Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Journal of Theoretical Politics

http://jtp.sagepub.com

How to Measure Constitutional Rigidity: Four Concepts and Two


Alternatives
Astrid Lorenz
Journal of Theoretical Politics 2005; 17; 339
DOI: 10.1177/0951629805052883

The online version of this article can be found at:


http://jtp.sagepub.com/cgi/content/abstract/17/3/339

Published by:

http://www.sagepublications.com

Additional services and information for Journal of Theoretical Politics can be found at:

Email Alerts: http://jtp.sagepub.com/cgi/alerts

Subscriptions: http://jtp.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.co.uk/journalsPermissions.nav

Citations http://jtp.sagepub.com/cgi/content/refs/17/3/339

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


Journal of Theoretical Politics 17(3): 339–361 Copyright & 2005 Sage Publications
DOI: 10.1177/0951629805052883 London, Thousand Oaks, CA and New Delhi

HOW TO MEASURE CONSTITUTIONAL RIGIDITY

FOUR CONCEPTS AND TWO ALTERNATIVES

Astrid Lorenz

ABSTRACT
This article deals with the (possible) explanation of constitutional development
by the rigidity of the constitution itself, for example the difficulty of the amend-
ment process. Four concepts of measuring rigidity are tested theoretically and
their respective outcomes concerning the frequency of constitutional reforms
are explained. The strikingly different results show how important it is to
discuss and to test operationalizations of abstract terms.
KEY WORDS . constitutional amendments . constitutional rigidity .
constitutional politics . index construction

In political science, changes to already existing constitutions in demo-


cratic states are seldom analyzed on the basis of a multi-case comparison
(Bogdanor, 1988: 1) although they seem to be an everyday norm: 32 of all
39 established democracies have amended their constitutions between 1993
and 2002, 18 of them five or more times. According to my calculations, the
average rate amounts to 5.76 reforms in ten years. Three states (Switzerland,
Finland and Poland have even introduced new constitutions (Lorenz, 2004:
Table C2).
Perhaps due to the very lack of knowledge about constitutional change,
the stable, normatively regulating and orientating function of constitutions
(Göhler, 1997), along with their irreversibility as a basic principle of ration-
ality (Goodin, 1996), is often overemphasized. Dieter Grimm, a German
expert on constitutional law, also criticizes the often ‘both undisputable and
nearly empty statements’ on constitutional development, citing a colleague
who writes that ‘the vitality and effectiveness of a constitution is based on
its ability to join together with the lively and spontaneous forces of the
times and to adapt the individual facts of the present to the future’ (Grimm,
1994: 316).
Donald S. Lutz (1994), Arend Lijphart (most recently 1999), Dag Anckar
and Lauri Karvonen (2002) and the author herself have studied various
aspects of constitutions and their changes. Among other things, they have
all developed indices for measuring constitutional rigidity as a precondition

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


340 JOURNAL OF THEORETICAL POLITICS 17(3)

for their quantitative analyses. The following discussion of their different


concepts may help to decide how best to measure constitutional rigidity by
presenting one progressive step forward towards filling the gap in knowledge
about constitutional politics in stable democracies.

Aims and Results of the Studies

Donald Lutz starts his analysis with the premise that institutionalists have
not pursued the empirical verification of their predicted consequences with
much vigor. After having studied 82 constitutions, he states that the rate
of amendment correlates highly with the degree of its difficulty. Nevertheless,
‘beyond a certain point, making the amendment process more difficult is an
‘‘inefficient’’ way to keep the amendment rate in a moderate range’ (Lutz,
1994: 365). Instead, Lutz explains the variance of the amendment rates by
the interaction of the length and the rigidity of the documents. He underlines
the inter-relationship of these variables with other constitutional institutions,
e.g. the operation of the legislature or the court. Ultimately he recommends
avoiding the extremes of either legislative dominance or referendum strate-
gies and instead combining either the legislative complexity or the interven-
ing election strategy with a relatively short document. Lutz concludes that,
first, institutions matter and, second, that they operate more or less indepen-
dently of cultural, economic or other structures or actors.
According to Arend Lijphart, constitutional rigidity is just one of a
complex set of variables which he uses to test the character and performance
of different patterns of democracy (consensus and majoritarian model).
Combined with the strength of judicial review, rigidity is seen as the central
explanatory constitutional variable. In Lijphart’s opinion, they both belong
to what the author calls the ‘federal–unitary dimension’ (Lijphart, 1999: 216).
Lijphart sees constitutional rigidity and judicial review as antimajoritarian
instruments while non-rigid constitutions without judicial review, to him,
permit unrestricted majority rule. As he has hypothesized before, the results
of his study show that rigid constitutions tend to have more judicial review
protection than more flexible ones, although the correlation coefficient is
only 0.39 (Lijphart, 1999: 229). Lijphart states that rigid constitutions are
linked with the other two variables of the federal–unitary dimension:
bicameralism and independent central banks.
Dag Anckar and Lauri Karvonen do not treat the amendment methods as
an independent variable but want to explain why countries prefer different
basic types of constitutional rigidity. Like Lutz and Lijphart, Anckar and
Karvonen reduce the great variety of amendment provisions to some basic
types and try out a set of explanatory factors, e.g. the age, the length and

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 341

the level of detail of a constitution, the present model of democracy (consen-


sual or majoritarian), the degree of ethnic and cultural heterogeneity, the size
of a country, the colonial heritage or the fact that it belongs to a certain world
region. In the end of their analysis, they state that the rigidity variation
cannot be systematically explained by any of the ten tested factors and
that the choice of an amendment procedure does not mirror a broader and
reflected constitutional orientation and strategy. There is an overall prefer-
ence for more rigid documents, as Lijphart had already observed before.
The authors suppose that the rigidity seems to perform better as an indepen-
dent rather than as a dependent variable.
My own research project examines the explanatory power of a set of insti-
tutional and contextual variables according to the frequency and significance
of constitutional reforms in different democratic political systems. The formal
difficulty of the amendment process is one of these variables. However, initial
tests indicate that the rigidity of a constitution may explain the principal
inclination of a political system to amend its core document but that it
fails to predict the frequency of those amendments.

Methodological Differences

All authors mention the overwhelming variety of national constitutional


amendment procedures including alternative provisions concerning different
articles or sections of the document(s), concerning normal or rapid pro-
cedures, or depending on who has initiated the reform. This variety presents
a methodological challenge. To limit the problems, all rigidity concepts dis-
cussed here refer to formal amendment provisions and not to alteration
mechanisms or implicit change, e.g. by judicial review or new conventions.

Donald S. Lutz
Lutz defines 68 empirically found actions that could be used in combination
to initiate and approve constitutional amendments and that ‘together cover
the combinations or virtually every amendment process in the world’ (Lutz,
1994: 361). Some basic scores of his index of constitutional rigidity are taken
from an empirical study of the US states’ constitutions, which in the period
between 1776–1991 are assumed to have had a similar context of variables
and, therefore, guarantee a quasi-experimental situation. First, Lutz studies
the actual success rate of different kinds of necessary majority votes or other
provisions. Then he divides the success rate of reform proposals initiated by a
simple (bicameral) legislative majority by the success rate of proposals
initiated by a two-thirds majority, a three-quarters majority, etc. Thus, the
index scores for different possible ‘actions’ are generated. Afterwards, the

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


342 JOURNAL OF THEORETICAL POLITICS 17(3)

author’s catalogue of actions is supplemented by other requirements found in


32 other selected national constitutions. These are assigned scores which,
according to Lutz (1994: 361), are ‘reasonable in the context of the other
index scores’ or because of some theoretical premises. In order to measure
the constitutional rigidity of a case, one has to add up the score values for
each necessary amendment action.
At first glance, this detailed approach seems to be precise and realistic but
the genesis of the Lutzian scores is not without problems. Some of the num-
bers depend completely on the US evidence. The other part of the scores is
estimated, e.g. respective unicameral legislatures, which the author could
not observe in American states. Estimating values is possible and a common
practice. But it makes the exactness of scores fictitious. In the end, many of
them are similarly ‘handmade’ like, for instance, the indices presented later.
The specific way in which scores are generated could endanger the universal
usability of the concept. Since the catalogue contains diverse but not all
theoretically possible actions, the rules for estimating missing values could
be described more precisely. I personally failed to apply the Lutzian index
successfully to measure the rigidity in countries which are not mentioned
by him because he had not offered scores for many of their procedure devices.
Apparently, concepts of rigidity have to be left open for the endless creativity
of constitutional engineers.
Lutz’s test of his index using data from 32 constitutions shows a very high
correlation coefficient of 0.7970 between the amendment rate and the index
of difficulty. His explanation of the amendment rate by only two variables
(rigidity and length of the constitution) results in a unique adjusted R 2 of
0.93, i.e. 93 percent of the variation is explained. However, taking into con-
sideration that the Lutzian choice (and non-choice) of constitutions and of
time periods is not based on strict premises, this should be re-examined.
Some subsequent tests of all indices will therefore, inter alia, compare their
explanatory outcomes concerning all fully established democracies in the
same period of time.

Arend Lijphart
Lijphart’s concept reduces the array of amendment provisions to a four-
point scale of rigidity. The classes are defined as approval by an ordinary
majority as the easiest way (1.0); by more than ordinary but less than two-
thirds majority or ordinary majority plus referendum (2.0); by two-thirds
majorities or equivalent (3.0); and by more than a two-thirds majority or a
two-thirds majority combined with other requirements, such as approval
by state legislatures (4.0). In every case, Lijphart considers the least restrain-
ing legal method of amendment. This simplifies the measurement and is
empirically justified because most reforms seem to be realized by the easiest

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 343

possible way. This is even conceded by George Tsebelis (2000), who, in his
own approach, refers to the maximal barrier for decisions, e.g. who could
veto.
However, this rule competes with another one: when different provisions
apply to different parts of the constitution, the provision pertaining to amend-
ments of the most basic articles should be counted. One must admit that
many constitutions contain specific, more rigorous procedures for amending
such core sections. Whether this counting rule gives an adequate picture or
not depends on the aim of the analysis. When we seek to measure the rigidity
of the basic principles of a political system, this approach can be right. But
what about, for instance, in Germany, where the meaning of some clauses
is completely unchangeable? In his own measurement, Lijphart does not
take this eternity rule into consideration when he assigns the Grundgesetz
a rigidity of 3.5. Indeed, the decision to do so gives a more adequate picture
of German basic law.
If we want to reflect realistically on the development of entire constitu-
tions, then Lijphart’s second measurement rule can distort the picture,
because many, if not most, recent constitutional reforms have not, in fact,
changed the basic principles of the constitutions (Lorenz, 2004). That is
why it is problematic to correlate, for instance, the frequency of amendments
with the Lijphart scale. An impressive example is Canada which has a very
rigid constitution, assigned by Lijphart the maximum of 4.0 points. How-
ever, none of the seven constitutional amendments in the period between
1993–2002 had to pass this high barrier. So it may be misleading to explain
the number of reforms by the maximum rigidity. This, of course, is not a
critique of Lijphart who had other aims in his analysis.
Doubts have provoked the author to classify plurality systems with a
supermajoritarian amendment rule in the category below the one to which
they technically belong because the supermajority device is much less con-
straining there than in proportional representation systems. I tested all ten
states with plurality electoral systems which have been fully established
democracies from 1993 through 2002 (according to Freedom House).
During this entire period, there has been a party holding more than 50 per-
cent of the mandates in only five of these cases. No party has reached a
majority of parliamentary mandates in Papua New Guinea, Australia (1998
until now), France (until 2002), Mauritius (1991–95) and Mongolia (1996–
2000). Obviously, the correlation between the voting system and actual
majorities is overestimated. Heterogeneous societies, social disaggregation
and declining party commitment, i.e. trends which are characteristic of
most contemporary democracies, counteract supermajorities. So if we want
to consider the real allocation of political power in an index of constitutional
rigidity, then it is more precise to measure it empirically. That, of course,
would endanger the elegant simplicity of the concept and the validity of

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


344 JOURNAL OF THEORETICAL POLITICS 17(3)

case classifications concerning longer time periods. I prefer to construct a


purely constitutional variable which may be tested alone or together with
other independent variables, e.g. the presence of absolute majorities.
Another problem with the Lijphart model is the unsystematic considera-
tion of referenda. The author mentions in passing that he assigns the
approval of an amendment by more than an ordinary and less than a two-
thirds legislative majority a score of 3.0 if a referendum is necessary. Besides
this, referenda are beyond Lijphart’s interest although many amendment
provisions require them. It seems to be evident that plebiscites are real addi-
tional barriers to constitutional reforms, especially the larger and more
complex the state (see Anckar and Karvonen, 2002: 11). The first Danish
referendum on the treaty of Maastrict produced a negative result, Norwegian
people rejected integration into the European Union (EU) in 1994 and, in
several candidate countries, it was unclear whether the people would approve
accession to the EU. In Portugal, a 1998 referendum narrowly defeated a
proposal to allow abortion on demand. Three constitutional amendment
proposals were rejected in Ireland in 1992 and 2002. These examples under-
line the fact that no matter where a referendum takes place, it may fail,
especially in view of the perceived distance between political elites and society
in many modern democracies. In my opinion, this should be systematically
considered by the index construction.

Dag Anckar and Lauri Karvonen


Dag Anckar and Lauri Karvonen do just this. They establish a two-
dimensional typology of amendment methods by distinguishing between three
kinds of legislative votes (none, ordinary, qualified majority) and the same
kinds of popular votes. A constitutional amendment may be voted on by
‘none’ (1 point), an ordinary majority in the legislature (2 points), a qualified
majority (5); by an ordinary majority vote of only a referendum (3) or a
qualified majority of only a referendum (7); by an ordinary majority of both
the legislature and a referendum (4), by a qualified majority of the legislature
combined with an ordinary majority vote in a referendum (6) or with a
qualified majority in a referendum (9) or an ordinary majority of the legisla-
ture combined with a qualified majority of a referendum (8). Like Lijphart,
they measure the rigidity of the constitutional core and not the rigidity of
the constitution as an entire document. Against the background of the
empirical evidence, one can expect that this will make a rather big difference,
as I argued earlier.
Anckar and Karvonen’s concept is clearer than the Lutzian one and more
precise than Lijphart’s because it systematically considers whether and which
majority is necessary in a referendum. Unfortunately, the breakdown of
the types of votes is not convincing. Non-approval by the legislature is too

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 345

small a category when we have only three to begin with. Furthermore, the
combination ‘no legislative approval’ plus ‘no referendum’ is unrealistic in
democracies (which is mentioned by the authors themselves). Finally, the
category of qualified majority seems to be too broad. Therefore, the very
uneven distribution of the 85 empirical cases is not surprising. While 76.5 per-
cent of the cases are concentrated in only two categories (5 and 6), two
categories have no cases and two other categories have only one or two
cases. The authors themselves state that their results (no impact by the ten dif-
ferent independent variables on the choice of amendment types) are already
methodologically grounded because of this uneven distribution of cases.
Moreover, the term ‘inclusion of the people’ (Anckar and Karvonen, 2002:
Table 1) for one of the two dimensions is not very precise. In Canada, for
instance, an ordinary majority of the Province’s parliaments must approve
a constitutional amendment, which represents at least 50 percent of the
entire population. Anckar and Karvonen handle this like a vote in a referen-
dum. One may argue that the provincial level is an additional voting arena
that should be considered for measuring constitutional rigidity but the
label ‘inclusion of the people’ is problematic.
The authors have noted many of these critical points and reacted by
regrouping and renaming the categories and forming broader classes so
that each comprises a larger number of cases. Thus, the new model becomes
simpler: the first dimension distinguishes between cases in which legislative
decision amends the constitution and cases that require additional approval
by a referendum or state ratification. The second categorization refers to the
kind of majority vote but is now divided into ‘flexible amendment’ (including
ordinary majority), qualified decisions and more rigid procedures. This
redesign of the model is much more convincing but still not without problems
if we seek a universal model for measuring rigidity.
Anckar and Karvonen, for instance, differentiate between parliaments and
referendums but not between different houses within parliaments. If differen-
tiation is the aim, then why stop with plebiscites? It seems obvious that it is
more difficult to reach two majorities than one, especially when the members
of the houses are elected in different ways. This causes varying interests and
role perceptions of mandate holders and the chance for a diverging relative
strength of parties in both houses. Thus, the passage of a law by the
German federal chamber, the Bundesrat, presents a rather strong barrier
because the opposition party often dominates.
Probably, Anckar and Karvonen ignore this fact because, to them, the
rigidity is the dependent variable. In bicameral systems, the provision that
an amendment must be passed by both houses is really less important than
the decision to require an additional referendum, while, in unicameral sys-
tems, it is completely irrelevant. How many houses must pass an amendment
seems to be more a function of the formal structure of the political system

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


346 JOURNAL OF THEORETICAL POLITICS 17(3)

than an independent decision. But if we try to explain the amendment rate by


the difficulty of reaching all necessary assents, then the number of houses as
different voting arenas is significant in my opinion. This argument empha-
sizes that every index is constructed on the basis of a specific general frame-
work of analysis.

Astrid Lorenz
The presented concepts of constitutional rigidity vary in many respects (see
Table 1). My own approach combines some ideas that have been discussed
and positively assessed earlier. It integrates the type of majority rule with
the number of voting arenas or actors. For measuring the constitutional
rigidity, I add up scores for the necessary majority in each of the necessary
arenas with different voters, e.g. in the parliament, in each chamber of a
bicameral legislative or in a referendum. ‘In each arena with different
voters’ implies that a legislature has to be assigned a score twice if it has
been newly elected between the votes and once when two or more votes take
place in one legislative term. In contrast to the last two concepts, I strictly
assess the easiest procedure for amendment. For scoring the majority rule in
each voting arena, I use the Lijphart scale from 1.0 to 4.0, modified in one
respect: I treat a three-fifth majority like two-thirds majorities (3.0 points)
and assign four points to majorities greater than three-fifths. This affects,
for instance, Benin where a four-fifths majority is needed for approval.
Combining Lijphart’s majority scale with the systematic consideration of
different voting arenas implies that non-parliamentary organs or actors
must be considered when they explicitly have to consent to an amendment.
A constitutional court, a government or the president are assigned 1.0
(for a simple majority vote). I do not consider whether a required assent is,
de facto, ritualized or really difficult to achieve, because in the case of
doubt or a crisis every player will strive to use all of her formal veto potential.
According to Lijphart’s scale of majorities, an approval by referendum may
be counted as two points (ordinary majority) or more, when a larger than

Table 1. Distribution of Cases

Index of rigidity Cases Mean Standard Minimal Maximal


observed deviation value value

Lutz 22 2.61 1.37 0.5 5.1


Lijphart 26 2.77 0.89 1 4
Anckar & Karvonen 39 4.97 1.27 2 9
Lorenz 39 4.85 2.19 1 9.5

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 347

absolute majority is necessary. In Namibia, for instance, where a rare two-


thirds referendum majority is required, this action is scored 3.0 and not 2.0.
‘Pre-final votes’ are an exception to this scoring rule: if a required vote does
not relate to the final document but only to the declaration of a need to
amend the constitution, a first, still general draft or similar, then this vote
is scored with half of the normal points for the respective action. According
to this logic, if an amendment project has to be introduced during a legisla-
tive term before the final vote takes place (as in Norway), then it receives 0.5,
i.e. half of the lowest majority score. This 50-percent rule is based on the
premise that majorities for draft documents are easier to achieve. I have to
point out that this rule does not refer to the first vote on a bill which has
to be voted again after an election but whose wording has already been
finalized. Such a first vote would be assigned the regular score corresponding
to the necessary majority.
Thus, the index uses differentiated scores instead of classes, like Lutz, a
relatively simple method of assigning values to necessary majorities, like
Lijphart, and it also systematically considers different voting arenas, like
Anckar and Karvonen.

Differences in the Samples


The authors applied their concepts to different samples. The broadest choice
of cases is found in that of Lutz who studies ‘true constitutional systems’, in
which ‘constitutions are taken seriously as limiting government and legiti-
mating the decision-making process they describe, important modifications
in the operation of the political system need to be reflected in the constitu-
tion’ (Lutz, 1994: 356–7). He first analyzes 50 US states and, at a second
stage, 32 other cases (including the USA as one case) in very different time
periods, e.g. Argentina 1853–1940; Norway 1814–1982; France 1968–92;
Western Samoa 1962–84; Germany 1949–92. Obviously, a country does
not have to be a consolidated democracy to be included in the Lutzian
sample. This and the calculation of national amendment rates for very differ-
ent time periods raise questions concerning the meaningfulness of the results.
Lijphart explicitly deals with democracies but has a quite pragmatic under-
standing of what a democracy is. Like Lutz, he assigns the national constitu-
tions index values for a relatively long period, 1945–96. This is justified with
his observation that constitutional amendments tend to be quite stable and
with his rather simple four categories which are quite elastic in relation to
changes in the amendment procedures.
Anckar and Karvonen choose their cases on the basis of the Freedom
House ratings for the year 1999. They analyze the constitutions of 85
states that have been classified as free.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


348 JOURNAL OF THEORETICAL POLITICS 17(3)

The author makes the narrowest choice. In order to be included in her


sample, a country has to be classified by Freedom House (http://www.
freedomhouse.org/ratings/index.htm) as free not merely for one year but
rather for the entire period of 1993–2002. This should guarantee the presence
of stable democratic standards. Moreover, the country must have a popu-
lation over one million to guarantee a minimal degree of complexity of the
political system. Finally, it is required that there have been no civil wars,
revolts or states of emergency on the national level because these can strongly
affect constitutional politics and, thus, minimize the potential of explanation
of the tested variables. These conditions of choice result in a sample of
39 cases.
The different operationalizations, different samples, and different periods
of analysis (for a summary, see Table A1) may be expected to influence the
results of analysis. This will be tested in the following sections.

Correlations Between the Indices

The first test compares how the constitutional rigidity of the same set of
countries is valued by the different indices. How does one choose these
countries? According to my own approach already mentioned, I chose only
those cases which have been fully established, peaceful democracies between
1993–2003. As already stated, this guarantees that the institutional variable
is really respected by the national actors. All 39 countries chosen are covered
by at least two samples. Other data provided by the authors are ignored,
e.g. concerning Andorra, Belau, Brazil, Cyprus, Guyana, Kenya, Kiribati,
Malaysia or Micronesia. Table 1 contains some basic data concerning the
distribution of cases.
As Table 1 reveals, the indices of Anckar and Karvonen (first version) and
Lorenz result in the broadest ranges, probably because the absolute numbers
of the scores and classes are higher than those assigned by Lutz and Lijphart.
The addition of scores for necessary actions, a method used by Lutz and
Lorenz, seems to lead to a larger standard deviation than closed scales
because of the creativity of the design of constitutional amendment pro-
cedures. Thus, the addition of scores provides a more realistic picture of
diversity by considering demanding requirements beyond majority pro-
visions but at the same time it may cause other problems, e.g. an uneven
distribution of cases or more statistically deviant cases.
Table A2 shows the absolute values of constitutional rigidity, the respec-
tive rank, and the average and rank based on the standardized indices for
all 39 countries. The rankings are, of course, less precise but do facilitate a

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 349

Table 2. Pairwise Correlation between Indices, Based on Table A2

Donald Arend Dag Astrid Average Imputed


Lutz Lijphart Anckar Lorenz of stan- factor
& Lauri dardized
Karvonen values

Donald Lutz 1.00

22
Arend Lijphart 0.53 1.00
0.01
21 26
Dag Anckar & 0.33 0.46 1.00
Lauri Karvonen 0.14 0.02
22 26 39
Astrid Lorenz 0.60 0.61 0.37 1.00
0.00 0.00 0.02
22 26 39 39
Average of 0.77 0.82 0.70 0.83 1.00
standardized values 0.00 0.00 0.00 0.00
22 26 39 39 39
Imputed factor 0.78 0.83 0.61 0.88 0.99 1.00
0.00 0.00 0.00 0.00 0.00
22 26 39 39 39 39

quick overview. They show that some cases are assessed very differently:
Australia takes rank 6.5 according to Anckar and Karvonen but rank 36
according the standardized indices. Denmark receives rank 6.5 according
Lijphart in contrast to Anckar and Karvonen (33.5) and Lorenz (34).
Sweden takes rank 2 according to Anckar and Karvonen but rank 17.5
according to Lorenz.
Since majority votes are assessed in relatively similar hierarchies by all
authors, the greatest variations concern those cases whose constitutional
amendments require other actions, e.g. a referendum, elections between the
votes, the consent of federal units or further provisions. This is also reflected
by the results of the second test. Table 2 shows varying correlations in pairs
between the four indices. The highest correlation coefficient exists between
the indices of Lorenz and Lijphart (0.61) and Lutz and Lorenz (0.60). It is
interesting to note that the simplest, abstract index of Lijphart and the
subtly differentiated, partly empirically based index of Lutz correlate rela-
tively strongly with 0.53. The values assigned by Anckar and Karvonen
show the weakest relationship to the other indices, including the average of
standardized values and the imputed factor.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


350 JOURNAL OF THEORETICAL POLITICS 17(3)

A principal component factor analysis1 of the four indices results in


one important factor. This shows that all indices measure the same latent
construct and confirms in this way the impression given by the qualitative
analysis of the index constructions. The generated factor explains 62.32 per-
cent of the whole variation.2 In addition, using the regression method, factor
scores were calculated for the 21 countries that had no missing values on any
of the four indices of rigidity. In a way, these factor scores are similar to the
average of the standardized indices. However, more weight is given to those
indices that have a higher correlation with the other indices. The factor scores
calculated in this way correlate highly with the average standardized indices
ðr ¼ 0:998Þ and the single indices. Moreover, the best subset regression
method allows estimating (imputing) values for those countries for which
no data are available (Table A2).3 The imputed factor which now covers
all 39 cases correlates marginally stronger with the rigidity indices than the
average of standardized values, except for the index of Anckar and Karvonen
(see Table 2).
The average of standardized values, such as the imputed factor, are
included in the following tests as theoretically possible alternatives to the
basic indices. They represent methods which are often used in order to
reduce errors.
The next section shows how the indices influence the statistical relationship
between constitutional rigidity and the amendment rate.

How the Indices Impact on the Results of Analysis

First of all, the amendment rates have to be ascertained. I decided to analyze


two rates: the Lutzian, based on different individual time periods, and the
amendment rate between 1993 and 2002, which I calculated myself. Because

1. The principal-component factor-analysis (based on a principal-components analysis) is a


specific model of the statistical package ‘stata’ (version 7). It includes an implicit model of mea-
surement error. The factor values show minimal deviation from other models with explicit mea-
surement error, such as the Maximum-Likelihood-Factor-Analysis ðr ¼ :99Þ.
2. A similar analysis using pairwise (instead of listwise) deletion of cases results in the same
factor explaining 61.67 percent of the variation.
3. The statistical package ‘stata’ (version 7) was used to estimate the missing values. ‘Stata’
performs imputation by best-subset regression. First factor scores are calculated from available
values of the rigidity indices using principal component factor analysis. In a second step missing
factor scores of the new variable, due to listwise deletion, are imputed using the prediction of the
best available subset of otherwise present data. Stata organizes its calculations by patterns of
missing data, so the missing-value regressions can be conducted efficiently. For more details,
see Goldstein (1996). For a discussion of different methods of imputation, see Little and
Rubin (1987).

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 351

of his deviating choice of cases, Lutz provides data for only 22 of the 39 fully
established democracies that are covered by my sample.
Table A3 contains the very different rates. It is interesting that they do not
show a strong relationship. The coefficient of the paired correlation between
the amendment rates is 0.35. In 14 of the 22 cases, the rates concerning 1993
through 2002 are lower and, in one case, it is the same (Japan). The amend-
ment rates seem to have increased only in Chile, Costa Rica, Finland,
France, Ireland, Italy and Switzerland. The data suggest that constitutional
modifications are rarer today. According to scholars who observed a trend of
convergence of normal legislation and constitutional policy (Colomer, 2001:
238f.; Riker, 1980), this could be surprising. Their thesis, if true, may be valid
only for contemporary democracies in the last decades. But again, the data
have to be treated with care.
My spot check of the Lutzian data produced a rather negative result. I was
only able to verify the rates for Germany, France and Ireland, and found
them incorrect. I calculated 0.24 instead of 0.55 for Ireland (www.taoiseach.
gov.ie/); 0.88 instead of 2.91 for Germany (www.verfassungen.de/de/
gg-index.htm); 0.13 instead of 0.19 for France (www.verfassungen.de/f/
fverf58-1htm), considering in every case the time periods used by Lutz.
This phenomenon may probably be explained by the difficulty of finding sys-
tematical information about constitutional reforms. Another cause for rate
differences could be a diverging understanding of amendments or of what
the constitution is. In New Zealand, for instance, four documents are
commonly regarded to have constitutional status. Nevertheless, the Imperial
Laws’ Application Act 1988 allows it to assign this status to even more
laws. That could be one reason explaining why the amendment rates for
New Zealand differ so much with 13.4 by Lutz and 0.1 by my calculation
for the period 1992–2003. Since the results of analysis depend much on the
data used, this should be kept in mind.
Following the present data, the weakest relationship between constitu-
tional rigidity and Lutz’ amendment rates is produced by Lijphart’s index,
with a correlation coefficient of 0.37 and the strongest by the index of
Lutz himself (0.75, see Table 3). This is interesting since, as reported, the
indices of rigidity of these authors (applied to the analyzed countries) corre-
late relatively strongly.
The much weaker relationship between rigidity and amendment rates
is particularly eye-catching in the period 1993–2002 when the indices of
Lijphart (0.05), Lutz (0.2) and Lorenz (0.15) are used. Again, the
Lijphart index produces the weakest correlation and indicates no (linear)
relationship between difficulty of amendment and the reform rate. According
to these results, the rigidity of constitutional documents does not explain the
amendment behavior of democratic states very well. This is sharply con-
trasted by the index of Anckar and Karvonen which results in a correlation

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


352 JOURNAL OF THEORETICAL POLITICS 17(3)

Table 3. Correlations in Pairs between the Amendment Rate and the Constitutional
Rigidity Indices

Indices of constitutional rigidity

Lutz Lijphart Anckar & Lorenz Average Imputed


Karvonen values of factor
standard- values
ized indices

Amendment rate 0.75 0.37 0.45 0.59 0.69 0.69


(Lutz) 0.00 0.10 0.03 0.00 0.00 0.00
22 21 22 22 22 22
Amendment rate 0.2 0.05 0.51 0.15 0.31 0.24
(1993–2002) 0.35 0.80 0.00 0.38 0.06 0.15
22 25 38 38 38 38

coefficient of 0.51, more than twice the size of the others, and which is the
only one producing quite similar results for both diverging rates.
Anckar and Karvonen’s index significantly increases the correlation co-
efficients produced by the average of standardized indices and the imputed
factor. Irrespective of how the index is finally assessed, this refers to the
fact that using any kind of average instead of the basic data may not
always be the best way to reduce errors. At least, it does not release
anyone from the duty of calculating the general suitability of each basic
index.
Another step tested Lutz’s hypothesis that the rigidity and the length of
constitutions explain the amendment rate. For calculating the results con-
cerning the Lutzian amendment rate, the number of words of the constitu-
tions listed by him is used (Lutz, 1994). They refer to the already amended
documents in the end of the time periods. Concerning the amendment rate
1993–2002, the number of lines was used, based on equally formatted consti-
tutions in their English version valid in 1993.4 One may expect that the rela-
tionship between words and lines is very high, so this difference will not
undermine the comparability of the results. In fact, my test of 23 cases pro-
duced a correlation coefficient of 0.83. In order to improve the preconditions

4. One exception concerns Italy for which I could not find an official consolidated version
which was valid at the beginning of the time period tested. The number of lines mentioned
refers to the version of 2003. Since there took place five simple and two significant constitutional
amendments before, this is a potential source of errors. However, the number of cases is expected
to outweigh its possible influence.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 353

2
Table 4. Adjusted Determination Coefficient Radj in the Regression of Length and
Rigidity to the Amendment Rate

Lutz Lijphart Anckar & Lorenz Average Imputed


Karvonen of stan- factor
dardized
indices

Amendment rate 0.95 0.82 0.77 0.77 0.77 0.77


(Lutz) 22 21 22 22 22 22
Amendment rate 0.03 0.06 0.19 0.02 0.07 0.04
(1993–2002) 22 25 38 38 38 38

for regression (and as done by Lutz), the values of the rigidity were pre-
viously transformed linearly, dividing one by the rigidity.
Table 4 shows that the explanatory power of the model is high concerning
the amendment rates of Lutz but that constitutional rigidity and the length of
the constitutions do not explain the amendment rates 1993–2002. Regarding
Lutz’s rate, all rigidity indices result in high relationships. The highest are
produced by the most complex index of Lutz (0.95) and by the simplest
index of Lijphart (0.82). Quite the opposite effect may be observed concern-
ing the contemporary amendment rate. Here, only Anckar and Karvonen’s
index produces a better result (0.19). However, the 95 percent explanation
of rate variation presented by Lutz in 1994 is far from being achieved.
If applied to only those 22 fully established democracies which were covered
by Lutz’s sample, the imputed factor results in a similar adjusted determina-
2
tion coefficient Radj of 0.00. This means that the function of regression
would be unable to predict or explain the amendment rate if the same regres-
sion was estimated on the basis of a new sample with the same number of
cases.
Why do these difference exist? Perhaps the growing influence of other inde-
pendent variables, e.g. of political constellations, undermines the universal
validity of Lutz’s model of explanation. But methodological differences
may also affect the outcomes. So the amendment rates have been calculated
on the basis of different time periods in the first and identical periods in the
second case. Moreover, the length of the constitutions has been measured in
different ways. Lutz’s method of explaining the amendment rate by the
length of the product of this amendment rate, i.e. the already amended docu-
ment at the end of the time periods, confuses the causalities. In contrast to my
preliminary test of the relationship between words and lines of the same
documents, an assessment of the 21 cases covered by both samples resulted
in a surprisingly low correlation coefficient of only 0.19 between the words
of Lutz and the lines of my time period, so perhaps even the text basis was

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


354 JOURNAL OF THEORETICAL POLITICS 17(3)

different. My numbers of lines include, for instance, schedules and transi-


tional provisions, because they form important parts of the documents:
schedules often contain bills of rights and Latin American constitutional
amendments often relate to transitional provisions which contain, inter
alia, policy goals.
Finally, the general potential of the number of words or lines as an indi-
cator could be questioned. Anckar and Karvonen operationalize length as
the number of articles. Both can be seen as two dimensions of the worth test-
ing intensity of rule-setting. A combined index of articles and lines (or words)
could measure the general amount of rules such as their extent of detail. The
Irish constitution, for instance, needs 1173 lines for 50 articles, Lithuania’s
accommodates 154 articles in a comparable amount of space (1152 lines).
The premise of the proposed additive index of similarly weighted values
would be that both a higher number of lines and a higher number of articles
(or sections) increase the probability of amendment. Since no data are avail-
able concerning the numbers of articles in Lutz’s time periods, I abstained
from testing the rigidity indices and the intensity of rule-setting as explana-
tory variables for amendment rates. But apparently, every new variable
introduces a new methodological challenge.

Conclusion

Four indices of constitutional rigidity have been discussed in this article and
assessed along with two common statistical alternatives to reduce errors and
to estimate missing values (average of standardized values, imputed factor).
Although the different operationalizations and mechanisms for measuring
lead to different and partly contradictory results, it is evident that they tap
the same important dimension that clearly emerged in and was confirmed
by a factor analysis.
Lutz’s index tends to explain only those amendment data which he pre-
sented, i.e. his index is probably linked too much to the evidence of the US
federal states and the allocation of scores for other actions is too dependent
on his own estimation to guarantee universal explanatory power. Only the
index of Dag Anckar and Lauri Karvonen tends to confirm the hypothesis
about an inverse correlation between the difficulty and the contemporary
amendment rates. But neither does it reach the high correlation suggested
by Lutz (1994) nor is it free of doubts concerning its universal applicability,
bearing in mind that they treat the rigidity as the dependent variable. Accord-
ingly, in tests it tended to deviate from the index average and its explanatory
outcomes.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 355

The index of Arend Lijphart tends to result in the consistently weakest


relationship between rigidity and amendment rates. Although one may criti-
cize some aspects of his index, my own study of constitutions makes me think
that it reflects the actual difficulty of amendment procedures better than the
more detailed indices of Lutz and Anckar and Karvonen. My own index
seeks to reflect the empirical variety of amendment procedures by consider-
ing more than just majority requirements (e.g. elections between the votes)
and more voting arenas than just parliaments but it sticks to the relatively
straightforward Lijphart type of scaling scores.
All in all, the indices of Lutz, Lijphart and my own redrafting seem to be
more promising. However, they all result in weak correlations to contempor-
ary constitutional amendment rates. A model of explanation proposed by
Donald S. Lutz which relies on two independent variables, the rigidity and
the length of constitutions, could not be verified for a sample of fully estab-
lished democracies in the period of 1993–2002. To solve this puzzle should be
a challenge to the further debate on constitutional politics. Has the pro-
intuitive effect of institutional rigidity been overestimated? Or must we refer
to the all too common wisdom that any such comparative analysis which
necessarily must be based on only a few variables, surely falls short of
explaining everything with perfection?
Given the present state of research, it would be particularly interesting to
check (and redesign) Lutz’s amendments rates and to think about theoretical
reasons for possibly decreasing amendment rates as well as for the respec-
tively varying explanatory power of the constitutional rigidity. Such a task
will, of course, be impossible without assessing the impact of other institu-
tional along with contextual variables and thereby embarking upon new
challenges.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


356
Appendix
Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009

Table A1 gives a summary of the information presented in the article. Table A2 not only shows how the different indices
assess the constitutional rigidity in 39 consolidated democracies, but also lists the medium of the standardized indices such
as the imputed factor scores and the respective ranks and Table A3 presents the constitutional amendment rates in 22
democracies calculated by Lutz (1994) and the amendment rates for the period 1993–2002 calculated by the author.

JOURNAL OF THEORETICAL POLITICS 17(3)


Table A1. The Indices at a Glance

Donald S. Lutz Arend Lijphart Dag Anckar & Lauri Astrid Lorenz
Karvonen

What has to be explained Amendment rate of Performance of consensus Preferences of countries for Frequency, significance of
constitutions and majoritarian amendment provisions constitutional reforms
democracies

Index of rigidity Certain kinds of majority; Kind of majority vote; Kind of majority vote; Kind of majority vote; voting
Considered factors success rate; times of electoral system (partially); voting arenas (legislature, arenas and actors (not
voting; unicameral/ approval by referendum people) limited; times of voting
bi-cameral legislature; (partially); approval by implied)
initiative actors; special states’ legislatures
body or regular legislature; (partially)
need of elections between
two votes
LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY
Mechanism of measuring Addition of scores Classification Classification Addition of scores

Range of scores/classes 0.25-4.00 1.0-4.0 1-9 0.5-4.0

Number of measured 32 42 85 39
cases

Basis of choice ‘True constitutional Democracies Free democracies Free, peaceful democracies
systems’ with population >1mio.
Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009

Time period Different Same (1945–96) Same (1999) Same (1993–2002)

Applicability to other Medium (catalogue of High Medium–High High


cases scores limited)

357
358
Table A2. Constitutional Rigidity Indices Concerning 39 Consolidated Democracies
Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009

State Constitution Donald Lutz Arend Lijphart Dag Anckar, Astrid Lorenz Average of standar- Imputed factor
established in Lauri Karvonen dized indices

Value Rank Value Rank Value Rank Value Rank Value Rank Value Rank

Australia 1901 4.65 20 4.0 24 4 6.5 8.0 34.5 0.88 36 1.21 36


Austria 1920 0.80 .5 3.5 3.0 15 5 18.5 3.0 7.5 .5 5.5 6

JOURNAL OF THEORETICAL POLITICS 17(3)


0.47 0.61
Belgium 1831 2.85 12 3.0 15 5 18.5 9.5 39.5 0.64 33 0.85 34
Benin 1990 – – – – 5 18.5 4.0 17.5 0.18 15 a0.30a 16
Bolivia 1967 – – – – 5 18.5 9.0 37.5 0.96 37 a1.23a 37
Botswana 1966 1.30 5 2.0 .5 6.5 6 33.5 3.0 7.5 0.46 7 0.68 5
Bulgaria 1991 – – – – 5 18.5 3.0 7.5 0.41 .5 9.5 a0.60a 8
Canada 1867 – – 4.0 24 4 6.5 7.0 31.5 0.53 31 a0.85a 33
Chile 1981 3.05 14 – – 5 18.5 7.0 31.5 0.44 30 a0.55a 30
Costa Rica 1949 4.10 19 3.0 15 5 18.5 4.5 23.5 0.30 26 0.37 27
Czech Rep. 1993 – – – – 5 18.5 4.0 17.5 0.18 15 a0.30a 18
Denmark 1953 2.75 11 2.0 6.5 6 33.5 8.0 34.5 0.37 28 0.42 28
Finland 1919 2.30 9 3.0 15 18.5 4.0 17.5 0.08 18 0.12 21
2000 – – – – 5 4.0
France 1958 2.50 10 1.6 .5 3 4 6.5 4.0 17.5 0.64 4 0.78 4
Germany 1949 1.60 7 3.5 21 5 18.5 6.0 29.5 0.16 24 0.22 25
Greece 1975 1.80 8 2.0 .5 6.5 5 18.5 5.0 25.5 0.34 12 0.45 12
Hungary 1949 – – – – 5 18.5 3.0 7.5 0.41 .5 9.5 a0.60a 9
Ireland 1937 3.00 13 2.0 .5 6.5 4 6.5 4.0 17.5 0.43 8 0.52 10
LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY
Italy 1948 3.40 17 2.0 .5 6.5 5 18.5 4.0 17.5 0.16 17 0.24 20
Jamaica 1962 – – 3.0 15 5 18.5 5.0 25.5 0.12 23 a0.12a 24
Japan 1889 3.10 15 4.0 24 6 33.5 8.0 34.5 1.00 38 1.24 38
Korea/South 1948 – – – – 6 33.5 5.0 25.5 0.44 29 a0.23a 26
Lithuania 1992 – – – – 5 18.5 4.0 17.5 0.18 15 a0.30a 19
Mauritius 1968 – – 3.0 15 5 18.5 3.0 7.5 0.19 13 a0.31a 17
Mongolia 1992 – – – – 6 33.5 4.0 17.5 0.21 25 a0.07a 22
Namibia 1990 – – – – 6 33.5 6.0 29.5 0.67 35 a0.54a 29
New Zealand 1840 0.50 1 1.0 .5 1.5 2 2.5 1.0 1.5 1.91 2 2.29 1
Netherlands 1815 – – 3.0 15 5 18.5 8.5 36.5 0.65 34 a0.88a 35
Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009

Norway 1815 3.35 16 3.0 15 5 18.5 3.5 12.5 0.05 22 0.04 23


PNG 1975 0.77 2 3.0 15 5 18.5 4.0 17.5 0.36 11 0.46 11
Poland 1992 – – – – – 33.5 3.0 7.5 0.02 20 a0.38a 13
1997 – – – – 6 5.0
Portugal 1976 0.80 .5 3.5 3.0 15 5 18.5 3.0 7.5 0.47 .5 5.5 0.61 7
Slovenia 1991 – – – – 6 33.5 3.0 7.5 0.02 20 a0.38a 14
Spain 1978 3.60 18 3.0 15 6 33.5 6.0 29.5 0.58 32 0.67 32
Sweden 1975 1.40 6 2.0 .5 6.5 2 2.5 4.0 17.5 1.12 3 1.25 3
Switzerland 1848 4.75 21 4.0 24 – 4.5 4.0 25.5 0.36 27 0.59 31
2000 – – – – 3 7.0
UK – – – 1.0 .5 1.5 2 2.5 1.0 1.5 2.03 1 a2.24a 2
Uruguay 1967 – – – – 6 33.5 3.0 7.5 0.02 20 a0.38a 15
USA 1789 5.10 22 4.0 24 9 39.5 9.0 37.5 2.07 39 2.40 39

a
Predicted on the basis of a best-subset regression.

359
360 JOURNAL OF THEORETICAL POLITICS 17(3)

Table A3. Amendment Rate Lutz and Amendment Rate 1993–2002

State Amendment rate Lutz Amendment rate


(period tested) 1993–2002

Australia 0.09 (1901–92) 0.0


Austria 6.30 (1975–92) 2.1
Belgium 2.30 (1973–88) 1.6
Botswana 2.44 (1966–84) 0.2
Chile 0.64 (1925–70) 1.1
Costa Rica 1.26 (1949–82) 1.5
Denmark 0.17 (1953–92) 0.0
Finland (1919) 0.86 (1919–92) 2.1
France 0.19 (1968–92) 0.9
Germany 2.91 (1949–92) 1.3
Greece 1.32 (1975–92) 0.1
Ireland 0.55 (1937–92) 0.9
Italy 0.24 (1946–92) 0.7
Japan 0.00 (1945–92) 0.0
New Zealand 13.42 (1947–87) 0.1
Norway 1.14 (1814–1982) 0.1
Papua New Guinea 6.90 (1975–92) 0.9
Portugal 6.67 (1976–91) 0.2
Spain 0.18 (1968–92) 0.0
Sweden 4.72 (1974–92) 1.3
Switzerland (1848) 0.78 (1873–92) 1.9
USA 0.13 (1789–1992) 0.0

Note: Every constitutional amendment is counted 1. In 2000, new constitutions came into
force in Switzerland and in Finland. I calculated the amendment rates of their old
constitutions in order to guarantee the comparability of the rates. Like Lutz, I divided the
number of amendments relating to the document of interest by the number of years in which
they were in force, i.e. 7 years for Switzerland and Finland and 10 years for all other cases.

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009


LORENZ: HOW TO MEASURE CONSTITUTIONAL RIGIDITY 361

REFERENCES
Anckar, Dag and Lauri Karvonen (2002) ‘Constitutional Amendment Methods in the Democ-
racies of the World’, paper delivered at the XIIIth Nordic Political Science Congress, Aalborg,
Denmark, 15–17 August, URL: http://www.socsci.auc.dk/institut2/nopsa/arbejdsgruppe24/
anckarkarvonen.pdf.
Bogdanor, Vernon (1988) ‘Introduction’, in Vernon Bogdanor (ed.) Constitutions in Democratic
Government, pp. 1–13. Aldershot: Gower.
Colomer, Joseph M. (2001) ‘Disequilibrium Institutions and Pluralist Democracy’, Journal of
Theoretical Politics 13(3): 235–47.
Göhler, Gerhard (1997) ‘Der Zusammenhang von Institution, Macht und Repräsentation’, in
Gerhard Göhler et al. (ed.) Institutionen–Macht–Repräsentation, pp. 12–13.
Goldstein, Richard (1996) ‘Patterns of Missing Data’, Stata Technical Bulletin 32: 12–13.
Goodin, R., ed. (1996) The Theory of Institutional Design. Cambridge: Cambridge University
Press.
Grimm, Dieter (1994) Die Zukunft der Verfassung. Frankfurt am Main: Suhrkamp.
Lijphart, Arend (1999) Patterns of Democracy: Government Forms and Performance in Thirty-Six
Countries. New Haven, CT: Yale University Press.
Little, R.J.A. and D.B. Rubin (1987) Statistical Analysis with Missing Data. New York: John
Wiley & Sons.
Lorenz, Astrid (2004) ‘Stabile Verfassungen? konstitutionelle Reformen in Demokratien’, Zeits-
chrift für Parlamentsfragen 35(3): 448–68.
Lutz, Donald S. (1994) ‘Toward a Theory of Constitutional Amendment’, American Political
Science Review 88(2): 355–70.
Riker, William H. (1980) ‘Implications from the Disequilibrium of Majority Rule for the Study
of Institutions’, American Political Science Review 74: 432–46.
Tsebelis, George (2000) ‘Veto Players and Institutional Analysis’, Governance 13(4): 441–74.

ASTRID LORENZ is Assistant Professor at the Institute for Social Sciences,


Humboldt University of Berlin. Her current research project is on comparative
constitutional politics in consolidated democracies, her teaching and research
activities include comparative politics and political institutions as well as sys-
temic change and European integration. She also heads the ‘East European
Societies in Comparative Perspective’ section of the German Association for
Political Science. ADDRESS: Institute for Social Sciences, Humboldt Univer-
sity, Berlin [email: astrid.lorenz@sowi.hu-berlin.de]

Downloaded from http://jtp.sagepub.com at CAPES on August 11, 2009

You might also like