suffer from similar measurement and validity problems in social sciences. So the authors claim that it is better to use the information offered by the indices and make them more reliable.
For Doyle and Elgie, there are problems in measuring presidential powers based exclusively on constitutional sources, because constitutions can be imperfect measures of actual power. This does not mean, according to the authors, that there are no problems in considering other sources and measuring behavioral powers, as Siaroff (2003) does, because there is the risk of capturing the impact of factors such as party competition rather than presidential power itself.
Santos, Pérez-Liñán, and García Montero (2014) measure the institutional capacity of Latin American presidents using an index elaborated by García Montero (2009). This indicator goes beyond the constitutional guarantees of influence on law and policy formulation by including the rules of the legislative process, with the goal of measuring the chief executive's role in every stage of this process. On the other hand, factors such as the president's leadership over their party or coalition and informal mechanisms of influence are not considered. Negretto (2013) includes the exercise of government and elaborates an index for governmental powers and another for legislative powers, with Latin America as reference. If his proposal goes beyond the legislative dimension, it still follows the pattern of being restricted to the constitutional prerogatives of presidents.
The issue about the insufficiency of constitutions as sources for measuring presidential powers is especially relevant in the context of Latin America, where there is a gap between what the legal framework establishes and its realization, as emphasized by O'Donnell (1996). This has led to a scenario in which informal rules may be more rooted than formal ones. Helmke and Levitsky (2006) define informal institutions as socially shared rules, generally unwritten, that are created, transmitted, and reinforced outside of officially sanctioned channels. For the authors, despite the fact that the literature about presidentialism in Latin America was initially focused on the formal aspects of the institutional design, later works suggest that presidential system dynamics cannot be fully understood with a strictly constitutional framework.
The effects of informal practices on presidential powers can be to either reinforce or limit them. In the former sense, O'Donnell (1994O'Donnell ( , 1996 theoretically exposes how patrimonial practices enable a degree of executive predominance that exceeds what is constitutionally prescribed and undermines horizontal accountability. His assumption has been tested by case studies, such as Hartlyn (1998), who shows how patrimonialism in the Dominican Republic can reinforce, and be reinforced by, the concentration of power in the executive.
At the same time, informal institutions may limit presidential power, even in cases of constitutions that make room for an exaggerated presidentialism. By studying the Chilean case, Siavelis (2006) shows how informal institutions have contributed to distribute power within the government coalition and to social groups, limit the amount of power used by presidents and consolidate democracy. In an attempt to theoretically model the relationship between formal and informal presidential powers in Latin America and the United States, Morgenstern, Polga-Hecimovich, and Shair-Rosenfield (2013) state that the informal ones are more valuable to formally weaker presidents.

Methodology
This article seeks to operationalize the measurement of the different sources of presidential powers and to give empirical and broad ground to previous indications from theoretical or case studies that the formal aspects of institutional design are insufficient for assessing the powers of Latin American presidents. The intention is to create a comprehensive index, not only by including factors such as the president's leadership over their party or coalition and informal practices, but also by accounting for the variety of dimensions in which presidential powers are exercised. The indices are built through aggregation and include interactive effects.
The dimensions of analysis refer to the functions of a president as party leader, head of state, head of government, and legislator, based on constitutional prerogatives, capacity for leadership, and informal practices that may be associated with the exercise of each function. These dimensions, proposed by Negretto (2013), 1 are used because they cover a wide range of presidential tasks. Given that the roles of head of state and head of government are combined in presidentialism, they are joined in one dimension. For each dimension, a series of variables was defined, as Table 1 shows.
In total, there are 47 variables, distributed over the dimensions of party leader, head of state and government, and legislator. For the last two dimensions, variables used by authors such as Shugart and Carey (1992) and Negretto (2013) were improved and complemented with others that refer to informal practices. The variables for the party leader dimension are in the most part those used by Mainwaring and Shugart (2002b). In order to take account of informal practices, some variables were adapted, as well as others were added.
A Multiple Correspondence Analysis (MCA) was applied to test the results and identify the most important variables. MCA is a statistical technique that reduces the dimensionality of a set of variables, by looking for a pattern and proximity between observations and by determining which variables are most important.
To consider informal practices, at least one variable that is able to capture them was included in each dimension of analysis. The selection of these variables does not intend to be exhaustive. The intention was to select variables that could be measured and repeated throughout the cases. As an additional procedure, some variables used in the literature on executive-legislative relations have been operationalized in a way that could capture informal practices.
The main example is the way adopted to count the coalitions. The government coalition, formalized by a joint candidacy in presidential elections, a written agreement among parties or, the most commonly used criterion, positions in the president's cabinet, was not considered. The concept used was of a legislative coalition, considered as one that encompasses the parties that systematically vote in line with the interests of the current government. In this case, there is no need for formalization through a joint candidacy in the previous elections, a written agreement or cabinet positions. This procedure is able to take into account patrimonial and other informal practices that aggregate parties to the president's support base, such as the ones Mejía Acosta (2006) describes for the Ecuadorean case. According to him, some parties voted with the government in exchange for benefits and did not publicly admit to integrate the coalition in power. In fact, they systematically denied doing so, to avoid the political fallout of associating themselves with badly evaluated presidents and to be less exposed to criticism for negotiating their support, which the author defines as 'ghost coalitions'.
The discipline of the president's party or coalition offers another example. These variables have three categories, which are undisciplined, disciplined through individual negotiations, and disciplined through institutional negotiations. With these categories, it is possible to take into account both the level of discipline observed in the legislative votes of interest to the government and the practices adopted to achieve discipline in the president's party or coalition. 2 The operationalization of the variables about the number of parties and the discipline of the president's coalition applies theoretical contributions (O'Donnell 1994(O'Donnell , 1996Morgenstern, Polga-Hecimovich, & Shair-Rosenfield 2013) that highlight patrimonial and alternative ways that can reinforce formal presidential powers in Latin America.
There are also variables in the dimensions of head of state and government and legislator that contribute towards the president having more or less power according to the circumstances and that are not necessarily associated with formal aspects. Those are the cases of the relationship with the presidency of the lower chamber, in the legislator dimension; the state ownership of media companies, the country's dependency on foreign loans, and the subordination of the military to civilian power, in the head of state and government dimension. The last three variables were included because of the relevance that armed forces, international financial institutions, and private media have shown historically in Latin America by limiting or questioning the presidents' power through means that are not necessarily foreseen and controlled by formal rules. The four variables encompass aspects such as the effects of having or not an ally in the lower chamber's presidency; the possibility of the president to display their accomplishments and offer an alternative narrative to the one in the private media; the ability to reject demands in exchange for loans and to show control over economic policy; and the control over the military, whether to discourage them from rebelling or to judge them for violations during a previous authoritarian regime.
The variables for the party leader dimension were coded taking into account the presidents who occupied office The coding of the variables concerning their party and coalition blocs considers the number of elected representatives for each party that, at the time of the evaluation, was governing or taking part of the legislative coalition in the lower chamber. The same timeframe was used to code the four variables of the head of state and government and legislator dimensions that do not refer to constitutional powers. In total, 14 variables of the three dimensions used in the indices were coded by taking into account the situation in mid-2017.
The coding of the other 33 variables is based on constitutions and auxiliary legislation. The complete procedure can be replicated anytime using the categories of the variables and the explanations for the coding. The variables are placed in an ordinal scale ranging from the category with the lowest degree of presidential power to the category with the highest degree in each variable. Tables 2 to 7 of the appendix show: a) the categories and codes used for each power in the party leader, head of state and government, and legislator dimensions; b) the coding of these variables for the 17 countries analyzed; and c) the sources used to code the variables of each dimension.

Results
To elaborate the indices through aggregation, the variables were weighted so as to have equal weights. Although they still may have different effects on a Latin American president's power, it is not possible to quantify that difference nor consider that it is the same in all cases. Giving equal weight to all variables is the same procedure used by Shugart and Carey (1992), who have become an important reference till today for measuring presidential powers.
For the legislator dimension, two weighted indices were constructed. The first was calculated by aggregating the codes for each country and dividing the result by the total of variables, as with the other two dimensions. The second index was based on the interaction suggested by Negretto (2013) and achieved by multiplying the sums of the veto and agenda-setting powers. Table 8 shows the results.
Brazil has the highest index of presidential powers in the head of state and government dimension and the fourth highest in the legislator dimension, while Chile leads the latter and presents the third highest index for head of state and government. Considering jointly the two dimensions, in which there is a predominance of powers of constitutional origin, Chile and Brazil exhibit the highest results and are the only countries with values above 0.60.
The index for Brazil in the party leader dimension, the second-lowest, helps explain why Dilma Rousseff (2011-2016) had her second term interrupted in 2016 due to an impeachment process. Although the index refers to the presidency of Michel Temer, who was Rousseff's vice-president and took over to finish her term, the explanation is pertinent given that the removed president had experienced more troubles than Temer with her legislative coalition.
The result for Temer does not mean that his legislative coalition was not majoritarian, but quite the opposite, considering that it reached a qualified majority. Yet this relative strength was diluted in the set of variables for the party leader dimension because it was a coalition of many parties, electorally and ideologically fractioned, and disciplined through individual negotiations.
In the extreme opposite of the head of state and government dimension, with the lowest index, is El Salvador, a little behind Colombia and Venezuela, which present the same value. In the legislator dimension, Costa Rica has the smallest index, with Guatemala and Honduras, which have the same result, being the closest, all three with approximately 0.10. Considering both dimensions in a combined way, Costa Rica, El Salvador, Guatemala, Honduras, and Venezuela present the lowest values.
On the other hand, the executive held a concentration of powers during Hugo Chávez's period as Venezuelan president, from 1999 to 2013. Again, the partisan powers explain this, since they allowed Chávez to have a disciplined single-party majority and exert influence over accountability institutions through appointments made by that legislative majority. The index for Venezuela in the party leader dimension refers to the presidency of Chávez's successor, Maduro, who did not have a legislative majority at the time of coding, but his index in that dimension stayed well above the ones in the other two dimensions, 0.583 against 0.254 for head of state and government and 0.406 for legislator. In addition, Maduro took advantage of previous appointments to accountability institutions for revoking powers of the Venezuelan legislature.
The partisan powers are also what explain the concentration of powers in the executive identified by the literature in Bolivia and Nicaragua (Anria 2016;Mainwaring & Pérez-Liñán 2015) during the Morales and Ortega presidencies. Nicaragua and Bolivia present low or medium values in the head of state and government and legislator dimensions, respectively 0.369 and 0.367 for the two dimensions combined, against 0.622 and 0.605 of Chile and Brazil. When it comes to partisan powers, however, Nicaragua and Bolivia have the highest indices, respectively 0.850 and 0.767, which are also the highest values for the four indices calculated that do not include an interaction between variables. The interactive version of the index for the legislator dimension creates some results superior to one, because of the multiplication between the sums of the veto and agenda-setting powers.
The inclusion of that interaction, suggested by Negretto (2013), slightly alters the picture in the legislator dimension. Chile and Costa Rica still have, respectively, the highest and smallest indices. There is a change in the countries that present the second and third highest values, with Argentina and Brazil moving in front of Ecuador. That limited effect of the interaction between veto and agenda-setting powers for the result of the legislator dimension reinforces the validity of the aggregation method used here.
For the set of all three dimensions, Chile, Nicaragua, Ecuador, Bolivia, Brazil, and Uruguay had the most powerful presidents. In the cases of Chile and Brazil, this is due to the powers as head of state and government and legislator, the dimensions in which there is a predominance of powers of constitutional origin. In the cases of Nicaragua and Bolivia, it is due to the powers as party leader. Ecuador and Uruguay are intermediary cases, since their presidents had more powers in the legislator and party leader dimensions, thus combining sources both constitutional and partisan. In the extreme opposite, Guatemala and Costa Rica had the weakest presidents, the only ones with an overall index of 0.20, approximately.
Chile and Brazil show that having a president with more constitutional powers does not mean as a result a concentration of powers by the executive, since the factor that led to this result in Nicaragua and Bolivia is partisan powers. The role of these powers contradicts the claim of Shugart and Carey (1992) that presidencies with more constitutional powers are more problematic for sustaining democracy.
The problems of regime instability identified by the literature in Latin America (Anria 2016;Mainwaring & Pérez-Liñán 2015) do not refer to the presidencies from Chile and Brazil with the highest indices of constitutional powers, but to three of the five (from Nicaragua, Bolivia, and Venezuela) with the highest indices of partisan powers. The role of these powers is also consistent with recent findings of Pérez-Liñán, Schmidt, and Vairo (2019) concerning the effects that the concentration of power by the executive has on democratic instability in Latin America, since two of the four variables used by the authors are related to partisan powers.
The presidencies coded here are presented in Graph 1 according to the powers of presidents as party leader and, in combination, as head of state and government and legislator. That definition of the axes for the graph is based on Shugart and Mainwaring (2002), for whom the capacity of a president in Latin America to influence law and policy formulation depends on the relationship between their constitutional powers, which are concentrated in the head of state and government and legislator dimensions, and partisan powers.
In the superior right quadrant, with indices higher than half of the scale on both axes, are Nicaragua, Bolivia, Ecuador, Uruguay, and Colombia. In the inferior right quadrant, with values higher than half in the axis of the dimensions in which there is a predominance of powers of constitutional origin, is the largest group of cases composed by Chile, Brazil, Argentina, Peru, Mexico, Panama, and Paraguay. On the left side of the graph, Venezuela and El Salvador are in the superior quadrant, with indices larger than half of the partisan powers axis, while Honduras, Costa Rica, and Guatemala are in the inferior quadrant, with values smaller than the half of both axes.
A MCA was conducted to test the results and identify the most important variables. The top two dimensions of the MCA explain 24.03% of the results. If we interpret dimensions 1 and 2 as, respectively, party leader and head of state and government, Nicaragua and Bolivia, in the first one, and Brazil, in the second, are in the same positions as in the indices of these dimensions, with scores much higher than those of other countries.
Graph 2 displays the squared loading estimated for each variable in the top two dimensions of the MCA. This statistic estimates the contribution of each variable. Legislative majority of the president's party, threshold to approve an impeachment, appointment of the controller general, and impeachment option for political reasons are the most important variables. These findings show the great power that a single-party majority may offer to a president and the increasing importance that impeachment rules have acquired for the survival of presidents in Latin America. On the other hand, a single-party majority leads to power concentration in the executive, since Nicaragua and Bolivia were the only cases of qualified majority for the president's party.
Using the classification categories proposed by Shugart and Mainwaring (2002) and adding other types of power, since the authors limit themselves to the legislative authority of Latin American presidents, it is possible to find that Nicaragua, Bolivia, Ecuador, Uruguay, and Colombia had potentially dominant presidencies; Chile, Brazil, Argentina, Peru, Mexico, Panama, and Paraguay, proactive; Venezuela and El Salvador, reactive; and Honduras, Costa Rica, and Guatemala, potentially marginal. Table 9 presents the characteristics of each group of presidencies.
These classifications do not mean, for instance, that all presidencies defined as potentially dominant materialize this potential and that all classified as potentially marginal are in fact marginalized. Considering the five presidencies classified as potentially dominant, this has only led to a concentration of powers by the executive in Nicaragua and Bolivia, whose presidents obtained the greatest scores of partisan powers among the 17 countries analyzed here.

Variation between presidencies and over time
In order to show the utility of the indices of presidential powers, two countries were chosen: one where there was a change of president since the first coding was done in 2017; and one where the presidential term lasted from 2017 to 2021. The first is Brazil, where Temer was substituted in 2019 by Jair Bolsonaro, elected in 2018 for a fouryear term. The second is Ecuador, where Moreno was in the final year of his term in 2021. With these cases, it is possible to show variation between one president to another in a specific country and over time within the same presidency.
Bolsonaro was elected by a small party, which increased from one deputy elected in 2014 to 52 in 2018 and Graph 1: Distribution of Countries According to Presidential Powers. Source: Authors' elaboration.
became the second-largest force in the Brazilian chamber of deputies. Even so, Bolsonaro has never presided over the party and left it in his first year as Brazilian president to become an independent politician since then. Moreover, he is not sympathetic about formalizing a government coalition and giving ministries and power to allied parties.
On the other hand, his predecessor, Temer, had presided over his party before becoming president and took charge of formalizing a government coalition and Table 9: Classification of Presidencies According to its Combinations of Powers.

Powers in Common Countries
Potentially dominant Single-party with qualified or absolute majority and veto subject to override by a qualified or absolute majority Nicaragua, Bolivia, Ecuador, and Uruguay Majority coalition of three to five parties, decree subject to legislative authorization, and veto subject to override by an absolute majority Colombia Proactive Majority coalition of six or more parties, unilateral decree or subject to legislative authorization, veto subject to override by a qualified or absolute majority, and appointment of attorney-general subject to ratification by a qualified or absolute majority securing enough legislative support for being able to survive two attempts of prosecuting him. Since Bolsonaro has the same constitutional powers as Temer had, an index restricted to this source of presidential powers would not be able to capture variations between one president to another and to guide an evaluation of consequences. As Graph 3 shows, there is no variation in the amount of power as head of state and government and legislator between Temer and Bolsonaro, but the current president lost power in the party leader dimension in comparison to his predecessor, due to factors such as lack of experience as party leader and internal struggles among political and sectorial factions of his government. Considering this, Bolsonaro is more dependent on the legislative to organize his support base, especially the presidency of both chamber of the congress, and vulnerable to an impeachment process.
For Ecuador, there is even more variation over time within the same presidency. Moreno was in the first and the last year of his term as Ecuadorian president between the coding of his powers in 2017 and the repetition of this procedure in 2021. He was elected by a party that achieved a majority of seats in the national assembly, the only legislative chamber in Ecuador. He also presided over this party. In 2021, at the end of his term, Moreno was being challenged in the party and decided to leave it, which by itself did not hold a majority of seats in the national assembly anymore.
As Graph 4 shows, Moreno had the same powers as legislator between one year to another and slightly more powers as head of state and government due to an informal source, which is the fact that there was no security forces uprising in the last ten years, but he did experience an important loss of powers as party leader. Considering this, Moreno had been facing legislative setbacks and formal proceedings against some cabinet ministers in his final days as president.

Conclusions
This article contributes both by advancing from previous indices that are limited to constitutional powers and by showing that the factor which leads to the concentration of powers by the executive in Latin America is partisan powers. With comparisons between presidents of Brazil in 2017 and 2021 and between these years of the same presidency in Ecuador, the indices developed here show that they are able to capture variation in presidential powers even if constitutional powers remain the same from one moment to another. Indices restricted to constitutional powers, such as those of Negretto (2013), would not be able to capture this kind of variation and guide an evaluation of the consequences, as the indices proposed here do.
The presidencies classified as potentially dominant in terms of overall powers are from Nicaragua, Bolivia, Ecuador, Uruguay, and Colombia. However, this potential led to a concentration of powers by the executive only in Nicaragua and Bolivia, whose indices of partisan powers were the highest ones among the 17 countries analyzed here. This shows the relevance of partisan powers in explaining the concentration of powers by the executive in Latin America.
Chile and Brazil also illustrate that having a president with more constitutional powers does not mean as a result a concentration of powers by the executive and democratic instability. The problems in sustaining democracy identified by the literature in Latin America (Anria 2016;Mainwaring & Pérez-Liñán 2015) do not refer to the presidencies with the highest indices of constitutional powers, which are from Chile and Brazil, but to three of the five (from Nicaragua, Bolivia, and Venezuela) with the highest indices of partisan powers.
The role of these powers contradicts Shugart and Carey's (1992) claim that presidencies with more constitutional powers are more problematic for sustaining democracy. It is also consistent with recent findings from Pérez-Liñán, Schmidt, and Vairo (2019) concerning the effects that the concentration of powers by the executive has on democratic instability in Latin America. Overall, the article's results give an empirical and broad ground to indications from theoretical or case studies (Hartlyn 1998;O'Donnell 1994O'Donnell , 1996Siavelis 2006) that the formal aspects of institutional design are insufficient for evaluating presidential powers in Latin America.

Notes
1 The author does not operationalize the dimensions and restricts himself to legislative and governmental powers conferred to Latin American presidents by different constitutions throughout the history of each country. His unit of analysis, then, is the constitutions. 2 Due to the categories that were defined for the variable, the coding was based on information gathered from political news coverage. Although it would be desirable to do it based on the proportion of votes that followed the government's orientation, establishing a threshold from which it could be determined if the party or coalition is disciplined, there is no systematized data on party discipline for legislative votes of all the countries analyzed here. The same interpretive procedure was adopted to code the variable on the cohesion of the president's party or coalition.