Constitutional Referendums in the Countries of the World

This study of the countries of the world aims at understanding why 98 countries resort in constitutional amendments to mandatory or facultative referendums whereas 95 other countries do not. Two different explanatory approaches are tried out. The one departs from rationality assumptions and regards political institutions as instruments for problem-solving; the other departs from diffusion assumptions and regards institutions as outcomes from cultural and historical contexts. The main findings are consistent with the belief that the use of the constitutional referendum is rationality-driven. Given that amendment thresholds are much less constraining in plurality election systems than in proportional systems, a central hypothesis is that plurality elections promote the installation in amendment of methods for popular examination and approval. This is indeed the case: plurality elections spell referendums and especially so in the context of democratic politics.


Introduction
In his influential study some dozen years ago of Patterns of Democracy, Arend Lijphart observed that democracies when amending their constitutions "use a bewildering array of devices to give their constitutions different degrees of rigidity " (1999: 218).However, Lijphart hastened to moderate the implications of his observation, as he noted that the great variety of constitutional provisions may in fact be reduced to four basic types: approval by ordinary majorities, approval by two-thirds majorities, approval by less than a two-thirds majority but more than an ordinary majority (for instance, an ordinary majority plus a referendum), and approval by more than a two-thirds majority, such as a two-thirds majority plus approval by state legislatures (Lijphart, 1999: 219).This study of the countries of the world simplifies matters further, as it focuses on one division only, that, namely, between amendments that require the use in one form or another of a referendum device, and amendments that do not require this device.Adding to a growing literature on the use and consequences of the referendum in different parts of the world (e.g.Altman, 2011;Qvortrup, 2002), this study of the constitutional referendum institution aims at understanding why some countries resort in amendments to referendums, whereas other countries do not.
By way of introduction, one specific point of departure needs to be clarified.In one by now classic contribution to the field of political regime studies Maurice Duverger talks about "similarity of rules, diversity of games" (1980: 167), and he thus separates constitutional texts from political practices which may or may not correspond to the textual prescriptions.Explicitly, this study is about texts rather than practices.The aim is not to understand why a certain nation has executed a certain amount of constitutional referendums which is higher or lower than in other nations -such a research would be on an equality with the task of explaining by reference to the age of constitutions and similar circumstances why amendments and attendant referendums are more frequent in some systems than in others (Lutz, 1994).Rather, the study is about the constitutional prescriptions per se, and the research questions concern variations in the extent to which nations incorporate in their constitutions amendment methods that require the use of referendums.Or, to phrase this differently, the research is about the existence of methods and not about their actual use.The fact that there were up to the year 2000 only two referendums on constitutional issues in Armenia (Grotz & Rodrigues-McKey, 2001: 329) as against three in Kazakhstan (von Gumppenberg, 2001: 419) and four in Azerbaijan (Grotz & Motika, 2001: 357) does not constitute differences from the point of view of this study between the three countries.As they all subscribe to the use of the constitutional referendum, they stand for equivalence, not disparity.
The data on which the study is based cover all independent states in the world in the year 2010.Information concerning the amendment procedures in these states has been gathered from the constitutional texts that were valid in the year 2010; it goes without saying that these texts are for separate countries from very different years.For instance, the constitution of Iceland is from the year 1944 and the constitution of India is from the year 1950; on the other hand, the present constitution of the Maldives is from the year 2008 and the constitution of Kenya stems from 2010.These discrepancies carry consequences in terms of method.Namely, the older a constitution, the greater the need to alter its stipulations, and this notion is, in principle at least, valid also for the very stipulations that concern the amendment method itself.In other words, constitutions that are of age may have during the course of time altered their amendment clauses while leaving the main bulk of the constitutional text intact.Systematic controls of older constitutions have therefore been implemented in this analysis to detect late and still valid amendments on amendments.
Concerning the constitutional texts, two parallel sources have come to use.First, the web-source Constitution Finder, a database of national and state constitutions and related documents, provided by the University of Richmond School of Law (http://confinder.richmond.edu/)has been used systematically.Although rich and detailed, the materials from this source, however, are not in all respects satisfying.Some documents are available in un-accessible languages only -for instance, the constitution of North Korea is given in the Korean language, and the constitution of Sao Tomé and Príncipe in the Portuguese language.Furthermore, the latest constitutional developments and changes are not always recorded in full, and in a few cases summaries and abstracts rather than full documentations are available.However, the second source, the Constitutions of the Countries of the World, a series of updated constitutional texts by Oceana Publications (Blaustein & Flanz, various years), makes up for these shortcomings, and is an excellent guide to the constitutions and constitution-like texts from all parts of the globe.Not only contain the editions complete constitutional texts; in several cases the editors also provide expert commentaries as well as historical notes and reviews and annotated bibliographies.
The study is divided into five sections.Following this introduction, a second section provides a brief overall-view of the use of referenda in constitutional amendment.This section, still preliminary and descriptive, focuses the question: What types of constitutional amendment referenda are there and in how many countries?The following two sections have explanatory ambitions, as they depart from the assumption that constitutions and political institutions do not operate in a vacuum and that the choice of institutions does not happen at random.Still, no single theory can account for all varieties of institutional change, and the task of selecting adequate independent variables for explaining institutional variance remains difficult as well as challenging.Confronting the question "Do institutions matter?",Jean Laponce and Bernard Saint-Jacques give the following answer: "Yes, institutions do matter, as leaders matter, as social forces matter, as the weight of tradition matters, as accident matter" (1997,233).Very much the same list may be quoted as a guideline for seeking causes rather than effects of institutions; here, rather than focusing on one specific frame of reference, two competing as well as over-lapping theoretical frameworks are applied.The one operates from a rationality perspective and is dealt with in section three; the other operates from a diffusion perspective and is dealt with in section four.A final section gives a summation of findings.

A Brief Survey
First a methodological pitfall must be cleared off, which follows from the fact that several countries apply parallel but different methods of amendment, the threshold being higher for certain items than for certain other items (Anckar & Karvonen, 2002: 12-13).For instance, the constitution of Papua New Guinea makes use of three parallel methods, and states that "Nothing prevents different majorities being prescribed in respect of different aspects or subject matters of a provision" (Article 17; Anckar & Karvonen, 2002: 12-13).In cases where the use of referendum is prescribed for the amendment of certain but not all items, this parallelism obviously creates classification difficulties -are such countries constitutional referendum cases or are they not?Following a suggestion by Lijphart (1999: 221), classifications are guided here by a simple but reasonable principle, which states that the most rigorous requirement counts, except when evident that the requirement is valid for some very specific article or purpose only.For example, in St Vincent and the Grenadines the stipulation is that bills to alter the constitution must be approved in a referendum when and if they concern the protection of fundamental rights and freedoms, establishment of Parliament, election of representatives, appointment of Senators, matters of finance and public service, and the like (Constitution, article 38; Anckar & Karvonen, 2002: 13).It is evident from this enumeration that the referendum device, although in use for a defined set of matters only, is common enough to direct classification.Some basic distributions are given in Table 1, which has the form of a typology and also reports the frequency in the materials of the various types.As evident from the Table, the use in amendment of the referendum device is mandatory in a good third of the cases.It would appear, therefore, that he use of the device is a not so frequent feature, exception rather than rule.However, this is not the case.Namely, the remaining two thirds of the cases do not all distance themselves once and for all from the use of the device.In fact, almost one fourth of the remaining cases prescribe the use of referendum, given that certain conditions are satisfied that pertain to majority thresholds, the political institutions involved, situational political factors, and the like.These cases, then, operate a facultative referendum device (Suksi, 1993: 28-29).The over-all count is therefore quite even and balanced: against 95 countries which do not resort to the referendum stand 98 countries which make use of or may make use of the device.This crude division into two groups is basic to most of the calculations that will follow later in this analysis -against a group of countries that maintain mandatory or facultative referendums stands a group of countries that do not employ the referendum device.
There are in the materials two rather different types of facultative referendums.One declares referendums to be the rule, but allows for exceptions when and if defined prerequisites are at hand -in other words, referendums may be dealt with in terms of exemption.This strategy is applied in a good handful of cases; some examples may illustrate the mechanisms at hand.In Peru amendments require acceptance by a parliamentary majority plus a majority of votes in a referendum; however, the referendum stage can be avoided when and if the amendment proposal is accepted by Congress with a two-thirds majority in two successive regular sessions.In Kyrgyzstan, to take another example, amendments are decided by parliamentary majorities and a majority of votes in subsequent referendums; however, a fair amount of amendment issues may be decided by Parliament "upon proposal by the President, Jogorku Keresh, or initiative of not less than 300.000 voters".And in Benin, to give a third example, the constitutional amendment requirement is for a three-fourths parliamentary majority plus a majority of votes in a following referendum.However, if the parliamentary majority reaches the four-fifths threshold, no referendum phase is required.As evident from these examples, a central condition for the disposal of a referendum stage is an unusually high level of parliamentary and legislative support for the issue in question.On the other hand, the conditions for introducing referendums may rather imply that the popular vote is an exceptional device, which must, however, be resorted to when certain conditions are met.Referendums, then, may be dealt with in terms of insertion.Less than 10 per cent of all cases belong in this category, and again, as evident from some examples, the detailed prescriptions abound with variation.Draft constitutional laws require in Albania a two-thirds parliamentary majority, but must be submitted to popular vote when and if this is required by a two-thirds parliamentary majority.Accepted draft laws must be submitted to referendum when required by one fifth of Parliament.In Estonia amendment proposals are submitted to referendum if so required by a three-fifths parliamentary majority, and in Mongolia, where the amendment threshold is a three-fourths parliamentary majority, a ratifying referendum stage becomes necessary when called for by a two-thirds parliamentary majority.The amendment requirement in Togo is for a four-fifths parliamentary majority, but the President may always refer an amendment proposal to popular vote.Furthermore, referendum becomes mandatory when and if the amendment proposal is accepted by a parliamentary majority the size of which is less than four-fifths but still reaches the two-thirds threshold.
The place of referendums in the political decision-making sequence may vary.True, referendums are as a rule organized in order to obtain the people's approval of a measure, this meaning that referendums are about issues which have already been deliberated upon and approved by legislative bodies (Suksi, 1993: 10).However, this is a rule with some exceptions.In Zambia, for instance, amendments require a parliamentary majority of two-thirds in second and third readings; however, in entrenched issues, the deliberations in Parliament must be preceded by referendum.In the Seychelles, likewise, certain defined matters must be accepted in referendum with a 60 percent majority of votes before parliamentary deliberation.A final note in this section is about instances when the possibility of amendment is altogether denied, referendums or not.For instance, in Angola alterations to the Constitution must respect, among other things, "the dignity of the human person; national independence, territorial integrity and unity, the republican nature of the government", and "the uniting nature of the State" (Constitution, Article 236); in Benin "no procedure for revision may be instituted or continued when it shall undermine the integrity of the territory"; furthermore, "the republican form of government and the secularity of the state" shall not be made the object of revision (Constitution, Article 156).In El Salvador may "under no circumstances" the articles of the Constitution which refer to "the form and system of government, the territory of the Republic, and the principle that a President cannot succeed himself" be altered (Constitution, article 248), and in Haiti may no amendment of the Constitution effect "the democratic and republican nature of the State" (Constitution, Article 284-4).In Iran several individualized items are "unalterable"(Constitution, Article 177).As evident from these examples, the ban on amendment as a rule concerns very specific and fundamental material limits only.In the following analyses and presentations these stipulations are disregarded.

Referendums as Problem-Solvers
Reference was made here earlier to Arend Lijpharts treatise on Patterns of Democracy, the aim of which was to juxtapose two types of democracy, namely majoritarian and consensual democracy.In his discussion in this context of constitutional amendment and particularly the case of Barbados, Lijphart made the observation that the two-thirds amendment threshold in that country in fact, due to the single plurality election system, becomes clearly less constraining than in proportional system countries.In parliaments elected by plurality, so the argument goes, large majorities often represent much smaller popular majorities; moreover, these large parliamentary majorities are often single-party majorities.It follows, then, that majorities that are large enough to force through constitutional change in fact represent rather narrow population segments or political interests.Lijphart notes that while two-thirds majorities are required for amending the constitution of Barbados, in three of seven elections since 1966 such large one-party majorities were in fact manufactured from between 50 and 60 percent of the popular votes (1999: 219-220).Indeed, as evident from empirical research, in a number of cases in the English speaking Caribbean ruling parties have possessed the capacity to change or replace the constitution unilaterally without opposition votes (Elkins & Ginsburg, 2011: 16).This shows how in plurality systems supermajorities may occur by accident or even by chance and without being deeply rooted in the society.
This observation is from the point of theory highly relevant for an understanding of the introduction in amendment procedures of a referendum device.Namely, if the electoral system carries in its wake situations in which one single party may decide alone on constitutional reform, the need for corrections and balancing counter-measures becomes obvious.While the introduction of super-majority amendment thresholds like a requirement for a three-fourths legislative majority is one useful counter-measure (Anckar, 2012 a), the referendum is in like manner a balancing device, as constitutional outcomes of one-party or near one-party politics are now submitted to a general and popular control.The guiding hypothesis in this study, then, is that countries that have installed in national lower House elections plurality election methods are more prone than other countries to resort to the constitutional referendum -in this analysis, like in several other electoral system classifications (e.g.Sartori, 1994: 3-5), the term "plurality elections" covers majority elections as well.In other words: whenever the electoral system carries in its wake the risk of unfounded constitutional alteration, rational choice dictates that the alteration measures are subjected to popular examination and approval.
The basic distinction in regards to the independent variable, then, is between plurality elections and other elections, and the basic distinction in regards to the dependent variable is between an acceptance and a non-acceptance of the constitutional referendum.To these main distinctions, however, must be added a third distinction in regards to an intervening variable; this distinction is between democratic and non-democratic countries.The distinction is necessary because the functions of constitutions and, in consequence, the motives for introducing amendments and amendment methods may be assumed to be somewhat different in non-democratic than in democratic contexts (Derbyshire & Derbyshire, 1999: 16).Above all, the very tension between majority will and minority protection that is embedded in the democratic way of structuring government and is the target for the choice of amendment methods is not at issue to the same extent in non-democratic entities, which may be supposed to maintain a weaker commitment to popular rule.Only in democratic settings disproportionalities of electoral systems may produce real power changes and new power positions; therefore, a distinction needs to be introduced in the analysis at hand between democratic and non-democratic states.The ensuing expectation is now that the link between plurality election and the maintenance of the constitutional referendum will be more visible in a set of democratic than in a set of non-democratic countries.
Besides electoral systems, fragmentation is tried out here as a rationality-based explanatory factor.The expectation is that countries which are divided into competing and perhaps even hostile ethnic, language and religious segments will display an inclination to maintain the constitutional referendum.Again, this is because these countries face the unpleasant eventuality that some of the segments may reach a power position which makes it possible to force a moderate constitutional amendment threshold.If particular animosities prevail between the segments, the dread of power alterations will increase, as will the inclination to establish and preserve rigid amendment.The presidential election method in the Federated States of Micronesia and the motives for adopting this particular method serves as a good illustration of the mental dispositions, doubts and caution that are at play here.Although the political system of Micronesia is presidential in nature, the President is not popularly elected as in other presidential democracies, but is elected by Congress among the members that represent the state level.This deviation from a common pattern was introduced to lessen the possibility that a President will be elected solely because the single largest state has overwhelming electoral power (Burdick, 1988: 266-267).To decide empirically the level of fragmentation in the various countries, use is made here of an available listing, which reports for every country in the world three indices of fragmentation (Anckar, Eriksson & Leskinen, 2002).These indices are about ethnic, linguistic and religious fragmentation, and the list also combines these measures into an index of total fragmentation, which adds to the value for religious fragmentation the dimension of ethnicity or language, which ever returns the higher value (Anckar, Eriksson & Leskinen, 2002: 6).For the purposes of this study, the total fragmentation index comes to use; since the separate indices run on a scale from 0 to 1, it follows that the maximum total fragmentation value is 2.00.Findings are reported in Table 2. Instead, however, of dealing with separate explanatory factors, the analysis aims by grasping configurations of independent variables at looking for more complex relationships among the proposed causes.This is done in the form of a truth table, which is a basic tool of the Boolean algebra approach and presents all possible combinations of the values of the independent variables (Ragin, 1987;Peters, 1998: 162-171).This Boolean analysis, which requires that variables are made into dichotomies, classifies the available cases in terms of presence (Yes) or absence (No) of presumed determinants as well as presence or absence of the expected outcome.Concerning the independent variables, plural electoral system countries are compared to countries with other electoral systems; following a listing by Krister Lundell (2005: 45), single-member plurality, block vote and limited vote as well as one-ballot majority, two-ballot majority and alternative vote are classified here in the plurality-majority systems compartment.Furthermore, democracies are compared to non-democracies and the democratic status of the countries of the world is defined in terms of Freedom House classifications for the year 2010.Countries that are classified by Freedom House as "Free" are regarded here to be democracies whereas other countries are non-democracies (Anckar, 2011: 54-57).Finally, fragmented countries are compared to less fragmented countries, the cutting point being the value 1.00 on the above fragmentation scale.While the total number of countries in the present analysis is 193, due to missing data Table 2 deals with 181 cases only.Fragmentation data are not available for a small handful of countries, like Timor-Leste and Montenegro, and a good half-dozen of countries, like Brunei, Oman, Qatar, Saudi-Arabia, and Vatican City, do not have national elections and, in consequence, electoral systems.
An analysis of the impact of electoral systems makes a strong case for the institutions-as-problem-solvers approach.More than three fifths of the plurality method countries are referendum countries as against clearly less than half of the countries with other electoral systems -whereas, to give a few examples, Belarus, France, Grenada, and Mauritania are referendum cases, Bolivia, Lithuania, Macedonia and Thailand are not.In other words: when a specific electoral arrangement applies, referendum use is promoted, and when the electoral arrangement does not apply, referendum use diminishes.Much in line with expectations, the insert of a democracy variable adds to the explanatory power of the electoral system categorization.Of democracies with plurality elections seven cases out of ten are in a referendum category; of non-democracies with plurality elections a significantly smaller part, although still a majority, are in the same compartment.The predominance of democracies notwithstanding, the fact that many non-democracies behave in the same manner certainly serves to support the statement that referendums are frequently deployed in the settlement of constitutional questions even in countries with little or no democracy tradition (Tierney, 2012).Furthermore, again in line with expectations, the impact of the democracy categorization decreases when other than plurality electoral systems are considered; in fact, the amount of referendum countries is now clearly smaller than the amount of non-referendum countries.In the absence of the trigger factor (plurality elections), the consequence (constitutional referendum) likewise fails to come off.In total, then, much can be said in favor of the belief that the use of the constitutional referendum is rationality-driven.In the wake of plurality elections political constellations may follow which pave the way for hasty and narrowly-supported alteration; therefore, countries with plurality elections may be expected to ward off this threat by means of constitutional referendums.And indeed, this expectation is verified.Plurality elections spell referendums and especially so in the context of democratic politics.
Turning to fragmentation, a cursory examination of the materials offers little guidance as data appear contradictory.
A supportive finding is that there are indeed in the materials countries that are fragmented and exercise the referendum -among such countries are, to mention a few, Benin, Ivory Coast, Uganda, Vanuatu and Zambia.However, a contradictory and non-supportive finding is that there are countries that are fragmented but still do not resort to the constitutional referendum -among such countries are, again to mention a few, Belize, Bosnia-Herzegovina, Solomon Islands and Tanzania.And, on the other hand, some low fragmentation countries like Armenia, Italy and Marshall Islands make use of the constitutional referendum, whereas other low fragmentation countries like Finland, Norway and San Marino do not.The more systematic and over-all view that may be derived from Table 2 suggests, first, that there is a link, albeit perhaps a weak one, between fragmentation and the use of a constitutional referendum, and, second, that fragmentation does not appear to add to the impact that follows from plural elections alone.In other words, whenever plurality systems are in the picture, referendums often follow, and it does not matter much whether or not fragmentation is also in the picture.However, importantly: in other constellations heterogeneity certainly appears to make a difference.Non-plurality systems, when combined with heterogeneity, tend to promote to some extent the installment of referendums; on the other hand, when and if the heterogeneity factor is absent, non-plurality systems as a rule combine with non-referendums.Therefore, fragmentation appears to have a triggering function, which, however, surfaces only in the absence of the other triggering factor, namely plurality elections.And, also importantly: whereas fragmented democracies do not markedly favor the referendum device, the link between fragmentation and the constitutional referendum is much stronger when non-democracies are concerned.This rather counter-intuitive finding follows most probably from the position of several African regimes, to be dealt with at some length shortly.

Diffusion as an Explanatory Factor
According to a second approach to constitutional politics, rather than representing outcomes of rational choice, constitutional features reflect the cultural and historical contexts of which they are part.Here, the validity of this second approach is tested by means of two indicators.First, the colonial heritage of countries has been found to possess considerable explanatory power vis-á-vis variation in political institutions; in particular, countries belonging to the British Commonwealth of nations often display institutional similarities that set them apart from most other countries (e.g.Pinkney, 1993: 40-60).Since the doctrine of parliamentary sovereignty is usually conceived of as a center-piece of the Westminster Model (e.g.Lijphart, 1984: 9), and since this doctrine prescribes amendment by regular parliamentary majority only (Lijphart, 1984: 9;Strong, 1958: 65), it is a reasonable and diffusion-derived expectation that countries that are freed and independent from British rule have adopted the metropolitan mode of constitutional amendment and refrain from the use of amendment via referendum.Furthermore, the expectation is that there will be a marked difference in this respect between a group of former British colonies and a group of other states.These expectations will be tried out here, as a group of 54 countries that are freed since World War II from British control (Derbyshire & Derbyshire, 1999: 811-813;Anckar, 2011: 51-52) are compared to the other countries of the world.From this comparison, however, two former colonies are left out.Somalia does not presently have a recognized government and South Yemen does not exist anymore as an independent state.
Second, as it is probably more natural to imitate one's neighbors than distant states, patterns of imitation and diffusion often follow geographical boundaries.In consequence, countries that belong to particular world regions may be expected to display similarities with one another.For instance, whereas Latin American countries have presidential executives and representational legislatures, the nations of Western Europe and Scandinavia typically have parliamentary systems with proportional representation (Powell, 1992: 231).A reasonable expectation in the present context, therefore, is that there are distinct differences between regions in the use of the constitutional referendum -while some are imbued by adherence to the referendum, others dissociate themselves from the use of the device.The empirical testing of this assumption requires that individual countries are placed in spatial and politico-geographical frames, and while this is in most cases a rather straight-forward enterprise, border-cases are nuisances.The same is true to some extent of the frames themselves.These difficulties are avoided here simply by means of imitation.Namely, the classifications that come to use follow closely the geography classifications in the authoritative series of electoral data handbooks of the world that were published in the time span 1999-2010 by Dieter Nohlen and his research team (Nohlen, Krennerich & Thibaut, 1999;Nohlen, Grotz & Hartmann, 2001;Nohlen, 2005;Nohlen & Stöver, 2010).The application of the Nohlen framework results in a division of the countries of the world in six groups, four of which have from 31 to 52 country members (Africa, Americas, Asia, Europe), while the two remaining categories are smaller (Middle East, South Pacific).
The findings in regards to these two diffusion approaches are given in Table 3, where the relevant frequencies are given as percentages.As evident from the distributions, the findings are internally somewhat inconsistent and perhaps not very encouraging.The colonial heritage factor in fact appears almost irrelevant.True, in line with expectation, more than former British colonies, other states tend to resort to the use of constitutional referendums, the difference being less than obvious and more than marginal.However, importantly, former British colonies do not stand out as a homogeneous group.Indeed, while half of the British cases resort to referendum amendment, the other half does not.A rejection of the amendment via referendum device is therefore by no means a distinguishing feature of states that have formerly been under British control.Most probably, this deviation from a mechanical diffusion pattern resembles a rational choice outcome, and this test of a diffusion impact therefore serves to indirectly support the explanation that has been offered in this study as a plausible alternative to diffusion.Namely, more than in the metropolitan power, political life has in most former British colonies been marked by ethnic, social and rigidity-promoting heterogeneity and more imbued by challenges to the legitimacy of government.The colonies have therefore been less inclined to endorse a principle of unconditional parliamentary sovereignty (Anckar, 2012 b: 14).Indeed, more than 2/3 (11 out of 16) of those former colonies that display unusually high (> 1. 3) fragmentation values are in the constitutional referendum camp.While the findings that pertain to the geography dimension are in like manner somewhat inconsistent, they still offer more support to a diffusion hypothesis.On the one hand, there are equal or about equal proportions of countries favoring or rejecting referendums in the Americas, Europe, and South Pacific; these regions, then, do not bear witness to any regionally defined clustering of attitudes towards the amendment via referendum device.On the other hand, however, the situation is different in regards to Africa, Asia and Middle East.Close to two thirds of the Asian countries and four fifths of the Middle East countries avoid the constitutional referendum; in contrast, implementing in the 1960s several referendums to support regime changes and later in the 1990s several referendums to support transitions to multi-party systems and to finalize conflict resolutions (Kersting, 2009), more than two thirds of the African countries maintain the device.Africa therefore represents a pro-referendum region while Asia and Middle East represent quite the opposite.Given that these regions all have a less than satisfying democratic position, and given also the general link, as demonstrated above, between democracy and referendum, it is interesting to note the much different outlooks of the regions in regards to the referendum institution.The political uprisings that recently swept across the Arab world notwithstanding, Middle East still stands out as a region that seems immune to democratic change -in many Arab countries, "democracy movements have yet to reach even the initial milestone of forcing the resignation of their longtime rulers", it is said in the most recent Freedom House Full Report Essay on Freedom in the World (Freedom House.Freedom in the World 2012).Africa has been characterized as a continent "that is usually excoriated for an unsatisfactory postcolonial democratic record" (Dale, 1999: 128), and the democratic record in Asia is in like manner defective.Still, as noted, Africa cultivates the constitutional referendum, whereas Asia and Middle East do not.
Why is this so?How is the frequent occurrence of constitutional referendums over the African continent best explained?One possible answer may be derived from the fact that most African countries with a constitutional form of government that is based on elections have a presidential or semi-presidential system of government (Nohlen, Krennerich & Thibaut, 1999: 25-30).As is well known, the existence of two independent organs in such systems has a built-in tendency to promote executive-legislative deadlock and to promote also the origin of solutions that unbalance the powers by making the president the spearhead of government (e.g.Lijphart, 1992: 15-16;Linz, 1994).In Africa, in accordance with a tradition that "African rulers, Prime Ministers and Presidents, soon see themselves as taking the place of Kings, and that is the way the electorate of the ruling Party generally sees the situation too" (Machobane, 2011: 9), authoritarian leaders who wish to expand their terms of office and their constitutional powers have sought popular approval, often enough by resorting to rigged practices.Identifications of common patterns of Third World referendums have suggested the frequent occurrence of overwhelming majority outcomes; not seldom referendums have even resulted in a yes-vote of 99,9 % (Marques & Smith, 1984).Verifying the recent observation by David Altman (2011: 88) that under undemocratic systems, top-down referendum proposals are to be expected, in an abundance of cases, outcomes of African constitutional referendums have followed this pattern, adding turnout-reducing boycotts from the part of the political opposition.
For instance, when in 2009 a constitutional referendum was called by the President of the island state of the Comoros, the initiative was opposed by opposition parties and those from other islands who saw the referendum as a ploy to extend the presidential term of office at the expense of the other islands.The referendum was approved by 94 % in favor at a turnout of 52 %. (Comorian constitutional referendum, 2009).Further scattered examples are a constitutional referendum in Benin in 1990 approved by 93% at a turnout of 64% (Hartmann, 1999: 89) , a constitutional referendum in Burkina Faso in 1991 approved by 93 % at a turnout of 49 % (Grotz, 1999: 133), and a constitutional referendum in Gabon in 1995 approved by 97% at a turnout of 64% (Fleischhacker, 1999: 398).Of course, in such settings the essence of referendums, the transformation of citizens into legislators is wasted (Hague & Harrop, 2004: 162).

Closing
This study has been about a potential clash in constitutional amendment procedures between political and legislative sovereignty.As indicated by reviews of the relevant literature (e.g.Suksi, 1993: 15-24), categorizations of sovereignty types tend to overlap and interlock in a rather confusing manner.In the present context, however, it suffices to say that the dividing line is between a political sovereignty located in the people and a legislative sovereignty located in a body distinct from the people which still may or may not be representative of the peoplethis body may be a legislature or a Constitutional Assembly, or an individual ruler like an Amir (e.g.Bahrain, Kuwait), a King (e.g.Jordan, Swaziland), a Sultan (e.g.Brunei, Oman), or the Supreme Pontiff (Vatican City).The point of departure of the study was that legislative sovereignty remains unchallenged in some societies, whereas in other societies legislative sovereignty is by means of referendums subordinated to and controlled by political sovereignty.To find and apply suitable standards for an evaluation of this divide in terms of constitutional theory appears a difficult and complicated task (e.g.Kahan, 1999;Tierney, 2009); however, it is also a task that has not been undertaken here.The main research task has been to map, understand and explain the shape of the divide.
Two different and seemingly contradictory approaches have been tried out.The one departs from rationality assumptions and regards political institutions as instruments of problem-solving: from this perspective amendment methods are chosen because they lend themselves to managing problems of low-threshold constitutional alteration and fragmentation.The second approach departs from diffusion assumptions and regards institutions as outcomes from cultural and historical contexts: from this perspective amendment methods are chosen because they correspond to familiar, induced or perhaps even forced models.The findings are in the main consistent with the belief that the use of the constitutional referendum is rationality-driven; one important individual finding is that plurality elections tend to promote the use of the constitutional referendum.A generalization of this finding is that institutional design may really be explained by institutional design, this meaning that constitutional choices are dependent on each other, so that the choice of one device (referendum) follows from the choice of another device (plural electoral system), like when two chambers are usually regarded necessary in federal systems.From this generalization follow two final notes on the interplay between frames of reference that build on rationality and diffusion: First, in the social sciences the demarcation lines between approaches and findings often remain to some extent suggestive rather than definite.For instance, research on small state democracy has suggested that small size does not appear to associate systematically with culture to the exclusion of rationality or with rationality to the exclusion of culture (Anckar, 2008: 82-83); this observation on the complementary and parallel nature of the approaches is certainly valid in the context of this study also.For instance, although final estimations must await the outcome of series of case studies which cannot be pursued here, it is still tempting to interpret as an outcome of parallel impact the fact that half of the former British colonies have opted for the constitutional referendum whereas the other half has not.Second, besides being complementary, the two frames appear in the present context in a causation pattern, the one explaining the other.Namely, of the 82 countries with a plurality election method that are included in the calculations in Table 2, clearly more than half, namely 45, are former British colonies.This means, first, that the colonies contribute extensively to the total quantity of plurality systems, and, second, that a vast majority of the colonies have adopted the electoral system of the metropolitan power or variants of this system.Diffusion, in other words, emerges as a powerful explanatory tool in regards to the choice of one institutional arrangement.This arrangement, however, may have in its wake inferior consequences in terms of constitutional alteration, and therefore calls for moderating efforts, the implication of which is that legislative sovereignty is controlled by popular sovereignty.Diffusion breeds rationality, so to speak.

Table 1 .
Referendums in constitutional amendment: An empirical typology

Table 2 .
Explaining the use of constitutional referendums: A truth table

Table 3 .
Colonial heritage and geographical location as determinants of constitutional referendum occurrence