Advanced Search
  Home  
  News  Expand   News  
    About    Expand     About    
  Projects  Expand   Projects  
  Members  Expand   Members  
  Publications  Expand   Publications  
  Meetings  Expand   Meetings  
  Fellowships  Expand   Fellowships &nbsp
  Prizes  
  Contribute  
  Member Login
Home > Summer 2017 Bulletin > The Time Inconsistency of L...
Download or Print PDF
Summer 2017 Bulletin

The Time Inconsistency of Long Constitutions

George Tsebelis

The articles of the constitution that specify the requirements for amendment have been considered for centuries as “the most important part of the constitution.”1 There are two reasons for the importance of amending clauses in a constitution. First, more stringent requirements will lead to a more rigid constitution. Second, a rigid constitution gives the constitutional courts more discretion to interpret the constitution (without fear that they will be overruled).

Figure 1 provides a graphical representation of these arguments. Let’s assume that constitutional amendment requires the agreement of three different actors (which I call “constitutional veto players,” such as the President, the House, and the Senate).2 The figure represents the ideal points of the three constitutional veto players in a two-dimensional space. Each player prefers points closer to his ideal point over points farther away (giving each a “circular indifference curve,” which is not depicted in the figure). These preferences define the triangle 123, which constitutes the “constitutional core” of the country: that is, the set of constitutional provisions that cannot be amended given the rules of the game, but also the preferences of the constitutional veto players.3 Indeed, at least one of the constitutional veto players will object to a modification of the constitution from one point inside the triangle to another. For example, player 1 would block a movement from point L1 to L2. On the other hand, all three players would agree to a modification of a point outside the core to one within the core. For example, a movement of point J or K to J’ or K’, respectively, is constitutionally feasible and desirable.

Figure 1
Figure 1. Depiction of a Generic Constitutional Core in Two Dimensions

Unlike constitutional veto players, constitutional courts (for example, the U.S. Supreme Court) can unilaterally move (through constitutional interpretation) from any point outside or inside the constitutional core to any point inside the core (such as a movement from point L1 to L2) without the possibility of interference from the constitutional veto players. Movements from outside to inside the core would have the unanimous support of the constitutional veto players, while movements within the core would be unstoppable (because the veto players cannot agree on an alternative movement to the one initiated by the court).

Constitutional change, therefore, can happen in two basic ways. First, it can occur with movements from points outside the constitutional core to inside the core. Such movements can be performed either by the constitutional veto players or by the courts. Second, constitutional change can occur within the core; however, movements within the core can occur only through court decisions. Movements from inside the core to the outside are not possible, as such movements would never be initiated by the political system (as one of the required players would always disagree). Similarly, if attempted by the court, such changes would be aborted by the constitutional veto players.

This analysis indicates that, in democratic countries where the constitution is the basic rule of law, the size of the constitutional core affects whether constitutional modifications will be made through the constitutional court (as interpretations) or by the political system (as amendments). The larger the core, the more discretion is given to the courts, and the fewer possibilities for the political system to amend. For example, the exceptionally large constitutional core of the United States has allowed the Supreme Court to render decisions on major issues such as school segregation, abortion, and marriage equality.

The size of the constitutional core is defined by the constitutional amendment rules and by the preferences of the constitutional veto players. More specifically, increasing the number of constitutional veto players (for instance, by adding a referendum to the amendment requirements), raising the required majorities of collective actors (from, for instance, one-half to three-fifths or two-thirds), requiring a second vote of the same legislature after an election, or widening the differences in the preferences of different actors (political polarization) will each result in an increase in the size of the constitutional core. Consequently, any such conditions render constitutional amendments more difficult to undertake and increase the discretion of constitutional courts.

In the empirical literature, researchers have developed a variety of indicators of constitutional rigidity, which they define generally as the difficulty of constitutional amendment within a country. Most are based on constitutional rules alone, such as the requirement of qualified majorities, the requirement of multiple actors to approve an amendment, or a combination of the two.4 Others combine the rules and the actual frequency of amendments.5 These measures correspond poorly with one another, but they all share the common feature that they present a low level of correlation between the constitutional rules and the actual frequency of amendments. This puzzling lack of correspondence between theoretical arguments and empirical reality has been identified a number of times in the literature, leading some scholars to argue that institutions do not matter at all.6 Others abandon the use of rules altogether and refer only to the (in)frequency of amendments as their measure of rigidity or “entrenchment.”7

However, instead of abandoning the concept of rigidity, I argue it is more useful to focus on the discrepancy between theoretical expectations and empirical reality and to try to identify the conditions under which there is a high level of this discrepancy. Following the analy­sis of constitutional rigidity developed above, I proposed the concept of “time inconsistency”8 as a measure of the discrepancy between the intentions of the framers of the constitution and the empirical reality of the constitutional life a country experiences–that is, the difference between the actual and the expected (on constitutional grounds) frequency of amendments. A high degree of time inconsistency characterizes rigid or “locked” constitutions that nevertheless change frequently.

Characteristic examples of countries with high levels of time inconsistency include Mexico and Brazil, which, despite the stringent constitutional amendment rules, pass many constitutional modifications almost every year (close to one hundred successful amendments in each one of these countries from 1990 until today). These constitutions are highly time-inconsistent, in that provisions have been locked inside these constitutions that are considered inappropriate by the current veto players (concurrent qualified majorities in both chambers, as well as a majority of states in Mexico).

Similarly, the Pinochet constitution currently in effect in Chile, besides being difficult to amend,9 creates a special class of legislation in Article 63 called “organic laws” that function almost identically to formal constitutional amendments. Such laws require a vote by four-sevenths in each chamber of the bicameral legislature and the approval of the President of the Republic. This particularly rigid kind of legislation is mandated in the constitution sixty-nine times.

When writing the constitution, the framers must make a series of decisions: whether to include an item, how detailed the rules related to the new item should be, and how locked (protected) the rules will be from future modifications. The first two decisions are connected with the length of the constitution, and the last with the amendment rules. Time inconsistency emerges by locking too many items, and/or providing too much detail. In Jeremy Waldron’s terms, “any alternative conception that might be concocted by elected legislators next year or in ten years’ time is so likely to be wrong-headed or ill motivated that his own formulation is to be elevated immediately beyond the reach of ordinary legislative revision.”10

Using three different indicators of rigidity,11 I calculate three different measures of time inconsistency (the difference between the actual frequency of the amendments and the institutionally predicted frequency). All these measures are positively and significantly correlated with the length of the constitutions. Figure 2 provides a graphical representation of the time inconsistency as a function of constitutional length using a new measure of constitutional rigidity12 based on a veto-player’s analysis of the amendment provisions of ninety-two democracies.13 This indicator of rigidity, besides covering a larger number of countries than all other institutions-only indicators, produces a small negative correlation with the frequency of amendments (as theoretically expected). Using this measure of rigidity, time inconsistency is calculated as the difference between the actual frequency of amendments and the expected frequency generated from this indicator of rigidity. In the graph, the variables are standardized, so that the slope does not depend on the units of analysis. The reader can verify that this coefficient is highly statistically significant (p=.002), indicating that long constitutions are time inconsistent.

Figure 2
Figure 2. Log Length and Time Inconsistency

For those interested in reducing time inconsistency, there are two different options: decreasing the number of items in a constitution or the amount of detail in each item will reduce the length of the constitution; and unlocking the constitution makes it more flexible. The empirical literature has followed these two directions.

One branch has focused on the frequency of amendments alone, and has noticed that this frequency has increased dramatically after 1950. Researchers have coined the terms “statutory constitutions”14 or “unentrenched constitutions.”15 The underlying assumption is that political time has become denser and more amendments are required. If this assumption is true, the constitutional design should take these findings into account and reduce the locking of constitutions.

The second direction is based on Waldron’s analysis that the combination of long length and locking generates time inconsistency and indicates policy-making through the constitution. If lengthy and locked constitutions are frequently revised despite the difficulties specified in the locking mechanisms, it must mean that they are considered serious impediments in the countries that they regulate.16 In this case, constitutional length should be negatively correlated with undesirable aggregate indicators. Tsebelis and Nardi have found that, in OECD countries, longer constitutions are associated with lower GDP per capita and higher corruption.17 Tsebelis corroborates these findings with data from all (ninety-two) democracies.18 Brown goes one step further: with a time-series analysis on U.S. states for over a twenty-year period, and controlling for economic, demographic, and political variables, the author finds that there is a causal link (in the sense of Granger causality) between longer state constitutions and low GDP per capita.19

In conclusion, time inconsistency is associated with long constitutions. It may be generated out of unforeseeable conditions, in which case unlocking the constitutions would be a good remedy; but it may also have a systematic component, deriving from the excessive zeal of the framers to impose their preferences on future generations and perform policy-making through constitutional means. In this case, reducing the scope and the detail (i.e., the length) of constitutions would be the primary means to remedy time inconsistency.