1 The Politics of Semi‐Presidentialism

This chapter provides an introduction to the concept of semi‐presidentialism and establishes a framework for the study of the politics of semi‐presidential regimes. In the first part of the chapter, the evolution of the concept of semi‐presidentialism will be sketched, some of the main criticisms of the concept will be considered, a slight reformulation of the standard definition of the term will be proposed, and a list of semi‐presidential regimes will be identified. In the second part of the chapter, a framework for the comparative study of semi‐presidential regimes will be outlined. In this way, then, this chapter provides the basis both for the in‐depth country studies of semi‐presidentialism which follow and for the conclusion which examines the comparative experience of semi‐presidentialism and addresses the issue of whether or not countries should adopt a semi‐presidential form of government in preference to either presidential or parliamentary forms of government.

In a popular context the term ‘semi‐presidential regime’ was first used by the journalist and founder of the Le Monde newspaper, Hubert Beuve‐Méry, in 1959 (reprinted as Beuve‐Méry 1987).¹ At this time, though, the meaning of the term still remained rather vague and undefined. In an academic context the concept of semi‐presidentialism was first elaborated by the French political scientist, Maurice Duverger. Duverger first employed the term in the 11th edition of his textbook on political institutions and constitutional law which appeared in 1970 (1970: 277). He treated the subject in slightly more detail in 1974 (Duverger 1974) and his first full‐scale work on this theme appeared in 1978 (Duverger 1978). In France, then, the term was in regular use and was the subject of fierce debate by the end of the 1970s.

Elsewhere, interest in the concept took somewhat more time to develop. In 1980, the first article on the subject appeared in English written by Duverger himself (Duverger 1980). In 1983, an international conference was held on the theme of semi‐presidentialism which included contributions from, amongst others, Portuguese and Finnish political scientists (published as Duverger 1986b). In 1984, the first major study based on Duverger's work appeared in Spanish by a South American academic (Nogueira Alcalá 1986). In the early 1990s there was a growing German interest in the concept as the process of democratization gathered pace in Eastern Europe and the former USSR. (See e.g. Bahro and Veser 1995; and Steffani 1995.) Overall, by the end of the 1990s, reference to the term has become widespread and politics textbooks increasingly include a section on semi‐presidential regimes. Indeed, in 1997, Duverger's 1980 article was nominated and chosen as one of the most influential to have been published in the first 25 years of the history of the European Journal of Political Research.

The concept of semi‐presidentialism, then, has well and truly come of age. Since its first formulation, though, it has evolved. Moreover, it has consistently been the subject of criticism. Indeed, both the confusion that has surrounded and continues to surround the concept and the criticisms that have been directed at it suggest that a reformulation of the term is required before the study of the politics of semi‐presidential regimes can be undertaken.

The concept of semi‐presidentialism has been the source of a certain confusion over the years. In particular, there is confusion surrounding both the definition of semi‐presidentialism and the list of countries which should be classed as semi‐presidential regimes. In part, this confusion is caused by the development of the concept in Duverger's own work. In part, it is caused by how the concept has been applied in the work of others.

In 1970 Duverger provided the first definition of semi‐presidentialism. He stated that a semi‐presidential regime was ‘characterized by the fact that the head of state is directly elected by universal suffrage and that he possesses certain powers which exceed those of a head of state in a normal parliamentary regime. However, the government still consists of a cabinet formed by a prime minister and ministers who can be dismissed by a parliamentary vote’ (1970: 277).² At this time, according to Duverger, the list of semi‐presidential regimes comprised three Western democracies, Austria, Finland, and France, to which a fourth, Ireland, was added in the 12th edition of his textbook in 1971 (1971: 279). In 1974, though, Duverger altered the definition of semi‐presidentialism somewhat, now stating that a semi‐presidential regime exhibited three characteristics: ‘(1) the president is elected by universal suffrage. . . (2) opposite him, there is a prime minister and ministers who can only govern with the confidence of parliament. . . (3) the president can dissolve parliament. . . ’ (1974: 122). At the same time Duverger also revised the list of semi‐presidential regimes by both including Iceland and casting some doubt as to whether or not Ireland should in fact be classified as an example of such a regime (1974: 124). It is apparent, therefore, that up to this point at least there was a certain degree of confusion in Duverger's own mind concerning the concept of semi‐presidentialism.

In fact, Duverger only arrived at both his final definition of semi‐presidentialism and his stock list of semi‐presidential regimes in 1978 (1978: 17). It was this 1978 definition which was effectively restated in his 1980 article and which, subsequently, has become the standard English‐language definition of semi‐presidentialism. This definition is as follows:

[a] political regime is considered as semi‐presidential if the constitution which established it combines three elements: (1) the president of the republic is elected by universal suffrage; (2) he possesses quite considerable powers; (3) he has opposite him, however, a prime minister and ministers who possess executive and governmental power and can stay in office only if the parliament does not show its opposition to them. (Duverger 1980: 166)

Also according to Duverger, six countries should be classed as semi‐presidential: Austria, Finland, France, Iceland, Ireland, and the then recently established Portuguese regime. Subsequently, Duverger has consistently maintained this definition³ as well as the 1980 list of semi‐presidential regimes, although constitutional developments particularly in Central and Eastern Europe and the former USSR have led him to acknowledge that certain countries, such as Poland and Romania, should now be added to the list (Duverger 1992: 901).

Since 1978, therefore, Duverger has been consistent in both his definition of semi‐presidentialism and his classification of semi‐presidential states. Since this time, however, other writers have adopted different definitions of semi‐presidentialism and have identified different examples of semi‐presidential regimes. For example, O'Neill uses the term ‘semi‐presidential’ ‘to refer to those executive systems where (1) executive power is divided between a prime minister as head of government and a president as head of state, and where (2) substantial executive power resides with the presidency’ (1993: 197).⁴ This means that, for O'Neill, countries with directly elected but weak presidents, such as Austria, Iceland, and Ireland, should not be classed as semi‐presidential, whereas countries with indirectly elected but strong presidents, such as Albania and (formerly) Czechoslovakia, should be classed as such. In a similar vein, Sartori states that a political system is semi‐presidential if five properties jointly apply (1997: 130–1). These include the conditions that the president must be popularly elected, that the prime minister must be parliament‐dependent, and that the president must share executive power with the prime minister. This means that, for Sartori at least, the list of semi‐presidential regimes consists only of Finland, France, and, arguably, Sri Lanka. Furthermore, Linz argues, more succinctly, that semi‐presidential systems are those which ‘have a president who is elected by the people either directly or indirectly, rather than nominated by parliament, and a prime minister who needs the confidence of parliament’ (1994: 48). For Linz, this means that Finland, France, and Portugal are the primary examples of semi‐presidential countries.

It is apparent, then, that there has been and there continues to be a degree of confusion concerning the concept of semi‐presidentialism. Different people mean different things by the term and different people classify different countries as examples of semi‐presidential regimes. Needless to say, this causes problems both for the student of semi‐presidentialism and, it may be argued, for the very appropriateness of the concept itself.

Even though reference to semi‐presidentialism (however defined) has become widespread, there has always been and, indeed, there continues to be a certain amount of opposition to the concept. For example, in 1979 a leading French academic, Georges Vedel, stated that ‘at best, a semi‐presidential regime is only a convenient name given to a succession of contrary political practices closely linked to political changes’ (Le Monde, 19–20 Feb. 1979). Similarly, Shugart and Carey found Duverger's use of the concept to be ‘misleading’ (1992: 230) and instead preferred to formulate the concept of ‘premier‐presidentialism’. It goes without saying that Duverger himself is very aware of these criticisms. Indeed, as we shall see, he is quite dismissive of some of them, particularly those which emanate from his French colleagues. At the same time, however, it is necessary to examine the objections to the concept of semi‐presidentialism in order better to understand the meaning and implications of the term. In this context, four common criticisms of semi‐presidentialism can be identified. The first two, it will be argued, are misdirected, whereas the final two raise issues which need to be addressed and which necessitate a slight reformulation of Duverger's definition.

One frequent criticism of semi‐presidentialism concerns the term itself. Some writers simply object to Duverger's terminology. So, for example, Duhamel states that Duverger's use of the word is ‘disputable’ (1993: 158). In this context, there are two forms of this criticism. Some writers suggest that the term is satisfactory but that other terms are equally satisfactory, whereas others suggest that the term is unsatisfactory and should be substituted for a different term. Both objections, it might be argued, miss the mark.

The work of Linz and Stepan and Suleiman provides examples of the first terminological criticism. Both sets of writers accept the term ‘semi‐presidential’ but argue that it is synonymous with the term ‘semi‐parliamentary’ (Linz 1994: 48; Linz 1997; Stepan and Suleiman 1995: 394). For writers such as these, the term ‘semi‐presidential’ is, thus, potentially misleading because it can be substituted by another term which is equally valid. In response, it might be argued that this criticism is insignificant. After all, what does it matter which term is used if the subsequent methodology is valid? Duverger argues, however, that it does matter and that the term ‘semi‐presidential’ is the most appropriate one to use. In opposition to writers such as Linz and Stepan and Suleiman, he states that there is a significant distinction between the terms ‘semi‐presidential’ and ‘semi‐parliamentary’. For him, this distinction is to be found in the essential difference between a presidential regime and a parliamentary regime. In the former there are two sources of popular legitimacy (presidential elections and legislative elections), whereas in the latter there is only one (legislative elections). To the extent that in semi‐presidential regimes there are also two sources of popular legitimacy (presidential elections and legislative elections), then it is quite appropriate to call such regimes ‘semi‐presidential’ (Duverger 1986b: 8).⁵ For Duverger at least, then, it is clear that the terms ‘semi‐presidential’ and ‘semi‐parliamentary’ are not simply synonymous, that the term ‘semi‐presidentialism’ is used to mean something quite specific and that it is the most appropriate term to use to describe the countries that Duverger wishes to examine.

The work of Shugart and Carey provides an example of the second objection to Duverger's terminology. They have argued that the term is misleading because the use of the prefix ‘semi’ implies that semi‐presidential regimes are ‘located midway along some continuum running from presidential to parliamentary’ (Shugart and Carey 1992: 23). Consequently, they prefer to use the term ‘premier‐presidentialism’ where no halfway house situation is implied. To a certain extent Duverger is guilty of bringing this criticism on himself. In an early work he stated that the Finnish system is a ‘truly intermediate regime between a presidential and a parliamentary regime’ (Duverger 1974: 131). However, this is not what Duverger now argues and, in any case, the logic behind Shugart and Carey's argument is muddled. Sartori, for one, is quite dismissive of their approach. He believes that the term ‘semi‐presidential’ does not at all imply that such regimes are situated halfway along a presidential/parliamentary continuum. He points out that the prefix ‘semi is the Latin for “half”, and—as any dictionary would show for hundreds of expressions—does not assume any continuum because it proceeds continuum‐mania by well over two thousand years’ (Sartori 1997: 137). Lijphart, too, springs to Duverger's defence, although his language is rather more understated. He notes that ‘Duverger's concept of semi‐presidentialism is multi‐faceted and does not entail any intermediate distance between presidentialism and parliamentarism’ (Lijphart 1997: 126). In this sense, then, Shugart and Carey's line of thought appears fundamentally flawed.

Overall, then, it does appear as if Duverger has reasonable grounds to argue that the term ‘semi‐presidentialism’ means something quite distinct and that this term should be used in preference to alternative terms when examining the politics of particular countries.

Another common criticism of the concept of semi‐presidentialism is that a semi‐presidential regime is a mixed type of regime and, as such, is somehow intellectually out of place. Once again, it might be argued that this criticism is unfounded.

Some writers argue that semi‐presidentialism implies a mixed regime type. As such, it does not constitute a ‘pure’ regime type, like presidentialism or parliamentarism, and so it does not have the same conceptual validity as these other more standard types of regimes. So, for example, Pactet argues that ‘mixed regimes combine elements borrowed from presidential and parliamentary regimes, which with regard to the way that they function sometimes raises the problem of their coherence’ (1995: 153). Equally, Conac quotes de Tocqueville's objection to the very concept of mixed regimes and argues that semi‐presidential regimes function either as presidential regimes or parliamentary regimes (1992: 817). Similarly, Vedel in a much‐quoted article argues that the supposedly semi‐presidential French Fifth Republic is not a synthesis of presidential and parliamentary systems (and in this sense a stand‐alone regime type), but that it alternates between the two (Le Monde, 19–20 Feb. 1979: 2).⁶ Indeed, the notion that semi‐presidential states alternate between presidential and parliamentary phases is a further component of Shugart and Carey's objection to the term (1992: 23). In these ways, then, all of these writers object to the concept of semi‐presidentialism because it is considered to be an impure, hybrid, or ‘bastard’ concept (Bahro and Veser 1995).

Again, Duverger is at least partly responsible for provoking this criticism. For example, in his 1980 article he stated that semi‐presidential systems were ‘intermediary between presidential and parliamentary systems’ (Duverger 1980: 165). Even recently he stated that a semi‐presidential regime is ‘part presidential, part parliamentary’ (Duverger 1991: 109). However, whether or not Duverger is culpable, it is certainly the case that he pulls no punches in his opposition to those who voice this criticism. For example, in one article he stated: ‘the term semi‐presidential regime is still boycotted by French jurists who continue to venerate only two sacred cows: the parliamentary regime and the presidential regime’ (Duverger 1986c: 347). Similarly, in another he writes:

widely adopted in Portugal, accepted in Finland, used in Anglo‐Saxon countries, the notion of a semi‐presidential regime is still controversial in France . . . the majority of French constitutionalists (apart from the most serious of them) still maintain a fetishistic cult towards this dualistic vision [of parliamentary and presidential regimes] and consider anything which might complement it with a new model to be sacrilegious. . . (Duverger 1992: 901–2)

In fact, as Duverger implies, the mixed regime criticism is misdirected. There is no reason why a semi‐presidential regime should be considered a mixed regime at all. Instead, as Pasquino states, semi‐presidential regimes constitute a ‘specific and separate’ form of government (1995: 57). They possess their own ‘appropriately devised institutional features’ (Pasquino 1997: 129) and ‘what is required for the construction of semi‐presidential systems is an explicit, purposive and well designed act of institutional and constitutional engineering’ (ibid.). So, while it is certainly the case that Finnish constitution‐builders did not state that they were creating a semi‐presidential regime in 1919, that the Fifth Republic's founders were ignorant of the term in 1962, that the Bulgarian constitution states that the regime is parliamentary, and so on, it is also the case that the political institutions in these countries and others were arrived at purposefully and that collectively they constitute a specific and separate regime type. It follows, then, that ‘Presidential systems cannot simply, so to speak, lapse into semi‐presidential systems nor can parliamentary systems jump into semi‐presidential systems’ (Pasquino 1997: 129). As such, semi‐presidential countries do not alternate between presidential and parliamentary regimes. On the contrary, these countries simply exhibit various forms of political practice within the same basic constitutional structure and, in this sense, within the same regime type. In this way, semi‐presidential regimes are just as ‘pure’ as presidential or parliamentary regimes which also exhibit equally varying forms of political practice at different times (see below).

In contrast to the two previous objections, a relatively minor but nevertheless cogent criticism of semi‐presidentialism stems from the wording of Duverger's standard definition.

The first element of Duverger's definition states that ‘the president of the republic is elected by universal suffrage’. For some, this wording is problematic because it implies that the president is directly elected. And yet, certain countries which Duverger classifies as semi‐presidential appear not to meet this criterion. Most notably, this was said to be the case for Finland where the president was chosen by an electoral college prior to the reform of the country's electoral system in 1988. For example, Stepan and Skach argue that ‘from 1925 to 1988 the Finnish president was not so much directly elected but indirectly chosen by party blocs’ (1993: 5). Similarly, Shugart and Carey state that ‘Given its party‐centred character, [the election of the president] was not much different from election in parliament’ (1992: 213). It should be added that a similar criticism might be levelled against the Irish case. Here, there is a long tradition of parties conspiring to nominate an agreed presidential candidate (see Chapter 6). In this situation, the election is dispensed with altogether and the candidate is elected unopposed. Consequently, some Irish presidents have assumed power without having been directly elected at all which, again, might be seen to cause a problem for Duverger's analysis.

For Duverger, this criticism is largely irrelevant. For example, he acknowledges that prior to the 1988 reform the Finnish president was only elected indirectly and concedes that the nature of the electoral system was such that it led to ‘an election by notables much more than a popular election’ (Duverger 1978: 58). However, he also insists that election by notables is not the same as election by parties as in a parliamentary system (ibid.) and that the Finnish system resembled the US system where strictly speaking the president is also elected by an electoral college (Duverger 1978: 64). So, Duverger saw nothing in Finland's (pre‐reform) electoral system which was essentially incompatible with its status as a semi‐presidential regime. Equally, he saw nothing in the Irish propensity towards uncontested elections which might threaten its status either (Duverger 1978: 86).

While this may be a reasonable line of argument, it must also be acknowledged that there is at least some degree of ambiguity in this aspect of Duverger's definition and that this ambiguity needs to be addressed. So, for example, in his definition of semi‐presidentialism Sartori prefers to adopt a more stringent criterion. He states that an essential characteristic of a semi‐presidential regime is that the president ‘is elected by a popular vote—either directly or indirectly—for a fixed term of office’ (Sartori 1997: 131). Indeed, he insists on this wording at some length because he considers that US and (pre‐reform) Finnish‐style indirect elections closely resemble Latin American‐style direct elections particularly in that all are increasingly susceptible to what he calls ‘video‐politics’ or the opportunity for political outsiders to bypass the party system and manipulate television in the pursuit of votes. Thus, he prefers to reword Duverger's original definition so as not to risk excluding countries like Finland from the list of semi‐presidential regimes. To avoid confusion, this seems to be a sensible solution. It is appropriate, therefore, to adopt Sartori's approach.

The final problem with the concept of semi‐presidentialism is also derived from the wording of Duverger's standard definition. This time, the problem stems from the second element of the definition which states that in a semi‐presidential regime the president ‘possesses quite considerable powers’. This wording provokes writers to make one or other of two objections: either that the concept of semi‐presidentialism is incoherent and should be ditched altogether or that the concept is coherent but that Duverger's list of semi‐presidential regimes should be revised. The first objection, it might be argued, is unfounded, but the second does suggest that there is a problem with Duverger's definition which needs to be addressed.

The first objection of this sort is made by those writers who focus on Duverger's standard list of semi‐presidential regimes. These writers note that the list contains some countries with very strong presidents and others with very weak presidents. This, they suggest, undermines the whole concept of semi‐presidentialism because it lumps together countries which are too dissimilar. For example, Nogueira Alcalá states that from a purely legal point there are indeed six West European semi‐presidential regimes (1986: 134). However, he also states that from a political point of view the term ‘semi‐presidential’ leads to a ‘misleading appreciation’ of the Austrian, Icelandic, and Irish cases where there are weak presidents and argues that it is clearer and ‘more exact’ to classify all six countries as having a ‘dual executive with a presidential corrective’ (Nogueira Alcalá 1986: 135). Similarly, Cohendet states that Austria, Iceland, and Ireland are parliamentary, that Portugal is only intermittently semi‐presidential, that Finland was for a long time an exceptional case because of its proximity to the USSR and that the only real example of a semi‐presidential regime is France (1993: 74–5). Consequently, she prefers to distinguish between monist birepresentative parliamentary regimes (in which there are two sources of popular authority but only one controlling power, presidential or prime ministerial), such as Austria, France, Iceland, and Ireland, and dualist birepresentative parliamentary regimes (in which there are two sources of popular authority and two controlling powers, presidential and prime ministerial), such as Finland and Portugal (Cohendet 1993: 77). Finally, Shugart and Carey respond to this problem by distinguishing between premier‐presidential regimes, which indicates the primacy of the prime minister as well as the presence of a president with significant powers, president‐parliamentary regimes, which establish the primacy of the president and the dependence of the cabinet on parliament, and parliamentary (with president) regimes, where the popularly elected head of state is simply a figurehead (Shugart and Carey 1992: 18–27).

Duverger has always been quick to reply to this line of argument. His standard defence is to point out that similarly diverse political practices occur in other more uncontroversial regime types. For example, in 1978, he argued that, despite the fact that the German and Italian systems work so differently, ‘everyone puts [them] in the same category: parliamentary regimes’ (Duverger 1978: 18). He then goes on to add: ‘It is no more (or less) artificial to place France, the Weimar Republic, Finland, Austria, Iceland, Ireland and Portugal in another category: semi‐presidential regimes. . . ’ (ibid.). More recently, Duverger has reiterated this argument, stating that parliamentary regimes are just as diverse as semi‐presidential regimes: ‘you only have to compare the institutions in London with those in Rome to be aware of this’ (1991: 113). By 1992 Duverger was once again comparing the German and Italian systems, concluding that ‘parliamentary regimes demonstrate just as much heterogeneity [as semi‐presidential regimes]’ (1992: 902).

For Duverger, then, the fact that political practice in the six West European semi‐presidential regimes is so diverse does not mean that the concept of semi‐presidentialism is undermined. Instead, it is simply a reflection of the fact that countries with the same basic constitutional structure can operate in a variety of different ways. In this respect, Duverger's argument is sound. There is indeed just as much diversity amongst parliamentary regimes as semi‐presidential regimes. Indeed, it might be added that there is just as much diversity amongst presidential regimes as semi‐presidential regimes. For example, the US operates very differently from Mexico and yet both countries are unequivocally classed as presidential. Therefore, the fact that there is indeed a variety of political practice across semi‐presidential regimes does not undermine the fundamental validity of the concept itself.

The second objection of this type is made by those writers who take Duverger's definition literally and who then proceed to reconstitute the list of semi‐presidential countries. They accept the validity of the concept of semi‐presidentialism but note that it only includes countries which have presidents who possess ‘quite considerable powers’. They then proceed to eliminate countries with weak presidents from the list of semi‐presidential regimes. So, for example, when identifying semi‐presidential regimes Mainwaring states that ‘what matters is whether [presidential] offices are largely symbolic or, conversely, whether the office holders wield considerable power’ (1993: 203). On the basis of this logic, he argues that there are just two stable semi‐presidential democracies, Finland and France (ibid. 205). On the basis of a similar logic, Stepan and Skach also noted only two examples of semi‐presidential regimes, France and Portugal (1993: 9). Equally, Ieraci named just one, France (1994: 63). These writers, then, accept that there is such a thing as a semi‐presidential regime but classify only those countries with relatively strong presidents as examples of such a regime. As a result, the list of semi‐presidential countries varies from one writer to another according to each writer's subjective judgement as to what constitutes a ‘relatively strong president’.⁷

In contrast to the previous objection, it might be argued that there is some justification for this line of argument but, it might also be argued, this does not mean that the list of presidential regimes should be redrawn. It is certainly the case that the wording of Duverger's definition invites people to eliminate from the list of semi‐presidential regimes those countries whose presidents do not possess ‘quite considerable powers’. As things stand, therefore, writers such as Mainwaring, Stepan and Skach, and Ieraci are interpreting Duverger's definition quite logically and consistently. And yet, it might also be argued that there is a basic problem with their approach. This is because it allows different writers to provide their own interpretation of the powers of presidents and to draw up their own preferred list of semi‐presidential regimes on the basis of this interpretation. In other words, it allows the classification of regime types to become an essentially subjective exercise. (See the argument in Elgie 1998.) However, this subjectivity should be avoided because it poses problems for the study of comparative politics. The very reason for establishing concepts such as presidentialism, parliamentarism, and semi‐presidentialism is so as to be able to compare similar regime types more accurately. So, if different writers are able to draw up their own subjective list of semi‐presidential regimes, then it follows that those writers will not be comparing like with like and so the basis of the comparison is weakened. (This point will be considered in more depth in the last chapter.)

This suggests, then, a problem with the interpretation of Duverger's definition. This problem is caused by Duverger's stipulation that a semi‐presidential regime must exhibit a president who possesses ‘quite considerable powers’. In order to eliminate this problem, what is needed is a definition of semi‐presidentialism which excludes the opportunity for subjective classifications of semi‐presidential countries and establishes a clear‐cut list of semi‐presidential regimes. This can only be achieved if Duverger's definition is slightly reformulated. Only then will it be possible to arrive at an unambiguous list of semi‐presidential regimes which promotes the objective study of comparative politics (Elgie 1998).

It has been demonstrated both that the concept of semi‐presidentialism has been the subject of a degree of confusion over the years and also that the criticisms of both the first and second elements of Duverger's standard definition are to an extent justified. Thus, it is necessary to reformulate the concept of semi‐presidentialism so as to dispel the confusion and take account of these criticisms. To this end, therefore, the following reformulation will be proposed:

A semi‐presidential regime may be defined as the situation where a popularly elected fixed‐term president exists alongside a prime minister and cabinet who are responsible to parliament.⁸

This is a purely constitutional definition of the concept.⁹ Moreover, it is a definition which simply indicates the ways in which the head of state and head of government come to office and how they remain in office. It does not make any assumptions about the actual powers of these two actors. This is also a clear and straightforward definition of semi‐presidentialism which has the advantage of remaining very close to Duverger's standard definition.¹⁰ Moreover, it has two further advantages. First, it takes account of Sartori's point that the first element of Duverger's definition is potentially misleading. It does so by replacing the implication that a directly elected president is needed for a semi‐presidential regime with the notion that a popularly elected president is required, meaning a president who is directly elected or is elected in a ‘direct‐like’ manner. So, countries such as pre‐reform Finland can unequivocally be classed as semi‐presidential regimes. Secondly, it omits altogether the second element of Duverger's definition which refers to presidential powers. This means that the problems caused by this reference are removed. As such, countries with weak presidents, such as Austria, Bulgaria, Iceland, and Ireland, can unequivocally be classed as semi‐presidential alongside countries with strong presidents, such as France and Russia, as well as countries with some sort of limited presidency, such as Finland, Poland, and Portugal. In this way, then, semi‐presidentialism emerges as an example of a pure type of regime which exists alongside other such pure types, most notably, presidential regimes and parliamentary regimes (Elgie 1998).

On the basis of this definition, a list of regimes which can unambiguously be classed as ‘semi‐presidential’ can be established (see Figure 1.1).¹¹ This list includes the six West European countries that Duverger has consistently identified as being semi‐presidential. It also includes the large number of countries in Central and Eastern Europe and the former USSR which adopted a semi‐presidential form of government after 1989 or 1991. It includes the two most frequently discussed examples in South and South‐East Asia, Sri Lanka and South Korea, respectively. Finally, it includes a large number of semi‐presidential regimes in Africa and a small number in the Americas. Overall, if a head count of regime types in democratic political systems were to be conducted, it would find that semi‐presidentialism is more widespread than presidentialism (although less so in Central and South America) and is perhaps only slightly less widespread than parliamentarism (although more so in Central and Eastern Europe and countries of the former USSR).

By definition, semi‐presidential regimes share the same basic constitutional structure. They all have presidents who are elected in a direct or direct‐like manner and they all have prime ministers and cabinets who are responsible to the legislature. As has already been indicated, though, in practice semi‐presidential countries operate in many different ways. The constitutional power of presidents, prime ministers, and cabinets varies just as the political power of presidents, prime ministers, and cabinets varies. Most notably, constitutionally strong presidents are sometimes politically weak and constitutionally weak presidents are sometimes politically strong. Presidents sometimes dominate prime ministers. Prime ministers sometimes dominate presidents. Sometimes neither one dominates the other. In order to examine the politics of semi‐presidentialism, therefore, it is necessary to establish a framework which captures the variety of political practices from one country to another. This is the aim of the next section.

Duverger has frequently reiterated that ‘the purpose of the concept of semi‐presidential government is to explain why relatively homogeneous constitutions are applied in radically different ways’ (1980: 177). For Duverger, then, the concept of semi‐presidentialism is as much an heuristic device as a description of a particular set of constitutional arrangements (Duverger 1986d: 8; 1986c: 349; and 1982: 193). For him, the principal advantage of such a device is that it ‘permits the construction of an analytical model which allows the in‐depth explanation of how these regimes function. . . ’ (Duverger 1986d: 14). Moreover, also according to him, ‘it is not only a question of explaining past and present incarnations of semi‐presidential regimes but also predicting their future incarnations. . . ’ (Duverger 1978: 89–90). The basis of this analytical model is the identification of the appropriate set of variables which account for why semi‐presidential regimes operate in such different ways.

Those who have followed Duverger have frequently identified their own set of variables. For example, Bartolini states that factors exogenous to the institutional system need to be identified in order to account for why countries with the same constitutional features operate in practice so differently (1984: 225). For him, four factors are important: the politico‐cultural origins of the regime in question; the process by which presidential and parliamentary candidates are selected; the relationship between presidential and parliamentary electoral systems; and the relationship between the president and party‐system coalition‐building (ibid. 226–7). In a similar vein, Linz argues that ‘it is impossible to analyze the performance of a bipolar regime independently of the larger political system. . . ’ (1994: 51) and in this respect he singles out two factors which are particularly important, the party system and the ‘complex historical situation’ (ibid.). Equally, in his analysis of semi‐presidential regimes Pasquino focuses on two variables, the electoral system and the party system (1995: 59).

It is apparent from this list that there is at least some degree of consensus as to the factors which most appropriately explain the variety of practices to be found in semi‐presidential regimes. As might be expected by now, though, the list of factors which Duverger himself identifies has varied over the years.¹² In general, however, Duverger's list reinforces this consensus. For the most part, Duverger considers three variables to be of particular significance: the constitutional powers of the major political actors; the events surrounding the formation of the regime; and the nature of the parliamentary majority and the relationship between the president and the majority. Each of these variables will briefly be considered.

As noted above, by definition semi‐presidential regimes all operate within the same basic constitutional procedures. Over and above these procedures, though, the constitutional powers of presidents, prime ministers, and parliaments vary. This variety helps to account for the diversity of semi‐presidential politics.

For Duverger, variations in constitutional powers can be captured by reference to three general types of situations. The first type is where the president is merely a ‘controlling force’ (Duverger 1980: 177). In this situation, the president simply acts as the guardian of the constitution and may have the right, for example, to refer laws to the constitutional court and propose a constitutional referendum. The second type represents an intermediate situation in which the president enjoys these controlling powers and also has the unilateral right to dismiss the prime minister (ibid.). The third type is where the president is a ‘governing’ force (Duverger 1980: 178). In this situation, the president ‘shares in the running of the country, in collaboration with the prime minister and the cabinet’ (ibid.).

Even though Duverger distinguishes between these three types of constitutional situations, he also takes great pains to emphasize that constitutional rules and political practice do not always coincide. In some cases presidents who would appear to be in a position only to operate as a controlling force in fact operate as a governing force and in other cases the opposite is true. It would be wrong to conclude from this, though, that Duverger believes constitutional powers to be irrelevant to the practice of semi‐presidential regimes. Indeed, he clearly states that ‘the constitution plays a certain part in the application of presidential powers’ (Duverger 1980: 179). Nevertheless, he also states that constitutional factors remain only ‘secondary compared to the other parameters’ (ibid.).

In this way, therefore, in order to understand the comparative politics of semi‐presidential regimes and the reasons why such regimes operate so differently, it is appropriate to outline the constitutional powers of presidents, prime ministers, and parliaments. It is necessary to know whether the president can dismiss the prime minister, dissolve the legislature, appoint government ministers, assume emergency powers, and so on. Collectively, these powers indicate both the de jure balance of power between the various political actors and they also usually provide at least a hint (and sometimes more) of the de facto relationship between them as well.

The second factor which helps to explain the variety of semi‐presidential regimes concerns the historical, or politico‐cultural, context within which the regime was created. It is hardly surprising that this factor should be invoked to explain the variety of semi‐presidential regimes because, necessarily, each country's context is unique. This contextual factor, then, helps to engender national differences that persist over time and which can distort the operation of the set of de jure constitutional rules.

Each country operates within a given geographical area, against the background of a particular historical situation and according to the dynamics of a specific constitutional foundation. Nevertheless, certain similarities can be traced from one country to another. Three common types of context can be singled out. The first type concerns the situation where a semi‐presidential regime is adopted for purely symbolic reasons. This may occur, for example, when the adoption of semi‐presidentialism is associated with the process of national self‐determination. For example, if prior to independence the head of state was a foreign monarch, then the subsequent creation of a semi‐presidential regime with a popularly elected president may be motivated by the desire to reinforce the democratic credentials of the new regime rather than the desire to install a powerful head of state. In these cases, then, semi‐presidentialism may coincide with a weak presidency. The second type concerns the situation where a semi‐presidential regime is adopted for reasons of governability. This may occur, for example, when a semi‐presidential regime is adopted following the collapse of, say, a parliamentary system of government. Here, there may be a desire to create a strong leadership figure who will give direction to the new regime and prevent a repeat of the previous situation. In these cases, semi‐presidentialism may coincide with a strong presidency. The third type concerns the situation where a semi‐presidential regime is adopted during the transition to democracy. Here, one of several motivations may be present. For example, the presidency may be tailor‐made for the leading figure in the democratization process, so creating the conditions for a strong president. Equally, the presidency may be designed so as to prevent one person from assuming too much power, so creating the conditions for a weak president. Alternatively, the establishment of the regime may be the product of a ‘fudge’. In this case, the powers of president, prime minister, and parliament may be shared. Whatever the situation, the context surrounding the creation of the regime creates the opportunity for a great variety of political practices to occur across the set of semi‐presidential countries.

In his work, Duverger stresses the importance of the ‘combination of tradition and circumstances’ (1980: 180) in the evolution of semi‐presidential systems. Moreover, Duverger stresses not just the importance of, as it were, the moment of constitution‐building but also the conventions of political practice that endure thereafter. As Duverger states, countries develop a ‘factual tradition’ (ibid.). They initiate rules, norms, and procedures which subsequently become fossilized. In this context, the presidency of the first incumbent of the presidential office is often very important. If the first president is a figurehead, then the chances are that a figurehead presidency will become the norm. By contrast, if the first president is an authoritative decision‐maker, then, the likelihood is that a working presidency will be established. Whatever the outcome, it is apparent that events surrounding the formation of the regime are central to an understanding of the politics of semi‐presidentialism.

Since the first formulation of the concept Duverger has stressed that party political factors are fundamental to the operation of semi‐presidential regimes. For example, in 1971 he stated that ‘the structure of parties and the relationship between them is more important than constitutional powers’ when explaining why semi‐presidential regimes function so differently (Duverger 1971: 116). By 1978 Duverger had settled on his standard formulation of this variable stating that the mechanics of semi‐presidential regimes vary according to the nature of the parliamentary majority and the relationship between the president and the majority.¹³

The nature of the parliamentary majority can take a number of forms. The first case concerns the situation where there is an absolute parliamentary majority. Here, various scenarios present themselves. On occasions there may be a monolithic majority, implying that a single party enjoys a majority of seats in the legislature. On other occasions there may be a coalition majority with one dominant party, meaning that the position of the dominant party is strong but less so than in the previous situation. On yet other occasions there may be a balanced coalition majority, suggesting that power is shared between the majority parties in parliament. On all occasions the government's position is likely to be safe but only in the first scenario is the problem of inter‐party bargaining likely to be absent. The second case concerns the situation where there is only a relative or quasi‐majority in parliament. In this case, one party has more seats in parliament than any other but lacks an overall majority. Here, the position of the leading party may either be quite secure if it takes an ‘unholy alliance’ of political opponents to combine to bring the government down or it may be perilous if there is an alternative government waiting in the wings. The final case concerns the situation where there is no parliamentary majority at all. Here, the seats in parliament are shared between a large number of small parties and governments are supported by unstable and shifting coalitions.

Just as the nature of the parliamentary majority can take a number of forms, so too can the relationship between the president and the majority. For example, the president may be the leader of the majority or she or he may simply be a member of the majority. Equally, the president may be from the opposition or, alternatively, she or he may be a completely neutral figure altogether. By themselves, these various situations tell us very little about the type of semi‐presidential regime which is likely to ensue. Instead, they only help to explain the differences between semi‐presidential regimes when each is combined with the various forms of parliamentary majority that can occur. So, for example, Duverger argues that a president who is the leader of a monolithic majority will emerge as an absolute (republican) monarch (1980: 186). By contrast, a president who is simply the member of a party which only has a relative parliamentary majority will operate as a symbolic figurehead leader (ibid.). Overall, in his classic work on semi‐presidentialism Duverger identified 17 separate situations which may arise by combining the various forms of the parliamentary majority and the various relationships between the president and the majority (ibid.). Thereafter this number was reduced (Duverger 1982: 230), but the basic argument remained the same, namely that the politics of semi‐presidentialism varies as a function of party political factors.

This book examines the politics of semi‐presidentialism. In particular, it examines the politics of European semi‐presidentialism focusing on the experience of a large number of semi‐presidential regimes in Central and Eastern Europe, Western Europe, and the former USSR. In this context, the main question which is being asked is the following: why do countries which share the same basic institutional structure operate so differently in practice? As we have seen, Duverger's work on semi‐presidentialism provides a framework with which we can begin to answer this question. As such, in the chapters which follow particular attention will be paid to the constitutional powers of political actors, the circumstances surrounding the creation of the regime, the nature of the parliamentary majority, and the relationship between the president and that majority. At the same time, though, other factors will also be shown to be important in particular countries and these will be identified when and where appropriate. In the conclusion we will consider what the experience of semi‐presidentialism tells us about the academic debate concerning comparative institutional engineering and the pros and cons of presidential, semi‐presidential, and parliamentary regimes.