Replacing and Amending Constitutions: The Logic of Constitutional Change in Latin America

Abstract

Since 1978, all countries in Latin America have either replaced or amended their constitutions. What explains the choice between these two substantively different means of constitutional transformation? This article argues that constitutions are replaced when they fail to work as governance structures or when their design prevents competing political interests from accommodating to changing environments. According to this perspective, constitutions are likely to be replaced when constitutional crises are frequent, when political actors lack the capacity to implement changes by means of amendments or judicial interpretation, or when the constitutional regime has a power-concentrating design. It is further argued that the frequency of amendments depends both on the length and detail of the constitution and on the interaction between the rigidity of the amendment procedure and the fragmentation of the party system. The article provides statistical evidence to support these arguments and discusses the normative implications of the analysis.

Since 1978, all the countries of Latin America have either replaced or amended their constitutions. Replacement and amendment are, however, substantively different means of constitutional transformation. While the replacement of the existing constitution involves a political decision to re-create the basic legal structure of the state, amendments, like judicial interpretation, are mechanisms of legal adaptation that preserve the continuity of the constitution in a changing environment. The frequent replacement of constitutions thus puts into question the legal and political foundations of democratic regimes. What explains the choice between replacements or amendments?

It is argued here that constitutions are replaced when they fail to work as governance structures or when their design prevents competing political interests from accommodating to changing environments. According to this perspective, constitutions are likely to be replaced when constitutional crises are frequent, when political actors lack the capacity to implement changes by means of amendments or judicial interpretation, or when the constitutional regime has a power-concentrating design. It is further argued that the frequency of amendments depends both on the length and detail of the constitution and on the interaction between the rigidity of amendment procedures and the fragmentation of the party system. The article provides statistical evidence to support these arguments and discusses their normative implications. In particular, it is suggested that while new Latin American democracies may foster constitutional stability by adopting inclusive institutions, more flexible amendment procedures, and strong mechanisms for constitutional adjudication, it is likely that constitutional crises will continue to provide incentives for the constant renegotiation of constitutional agreements.

The article first considers the problem of constitutional change in comparative perspective. This is followed by a discussion in Section 2 of the reasons and various means for constitutional change. From this discussion emerge several general hypotheses about constitutional replacements and amendments, which are tested in Section 3 using different models of regression analysis for longitudinal data. The article concludes with a discussion of the implications of the analysis for constitutional design, and of the factors that work against constitutional stability in Latin America's new democracies.

Assessing Constitutional Change

The most important discussions in the research agenda on institutions in the social sciences hinge on the problem of institutional change and its conceptual antithesis, institutional stability.¹ Yet the meaning of change in comparative institutional analysis is inherently ambiguous.

Institutional change may imply the displacement of preexisting institutional forms or their adaptation to shifting environments. Adaptation, in turn, can occur by the introduction of formal alterations, by old rules being interpreted in new ways, or by informal rules and practices being developed that transform the meaning of existing institutions.² In addition, there is ambiguity in assessing the magnitude of change. It is not immediately obvious, for instance, when a formal revision should count as a significant instance of institutional change. Politicians may reform institutions without really altering their content.³ New regulations may reproduce previous ones or introduce marginal changes that are not expected to alter the effects observed under preexisting rules.

The study of constitutional change is an excellent starting point for understanding the general problem of change in formal political institutions. Given the role of constitutions as a “higher” law, both their nature and design work toward their self-preservation. Constitutions cannot, however, remain immutable; they need to be transformed to adapt to deep changes in the political, social, and economic environment. One way to change constitutions is through textual alterations, either through amendments or via wholesale replacement. Constitutions can also be modified over time without textual changes, typically by means of constitutional court rulings. Less visibly, constitutions may also be transformed by legislative and executive decisions, or by the informal practices of political actors (Ackerman 1991; Levinson 1995).

These are very different means of constitutional transformation. According to classical constitutional theory, amendments and judicial interpretation are the main mechanisms for adapting constitutions to changing circumstances. In practice, constitutions are also replaced, but this is not considered to be a regular means of adapting a constitution to new conditions (Lutz 1995: 243; Murphy 2007: 498). The enactment of a new constitution supposes the irruption of the constituent power of the people, which finds no limits in the existing constitution and implies in practice its legal abrogation. This is why most constitutions do not provide for their own replacement, thus turning this alternative into an extraordinary, usually irregular form of constitutional change.⁴

Given the disruptive nature of replacements, constitutional theory suggests that they should be exceptional events. Modern constitutions trace their legitimacy back to a sovereign decision of the people, which should take place only during extraordinary times, as in a revolution or in the midst of a major political crisis (Ackerman 1991). Constitution-making in established democracies seems to confirm this expectation. The current U.S. constitution, for instance, dates to 1789, the year it was formally ratified. In some western European countries, such as France, Spain, Portugal, and Greece, constitutional replacements have been more frequent, but several other countries of the region, such as Norway, Belgium, and Denmark, retain constitutions enacted in the nineteenth century. On average, the countries of Western Europe adopted 3.2 constitutions from 1789 to 2001, with a mean lifespan of 76.6 years.

Constitutions have been less enduring in other regions of the world, including Latin America. Since independence, a total of 194 constitutions have been enacted in this region, of which 103 have been in force from 1900 to 2008 (see Table 1).⁵ This is an average of 10.7 constitutions per country since the early decades of the nineteenth century, and an average of 5.7 constitutions per country from 1900 to 2008. The mean lifespan of constitutions has been 16.5 years for all the constitutions enacted since independence, and 23.3 years for those in force from 1900 to 2008.⁶

Table 1. Constitutional Change in Latin America

Source: Author's calculations, based on: Constituciones Hispanoamericanas (http://www.cervantesvirtual.com/portal/constituciones/); Political Database of the Americas (http://pdba.georgetown.edu/); and country sources.

Due to the exceptional durability of Latin America's new democracies, the rate of constitutional replacement decreased somewhat between 1978 and 2008. Even so, an average of almost one new constitution was enacted per country during this period. This is a relatively high rate of constitutional replacement, particularly if one considers that not all the countries of the region established new constitutions with the inauguration of democracy; that some democratic regimes (Costa Rica, Colombia, and Venezuela) had already been established by 1978; and that most democracies have since been stable. As of 2009, every Latin American country except for Costa Rica, Mexico, Panama, Dominican Republic, and Uruguay had adopted a new constitution and some, like Ecuador, had done so more than once.⁷

There is also considerable variation in the number and frequency of amendments to existing constitutions.⁸ Interestingly, however, constitutional amendments and replacements may be inversely related. The frequent replacement of constitutions obviously prevents the accumulation of amendments. At the same time, since constitutions need to adapt to changing circumstances, amendments may be essential for constitutional survival (see Elkins, Ginsburg, and Melton 2009; Negretto 2008). As shown in Table 2, the mean number of amendments is higher in Western Europe than in Latin America. One reason for this relationship is that constitutions tend to last longer in the former than in the latter region. But the mean amendment rate—that is, the number of amendments divided by the years a constitution has been in force—is relatively similar, which raises the question of what the proper rate of amendment is that enhances the durability of constitutions in a particular environment.⁹

Table 2. Constitutions and amendments in Western Europe and Latin America, 1789–2001

Source: Same as Table 1 for Latin America; Blaustein and Flanz (2008), and Elkins, Ginsburg, and Melton (2009) for Western Europe.

(1) 16 countries.

(2) 18 countries.

(3) Amendments to constitutions in force in 2001.

(4) Amendments by years of life.

The above analysis suggests that constitutional replacement and amendment work very differently as means of constitutional transformation: the first formally displaces an existing constitution, and the second implies its continuity. Thus the main goal of a theory of formal constitutional change should be to explain why political actors alter existing constitutions, and why they choose either replacement or amendment. A comparative theory of constitutional change should also explain the interaction between formal and informal mechanisms of constitutional adaptation, such as judicial interpretation. In what follows, I outline the basic elements of such a theory.

A Political Theory of Constitutional Change

Absent a state-of-nature situation, in which there is no legal order, constitutional change can be conceptualized as a two-step decision. The first consists of deciding whether to maintain or change existing constitutional structures. If change is decided on, the second step consists of choosing between alternatives of constitutional transformation. While the first decision is determined by how satisfied political actors are with existing institutions, the second is determined by the suitability and availability of different alternatives for change.

Incentives for Constitutional Change

It has been argued that because institutions establish obstacles to their own reform, change may be inhibited by even a modest level of uncertainty about the possible outcome of alternative institutional arrangements (Shepsle 1986: 75). The logic of this argument applies with particular force to constitutions. To create a new constitution it is generally necessary to convene a popularly elected constituent assembly, to approve the new text in a referendum, or both. Constitutional amendments usually require qualified congressional majorities, and sometimes a further level of approval, such as a second vote in a different legislative session or legislature, or ratification by voters or a number of states in federal countries. In addition, most constitutional provisions impose strong informational requirements to anticipate the effects of different rules under changing political conditions. Nevertheless, politicians do not always choose to maintain the status quo. Why is this so?

Since institutional change is always costly and the expected benefits of alternative institutions are uncertain, rational risk-averse politicians are unlikely to initiate revisions unless the payoffs obtained from the existing constitution become too low or negative. This suggests that the incentives to replace or amend a constitution crucially hinge on the factors that decrease the value of existing constitutional structures and increase the expected benefits of alternative arrangements. Some of these factors are quite general because they affect the life of all constitutional systems: political transformations at the state or regime level, the dysfunctional performance of the constitution, and balance-of-power shifts (see Negretto 2013). There may also be incentives for change when the constitution fails to adapt to technological changes, to new social values, or to policy shifts. In such cases, however, constitutional change is more likely to occur by the novel interpretation of old rules than by their textual alteration through replacement or amendment.¹⁰

Political Transformations at the State or Regime Level

Profound political changes, such as the founding of a new state or a regime transition, usually require a new form of legality. New states almost invariably symbolize their birth by enacting a constitution. The same may happen with regime transitions, but in such instances the scope for variation is greater. Authoritarian regimes may simply suspend an existing democratic constitution. Democratic regimes may opt to restore a pre-authoritarian constitution, to maintain a constitution enacted during the authoritarian period, or to introduce amendments to adapt an authoritarian constitution to new political conditions. The choice depends on which constitution is considered more capable of effectively and legitimately organizing the new democratic regime, and on the balance of forces between the outgoing authoritarian regime and democratic forces (Geddes 1990).

Dysfunctional Constitutional Performance

A second factor that is likely to render an existing constitution obsolete is its dysfunctional performance. Constitutions are governance structures that organize electoral competition, enable representatives to provide public goods, and maintain citizen support for a political regime. When constitutions fail to perform these tasks, politicians are likely to have an incentive to replace them or amend their provisions, usually under popular pressure for reform. The failure of the constitution as a governance structure inevitably involves a subjective component in the perceptions that citizens and elites have of constitutional and regime performance at a particular historical juncture. Yet there are almost always objective signs of the exhaustion or crisis of a constitutional system, such as frequent breaches of the constitution or inter-branch conflicts over the use and interpretation of constitutional prerogatives.

Balance-of-Power Shifts

Constitutional change may also occur when existing institutions no longer serve the interests of those with the power to change them, or when the losers under a particular set of rules organize a successful reform coalition. This form of constitutional change usually follows important shifts in party competition, such as when established parties collapse or decline, or when new parties and political leaders emerge. Transformations in the partisan context may also go hand in hand with changes in the programmatic or ideological content of public policies. These factors are expected to be important inducements for constitutional change within democratic regimes with unstable patterns of electoral competition. It should be observed, however, that the impact of shifts in the distribution of partisan power on constitutional change is often mediated by other factors, such as whether the nature of the existing constitutional design contradicts the interests of new actors. For instance, a sudden fragmentation of the party system is not likely to induce constitutional changes if the existing constitutional design is already inclusive and allows for the political survival of new parties.

All these reasons for constitutional change are well represented in the historical experience of constitution-making in Latin America. Regime transition was a frequent cause of constitutional change during the twentieth century and accounts for most of the constitutional replacements and amendments that occurred in the region during the late 1970s and early 1980s. As new democratic regimes became stable in recent decades, most constitutional replacements and amendments have been associated with the occurrence of constitutional crises or balance-of-power shifts among party actors.

Means of Constitutional Change

The preceding discussion points to the general factors that provide political actors with an incentive to introduce constitutional changes. However, this discussion is insufficient to shed light on the specific means of constitutional transformation that political actors will select at a particular historical juncture. An analysis of this choice is the most important step for understanding constitutional survival and adaptation and must start by explaining the option to replace a constitution, which is the most extreme form of change. We then need to consider how replacements relate to the most common means of constitutional adaptation; namely, amendments and judicial interpretation.

Replacements

In a recent important work on the durability of national constitutions, Elkins, Ginsburg, and Melton (2009: 76) argued that the choice between different means of constitutional change is mostly a function of relative costs and benefits, with intra-constitutional change typically being the preferred choice because it entails a lower political cost. It follows from this logic that replacing a constitution becomes an option only if amendments are unfeasible under the circumstances. This argument captures an important but partial aspect of the politics of constitutional replacement.

Replacements and amendments cannot always be used as interchangeable means of constitutional transformation. New constitutions are required to change the basic structure of the state and the political regime, whereas amendments are meant to revise procedural details or policy issues contained in the constitution. This is why some constitutions explicitly establish that amendments cannot alter certain fundamental political institutions or principles, such as the form of government, the territorial distribution of power, or particular rights or prohibitions, that belong to the essence of the state as defined by the existing constitution.¹¹ A change in these provisions implies the creation of a new constitution, which inevitably demands the direct intervention of the body of citizens via the election of an independent constituent assembly, popular ratification of the constitution, or both. In other words, when central aspects of the constitutional regime and the state need revision, amendments may not be an option, even if they are feasible.

The case of Colombia in the late 1980s illustrates this situation. During the 1980s, political leaders of the main parties in Colombia agreed that the 1886 constitution needed substantial revisions and that amendments were the appropriate mechanism for introducing these changes. In 1990, however, a constituent assembly was elected to replace the constitution in spite of the fact that the Liberal Party of Colombia had just won the presidential and congressional elections and had the necessary congressional majorities to amend the constitution. The reason for choosing replacement of the constitution was that the incapacity of the state to contain violence in the late 1980s convinced sectors of the political elite and above all citizens and the media that the 1886 constitution had failed to work as a governance structure (Dugas 1993). No amendment would have been sufficient to remedy this failure.

This analysis suggests that certain political events make replacement of a constitution more likely than amendment. For instance, convening a constituent assembly to draft a new constitution is the most reasonable option in the face of a constitutional crisis that puts into question the viability of the political regime. The same could happen during major political changes, as in a transition to democracy, particularly if the pre-authoritarian constitution was replaced or was widely considered to be responsible for the fall of the previous democratic regime. Put differently, some events call for a “new beginning” where it is necessary to reconsider the fundamental institutions that structure the polity. By contrast, some forms of adaptation naturally call for amendments, such as the need to accommodate the existing constitution to shifts in the balance of power among political actors or to policy changes that contradict existing constitutional provisions.¹²

To be sure, constitutions can also be replaced for strategic reasons. Although replacements are supposed to address basic constitutional decisions, there is a great deal of ambiguity about what is fundamental in a constitution or when the latter has failed as a governance structure. This ambiguity provides popular politicians with room to appeal to the people and strategically abrogate the existing constitution as a way of introducing changes they were unable to adopt by means of amendments. There are abundant examples of ambitious political leaders in Latin America who have managed to have their country's constitution replaced to remove a proscription on presidential re-election that was impossible to eliminate via congressional amendment. More generally, the inability to pass amendments prevents the gradual accommodation of competing political interests, which leads to an accumulation of pressures for change that over time may increase the probability of replacement.

The inclusionary or exclusionary features of constitutional design may also affect the possibility of strategic replacements. The mechanism here is not whether adaptation is feasible but whether the design of the constitution itself increases or reduces political pressures for change. In this respect, there is reason to believe that inclusive, power-sharing institutions are more competent than power-concentrating institutions to accommodate competing political interests to shifting environments. Since restrictive rules create absolute winners and losers, some degree of uncertainty regarding future outcomes provides both incumbents and challengers with an incentive to adopt more inclusive institutions (Colomer 2001: 210). Once created, pluralist institutions are not likely to face the same pressures for change because over time they encourage the emergence of a larger number of actors with a vested interest in their maintenance.¹³ Constitutions with an inclusive, pluralist design are thus more likely than those with restrictive institutions to be resilient to temporary shifts in the balance of power among political actors.

Amendments

In recent years, a growing number of political scientists (Lorenz 2005; Lutz 1995; Nolte 2008; Rasch and Congleton 2006) have formulated general propositions on the logic of constitutional amendments. Other authors (Elkins, Ginsburg, and Melton 2009) have considered amendments only indirectly, as a substitute for wholesale replacement. Only one author, Donald Lutz, has attempted to explain amendments both as an independent mechanism of constitutional adaptation and as an alternative to replacement. According to Lutz (2006: 155), the amendment rate should increase as the constitution grows in length and as the amendment procedure becomes more flexible. At the same time, he argued that constitutions should only endure if they are amended neither too often nor rarely because at both extremes they may suffer a loss of authority (1995: 245; 2006: 180). By implication, constitutions are likely to live longer if they include amendment procedures that finely balance flexibility and rigidity (Lutz 2006: 182).

Lutz's proposition that the amendment rate should increase with the length of the constitution is persuasive. Longer constitutions tend to have a higher number of detailed, policy-oriented provisions that require frequent amendment as they become obsolete or inadequate to adapt to environmental changes. However, his argument about the causal effect of amendment procedures on the amendment rate and about the impact of the latter on constitutional durability is debatable.

First, the rate of amendment cannot depend solely on procedural obstacles; the distribution of partisan power is just as important. The most rigid amendment procedure can become flexible in a dominant party system, as under the hegemony of the Institutional Revolutionary Party (Partido Revolucionario Institucional, PRI) in Mexico. By contrast, a flexible amendment procedure may become rigid in practice if party system fragmentation becomes very high, as has been the case in Ecuador since 1979. Further, the occurrence of amendments also depends on whether there is conflict about their content. Even a large number of parties may coordinate to adopt amendments, regardless of the amendment procedure, if there is a reform consensus. This consensus may emerge either because the amendment is non-controversial or because parties were able to exchange support by trading votes on different aspects of reform. Any of these mechanisms of agreement may explain cases such as Brazil, where the amendment rate has been high since 1988 in spite of the fact that no less than three parties have usually had to agree to pass amendments.

Second, although it makes sense to think that an excessively high amendment rate may weaken the authority of constitutions over time, it seems implausible to formulate a universal standard of what constitutes a moderate rate of amendment. This will depend on how frequently the constitution needs to be modified; and that, in turn, would vary across cases as a result of extra-constitutional factors, such as the relative stability of the political, social, and economic environment. While a low amendment rate may be adequate to preserve the constitution in a stable environment, it may undermine the constitution if environmental shifts demand frequent reforms. In other words, determining what is a “reasonable” amendment rate is an empirical matter highly dependent on the universe of cases under analysis. In unstable contexts, relatively frequent amendments should enhance the durability of the existing constitution simply because periodic adaptation prevents the accumulation of social and political pressures for change.

The foregoing suggests that in new democracies facing recurrent demands for institutional and policy reform, as is the case in Latin America, the frequency of amendments should have a positive effect on the durability of constitutions. On the other hand, since most of Latin America's new democracies have fragmented party systems, the amendment rate can only increase if amendment procedures are relatively flexible, or if legislators agree on what reforms should be undertaken.

Judicial Interpretation

Judicial interpretation constitutes an alternative to amendments as a mechanism of constitutional adaptation (Levinson 1995: 20). A body authorized to arbitrate in constitutional controversies and decide on the constitutionality of laws and executive orders may introduce significant changes to an existing constitution without altering its text. Clearly, just as amendments may be insufficient to overcome a deep constitutional crisis, judicial interpretation may not be an alternative to amendment when constitutional change requires explicit constitutional alterations or when the popular acceptance of reforms demands intervention by elected representatives. Further, given a moderate measure of judicial independence, political actors may be unable to use judicial interpretation to adapt constitutions as they can with amendments. Judicial interpretation is the best means to adapt a constitution to new social values, technological changes, or policy shifts in a gradual, decentralized way.

However, in relation to replacements, judicial interpretation can play a role similar to amendments. The more frequently a constitution is adapted to a changing environment through judicial interpretation, the lower the social or political pressures to replace the constitution should be. This may be more pronounced when amendments are difficult to implement due to the presence of a high number of veto players with conflicting preferences about the content and direction of reforms. Yet formal amendments and judicial interpretation may also reinforce or complement each other as mechanisms of constitutional adaptation. For instance, if courts have strong powers of judicial review and the constitution is easily amendable, judicial interventions may increase amendments by Congress since legislators would often resort to this mechanism to overcome controversial judicial interpretations. In addition, a constitution that incorporates substantive policies may encourage both amendments to incorporate policy shifts and frequent judicial interventions to decide on the constitutionality of legislation (Arantes and Couto 2008). In any event, when adaptation is made by both amendments and judicial interpretation, constitutional durability should be enhanced.

To be sure, it is not possible to observe directly whether judicial interpretation works as an alternative mechanism for constitutional adaptation or how active courts are in constitutional matters, except in single case studies. We can only infer these outcomes across countries by observing the formal powers of courts to interpret the constitution. According to some authors (Elkins, Ginsburg, and Melton 2009: 83), the existence of formal provisions of judicial review should make possible the use of judicial interpretation as a form of constitutional change. These provisions do not, however, provide sufficient variation for comparative analysis. Almost all constitutions in Latin America, for instance, authorize some form of judicial review. In addition, since there are many different forms of judicial review it makes sense to assume that the degree of judicial activism in constitutional matters should be related to the different instruments that expand or restrict the scope of constitutional adjudication.

The crucial variables that capture the importance of judicial interpretation as a mechanism of constitutional change are the scope, access, and effects of constitutional adjudication (Navia and Ríos-Figueroa 2005; Ríos-Figueroa 2011). Judicial interpretation is more likely to be used as a mechanism of constitutional adaptation when there is greater scope for constitutional adjudication to protect individual rights, to resolve constitutional controversies between branches of government, and to rule on the constitutionality of laws or decrees. Constitutional adjudication is also more likely to play this kind of role if both governmental actors and citizens can set in motion a constitutional review process, and if the decisions adopted by constitutional courts are universally valid and do not apply only to the parties involved in a judicial process. The addition of these different dimensions enhances the power of constitutional judges and makes it more likely that they will use judicial interpretation as a mechanism of constitutional transformation.

Observable Implications

The preceding analysis suggests several observable implications about the occurrence of constitutional replacements and amendments. I have argued that the choice between these two mechanisms of constitutional transformation depends, in the first place, on the type of event that triggers constitutional change. In particular, constitutional crises are likely to be addressed by means of replacement, insofar as they signal the failure of the constitutional system to work as a governance structure. Replacement is also likely to occur when the existing constitution becomes obsolete in the face of a deep political change, such as a regime transition. The first hypothesis could thus be stated as follows:

H1: Constitutional replacements are likely to increase with constitutional crises and regime transitions.

Constitutions may also be replaced for strategic reasons, when their design prevents the gradual accommodation of competing political interests to shifting environments. Since amendments and judicial interpretation are the ordinary mechanisms by which the constitution is adapted to changing contexts, the frequency of amendments and the strength of constitutional adjudication should be associated with less frequent constitutional replacements. The replacement of constitutions should also decrease with power-sharing institutions because they create a larger number of actors satisfied with the status quo. This reasoning leads to the following hypotheses:

H2: Constitutional replacements are likely to decrease with the frequency of amendments and the strength of constitutional adjudication.

H3: Constitutional replacements are likely to decrease with the existence of power-sharing institutions.

Just as replacements are appropriate means to address revisions to the basic structure of the state, amendments may be suitable for responding to shifts in the distribution of power among partisan actors. Yet this effect is difficult to predict as it depends on the direction of the shift and on the fit between the existing design and the interests of the main political actors. A more direct relation exists between amendments and the need to introduce changes in matters of procedural detail or policy regulation included in the constitution. From this perspective, amendments should increase with the length and detail of the constitution. As regards the capacity of political actors to implement amendments, I have argued that it depends not only on procedural features but also on the fragmentation of the party system and the degree of conflict among reformers about the content of reform. To be sure, since individual case studies are required to observe the heterogeneity or homogeneity of the institutional preferences of constitution-makers, a testable hypothesis in a large number of cases can only trace the impact of procedural rules and the distribution of partisan power on amendments. These arguments lead to the following hypotheses:

H4: The frequency of amendments is likely to increase with the length and detail of the constitution.

H5: The frequency of amendments is likely to decrease when amendment procedures are rigid and party system fragmentation is high.

The Determinants of Constitutional Change in Latin America

Latin America is an excellent testing ground for these hypotheses. Although the region has been prolific in constitutional change, there is a marked variation in the rate of replacements and amendments within and across countries. While Argentina's 1949 Constitution lasted six years, Colombia's 1886 Constitution lasted 105 years. Uruguay maintained its 1830s Constitution for 86 years, until 1916. Since 1917, however, the average duration of constitutions in that country has been 28.3 years. Similar variation can be found in constitutional amendments. As of 2008, the current constitution of Uruguay has had only 4 amendments in 41 years of life, while the Brazilian constitution has had 16 amendments in 20 years of life. Cross-country variation is also impressive in terms of constitutional design and political conditions.

Dependent Variables: Replacements and Amendments

The focus here is on the determinants of two outcomes: the duration of a constitution until its replacement, and the rate of amendments that a constitution experiences per year of life. In order to explore the mechanisms that explain these two outcomes, I collected data on constitutional replacements and amendments in 18 Latin American countries from 1946 to 2008.¹⁴

To achieve an adequate level of homogeneity among the units of analysis, I have limited my study to constitutions that have been minimally binding in regulating competitive elections for office and decision-making processes.¹⁵ The dataset thus includes only constitutions which were in force during years where presidents and assemblies were elected and more than one party competed in elections.¹⁶ Amendments, in turn, were included only if they were passed by an elected constituent assembly or congress, or approved in a popular referendum.¹⁷

The database contains 738 observations¹⁸ on 46 constitutions, 28 of which were replaced during these years, and 18 of which were still surviving by 2008, when the observation period ends.¹⁹ The sample represents 95 percent of the total number of constitutions in force in Latin America between 1900 and 2008 during years of competitive elections.²⁰ The mean time of survival of the constitutions included in the sample is 22.5 years. The mean number of amendments per constitution is 6, and the mean amendment rate per year of life 0.19.

As regards replacements, the database traces the life of a constitution from its enactment to its replacement.²¹ A constitution is considered to be new when, regardless of the procedure followed for its adoption, its drafters claim it is new, usually by including at the end of the text the abrogation of the previous constitution and all its amendments.²² In doubtful cases, country sources on the evolution and history of constitutions were consulted.²³ If these sources differ about whether a constitution was amended or replaced, I consider a constitution to be new when it is enacted by a popularly elected constituent assembly.²⁴

To count amendments, the database records the number of amendments each constitution experienced per year of life. The relevant outcome is the amendment rate, which is the number of amendments divided by the number of years the constitution has been in force (Lutz 1995: 243). This accurately measures the adaptability of a constitution by means of amendments, and controls for the durability of the constitution. There is some ambiguity, however, about whether amendments should be counted by article, by issue, or by the aggregate reforms approved in a year (see Rasch 2008). I opted for the latter measure because it is less open to interpretation and controversy about the counting rule, and because the institutional determinants of amendments usually remain constant within the same year. The amendment rate in a given year thus ranges from a minimum of 0 to a maximum of 1.²⁵

Explanatory Variables

I start by analyzing the factors that may affect the probability of constitutional replacement. One set of covariates of theoretical interest relates to specific events that may increase the risk of replacements, such as constitutional crises and regime transitions.

In order to rely on objective indicators, I have measured constitutional crises (CONSTCRISIS) by tracing the occurrence of irregular transfers of executive power and extreme forms of executive-legislative conflict in which the chief executive or congressional leaders attempt to terminate the constitutional term of the other branch.²⁶ Both events clearly signal the dysfunctional performance of the constitution. Regime transitions may also render the pre-existing constitution obsolete. However, since democratic constitutions have a stronger claim to legitimacy, the risk of replacement should be greater when the pre-existing constitution was enacted by an authoritarian regime. The variable (TRANSITION) thus takes a value of 0 when there is no transition, 1 when there is a transition and the pre-existing constitution was adopted under democracy, and 2 when there is transition and the pre-existing constitution was adopted by an authoritarian regime.²⁷ Although I do not expect shifts in the distribution of partisan power to have a direct impact on replacements, they may provide incentives to alter the existing constitutional equilibrium. I have thus included a variable (PARTYCHANGE) to test the effects of party change. It takes a value of 1 when a new party or coalition obtains 20 percent or more of the popular vote in legislative elections within ten years of its first appearance in the electoral arena.²⁸ In all these variables the effect of the event is considered to last 5 years from the date of its occurrence.²⁹

The second set of variables of interest is intended to reflect the impact of certain features of constitutional design that may decrease the risk of constitutional replacement. The main variables here are the ease of amendment, the strength of constitutional adjudication, and the power sharing features of the constitution in the areas of electoral competition and distribution of powers.

Different indexes have been proposed to measure the rigidity of amendment procedures (Lorenz 2005; Lutz 1995; Rasch and Congleton 2006). Two basic factors determine the obstacles to amending the constitution: the threshold of votes required in congress, and the number of institutional actors whose consent is necessary for approval (Rasch and Congleton 2006: 335). The problem is that these cannot be combined into a single ordinal scale of rigidity that could capture intermediate cases. For instance, it is not clear whether an amendment requiring a two-thirds majority in a unicameral congress is more rigid than another requiring an absolute majority vote in two chambers or in two different legislatures. Moreover, there is a potential negative correlation between the two measures because amendments passed by only one body (such as a unicameral congress) tend to require a qualified majority vote.³⁰ Given these measurement problems, it is not surprising that one often finds mixed and even contradictory results in studies that attempt to determine which of the proposed indexes of rigidity better explains the rate of amendments (Ferejohn 1997; Rasch 2008).

As a measure of procedural rigidity, I use the number of institutional actors whose consent is necessary to pass an amendment. This measure is intuitive and in bivariate analyses it seems to be a better predictor of the amendment rate than any other measure. The variable (VETOPOINT) is coded as a numerical variable indicating the number of instances an amendment must pass before it can be approved. It ranges from a minimum of approval in one chamber (0) to a maximum of approval in two chambers (or two different legislatures), plus approval by the executive or some additional procedure, such as a popular referendum or ratification by local legislatures (2). Intermediate scores (1) result from the requirement of any two instances of approval.

In order to measure the strength of constitutional adjudication I use Ríos-Figueroa's index of judicial power (2011). This index (ADJUDICATION) adds the number of instruments for constitutional review specified by a constitution, and considers whether adjudication has general scope, whether the effects of constitutional rulings are binding beyond particular cases, and whether the constitutional review process is accessible to all citizens. The index ranges from 0 to 8, with higher scores indicating greater institutional authority of constitutional judges to act as interpreters of the constitution.

To observe the degree of inclusion in electoral rules I focus on the rules to elect presidents: the formula and electoral cycle, the presidential term, and re-election rules.³¹ The formula for presidential election determines the number of candidates competing in this election; and indirectly, in combination with the electoral cycle, it also determines the number of parties competing in the legislative election (Golder 2006). The most restrictive rule is plurality rule with concurrent congressional elections (0); the most inclusive (2) is majority rule; and intermediate formulas (1) are plurality rule with non-concurrent congressional elections and presidential elections by qualified plurality rule (Negretto 2006). Presidential terms and re-election rules affect alternation in power and rotation in office. Presidential terms range from the least pluralist (0) of 6 or more years to the most pluralist of 4 or fewer years (2), with intermediate values (1) of 5 years. Re-election rules range from the least pluralist of consecutive (one or indefinite) re-election (0) to the most pluralist of no re-election (2), with re-election after one or two terms as intermediate (1) rules. Adding these scores together produces an index of electoral power-sharing (ELECTSHARE) ranging from a minimum of 0 to a maximum of 6.

To observe the degree of power-sharing in institutional rules I rely on the central features of the separation-of-powers system: congressional structure, presidential veto, and judicial independence. The first variable captures whether congress is bicameral; the second whether the president has a veto subject to qualified majority override in congress; and the third whether constitutional judges are granted sufficient institutional independence from political pressure.³² Adding these scores together yields an index of institutional power sharing (INSTSHARE) which goes from a minimum of 0 to a maximum of 3.

Five additional control variables are considered. The durability of a constitution may be related to its origins. For instance, there may be a greater incentive to replace constitutions that are established by non-elected authorities or unilaterally imposed by a dominant party as soon as the balance of forces changes. The variable ORIGINS therefore measures the degree of inclusiveness of the coalition that enacts the constitution: for non-elected authorities the value is 0; for a constituent assembly under the control of one party the value is 1; for a coalition of two parties the value is 2; and for a reform coalition including more than two parties the value is 3.³³ The variable DIFFUSION controls for the contagion effect of constitutional replacements in neighboring countries by measuring the percentage of countries in each sub-region (South, Andean, Central, and North) that have replaced their constitutions in five-year intervals.³⁴ LEGACY controls for the influence of previous failures on the probability of replacement and is a numerical variable indicating the number of constitutional replacements in a country in a given year since 1900. INFLATION and GROWTH are continuous variables measuring the average rate of inflation and growth in gross domestic product (GDP) per capita in five-year intervals.³⁵ These variables trace the impact of economic conditions on constitutional stability.

For the analysis of the amendment rate I kept all of the above variables, except the control variables specifically related to replacements, and added others that are relevant to explaining amendments. One of these is the level of detail of the constitution, measured as the words per issue covered in it (DETAIL).³⁶ The longer and more detailed the constitution is, the more likely it regulates procedural details and policy matters, which in turn increases the probability of amendments enacted to adapt the constitution to changing environments.³⁷ The other variable of interest is the level of party system fragmentation (ENPSEATS), which is measured using the Laakso–Taagepera (1979) index of the effective number of parties in the single or lower chamber of congress.³⁸ As I have argued, the rate of amendments should result from the interaction between this variable and the rigidity of amendment procedures (VETOPOINT).

Methods

I use an event history analysis to explore the factors that affect the probability of a country replacing its constitution. This type of analysis allows us to estimate the effect of variables on the hazard rate of an event (Allison 1984; Box-Steffensmeier and Jones 1997, 2004; Box-Steffensmeier and Zorn 2001). Specifically, I use a Cox proportional hazard model (Cox 1972).³⁹ This model does not require specifying a priori the relationship between the event of interest—constitutional replacement—and time, and assumes that each covariate has a proportional and constant effect on the risk.⁴⁰

Amendments deserve a different treatment because they are transformations that constitutions undergo without disrupting their legal continuity. The most appropriate model is thus one in which we can trace the rate of amendments a constitution undergoes in each year of life from its enactment to its replacement by another constitution. To explore the factors that affect the rate of amendments per year of constitutional life, I have used a Tobit cross-sectional time series model, which accounts for the limited maximum variation of the amendment rate.⁴¹

Results

Table 3 shows the results of the proportional hazard analysis of constitutional duration.⁴² All regressions provide robust standard errors clustered by country, to control for possible correlation among observations within each country.⁴³ Results are presented in coefficients instead of hazard ratios. These indicate whether an independent variable increases or decreases the hazard rate, using standard errors to determine statistical significance. A positive (negative) coefficient means that the variable increases (decreases) the hazard rate, in this case the rate at which a constitution is likely to be replaced.

Table 3. Determinants of constitutional replacements in Latin America, 1946–2008

Numbers in parenthesis are robust standard errors clustered by country.

*** p < 0.01; ^**p < 0.05; ^*p < 0.1.

The first model supports the hypothesis that constitutional crises systematically lead to a highly significant increase in the risk of constitutional replacement.⁴⁴ Regime transitions also appear to increase this risk but the effect is not statistically different from zero. This result is probably related to the fact that most of the constitutions surviving by the end of the observation period (2008) have not experienced any event of regime transition. Party system change does not have any discernible significant effect on replacements. As predicted, Model 1 also shows that the risk of replacement significantly decreases the higher the amendment rate is.⁴⁵ A similar effect is observed with the strength of constitutional adjudication and with institutional power sharing.

Model 2 replaces the rate of amendments by the level of rigidity of amendment procedures and shows, as expected, that more rigid amendment procedures significantly increase the risk of replacement. The remaining variables have effects similar to those observed in Model 1, except that now electoral rather than institutional power sharing appears to significantly decrease the likelihood of constitutional replacement. The overall fit of the model is good; we may reject the null hypothesis that the coefficients are jointly zero at the 0.001 level.

Table 4 shows three models for the analysis of the determinants of the amendment rate. The first includes the level of detail of the constitution, amendment procedures and party system fragmentation; the second adds a term for the interaction between amendment procedures and party fragmentation; and the third the remaining variables.

Table 4. Determinants of constitutional amendments in Latin America, 1946–2008

Numbers in parenthesis are robust standard errors clustered by country.

*** p < 0.01; ^**p < 0.05; ^*p < 0.1.

Model 1 shows that the amendment rate tends to increase as the constitution becomes more detailed and decrease as amendment procedures become more rigid. At the same time, party system fragmentation significantly increases the rate of amendments. These results call for interpretation. The likely reason for the positive effect of party system fragmentation on amendments is that as the number of parties and partisan interests in the system increase, there may be more demands for constitutional adaptation through amendment. At the same time, however, a higher level of party system fragmentation should lead to a lower amendment rate if the relevant procedures are stringent.

Model 2 tests this effect by including an interactive term between party system fragmentation and the stringency of amendment procedures. The results show that party system fragmentation increases the amendment rate when the amendment procedure is most flexible (i.e., vetopoint = 0). But it is not apparent what happens when both the amendment procedure becomes more rigid and party system fragmentation increases. Figure 1 illustrates this effect. The solid sloping line shows how the marginal effect of the rigidity of the amendment procedure changes as party system fragmentation increases. The dotted lines around the solid line indicate 95 percent confidence intervals. The rigidity of the amendment procedure has a statistically significant effect on the amendment rate whenever the upper and lower bounds of the confidence interval are both above (or below) the zero line. The confidence intervals show, as expected, that the amendment rate tends to decrease only when amendment procedures become more rigid and party system fragmentation increases; specifically when ENP ≥ 2.5.⁴⁶

Figure 1. Marginal Effect of the Stringency of Amendment Procedures as Party System Fragmentation Increases.

Model 3 is the full model. All the main variables, including the interactive term, maintain the previous effects. Interestingly, none of the political events that were predicted to increase the risk of constitutional replacements significantly increase the rate of amendments. And one such event—constitutional crisis—has a highly significant negative correlation with amendments. As expected, then, extraordinary political events such as constitutional crises provide incentives for constitutional change, more often through replacement than amendment. Similar to the case of replacements, party system change does not appear to have any direct significant effect on the amendment rate.

Model 3 also shows that pluralistic electoral rules increase the rate of amendments. This is consistent with the effect of party system fragmentation because there is a significant correlation between the latter and some components of electoral power sharing, such as the formula for electing presidents. Institutional power-sharing, however, decreases the amendment rate, just as it decreases the likelihood of constitutional replacement. This result may reflect the lower demand for change that exists when decision-making is inclusive. The strength of constitutional adjudication is positively and significantly correlated with the rate of amendments. This provides prima facie evidence that amendments and constitutional adjudication, as I argued before, may reinforce or complement each other as means of constitutional adaptation. The overall fit of the model is good and we can reject the null hypothesis that the coefficients are jointly zero at the 0.001 level.

Conclusion

I have argued that one important reason for replacing a constitution is its failure to work as a governance structure. Certain events demand a revision of the basic structure of the state, thus increasing the probability of constitutional replacement regardless of whether other means of constitutional transformation are available. A deep constitutional crisis is one such event. As I have shown, constitutional crises, in the form of irregular transfers of executive power or extreme executive–legislative conflicts, work particularly strongly against the survival of constitutions. Regime change seems to have a similar effect although its impact is not systematic. The reason is that as democratic regimes become stable, as they did in Latin America following the last wave of democratization, regime transitions as inducements for replacement tend to decline.

These results make sense of recent political developments in the region. Since 1978, open constitutional transgressions in the form of military or civilian coups have been rare occurrences in Latin America. But the region still suffers from governmental instability and inter-branch conflict. In recent years, these events have triggered several processes of constitutional replacement (Peru in 1993, Ecuador in 1998 and 2008, and Venezuela in 1999), suggesting that constitutional instability is likely to persist as a response to the dysfunctional performance of constitutions.

I have also argued that a constitution can be replaced for strategic reasons when its design prevents the accommodation of competing political interests to changing environments. In this regard, this article builds on and expands recent research on constitutional change that points to the importance of constitutional design as an explanation of the rate of constitutional replacements. In particular, it shows that constitutional replacements tend to increase when political actors cannot use amendments or constitutional adjudication as alternative means of constitutional transformation, and when the constitution has a power-concentrating design. These findings open an interesting avenue of research on the factors influencing the choice of constitution makers among alternative forms of institutional design.

Finally, this article also provides an explicit analysis of amendments, arguing that the amendment rate depends both on the length and detail of the constitution and on the interaction between the rigidity of amendment procedures and the fragmentation of the party system. Specifically, it shows that while the amendment rate tends to increase when constitutions are lengthier and more detailed, it is likely to decrease when amendment procedures are rigid and party system fragmentation increases. Whereas particularistic content in the constitution increases pressures for adaptation through amendment, the interaction between amendment procedures and party system fragmentation determines the actual capacity of political actors to use this mechanism of constitutional change.

The preceding comparative analysis is inevitably incomplete. The impact of some potentially important explanatory factors of constitutional survival and adaptation may only be traced in single case studies by means of qualitative analysis. For instance, the perceived legitimacy of constitutional origins may affect constitutional reform strategies. A constitution of revolutionary origins or one sealed by a national pact may be more likely to survive than one that is perceived to be the outcome of a self-interested bargain among political elites detached from the concerns of ordinary citizens. Public trust in representative institutions and constitutional courts may also affect the choice of means to change a constitution. Amendments must usually be approved by elected congresses, which voters in many Latin American countries see as corrupt and scarcely representative of their interests. An equally negative public opinion often exists about the courts that are responsible for interpreting the constitution. In this context, it should not be surprising to observe that both citizens and political elites support the replacement of the constitution in spite of the formal existence of other means of constitutional adaptation.