An Intellectual History of the Concepts of Premier-Presidentialism and President-Parliamentarism

The Intellectual History of the Two Concepts

The original definition of the concepts of premier-presidentialism and president-parliamentarism can be found in Presidents and Assemblies. Shugart and Carey (1992: 23) defined premier-presidentialism as a regime ‘in which there is both a premier (prime minister), as in a parliamentary system, and a popularly elected president’. Specifically, they repeated Duverger’s three definitional criteria for semi-presidentialism as being constitutive of a premier-presidential regime, namely:

Then, Shugart and Carey (1992: 24) referred to president-parliamentarism as the situation where ‘both the president and the parliament have authority over the composition of cabinets’. Again, specifically, they identified four criteria as being constitutive of a president-parliamentary regime. They were:

Thus, there is a sense in which the concepts of premier-presidentialism and president-parliamentarism came into the world fully formed and ready to go. That said, there have since been some not insignificant changes to the two concepts relative to how they were originally presented in 1992 and to how they are now commonly defined.

The first change concerned the relationship between Duverger’s (1980) idea of semi-presidentialism and Shugart and Carey’s two concepts. The issue is whether premier-presidentialism was simply another name for semi-presidentialism with president-parliamentarism being a different regime type altogether, or whether premier-presidentialism and president-parliamentarism were two sub-types of semi-presidentialism. In Presidents and Assemblies, the very first reference to premier-presidentialism might imply that they are sub-types. Shugart and Carey (1992: 6) state that ‘[s]ome of these regimes [with the combination of an elected presidency with a cabinet dependent on the assembly’s confidence] constitute a regime type we shall define as premier-presidential’. The use of the word ‘some’ here might be taken to mean that some are, but some are not. In that event, those that are not might be the ones they define as president-parliamentary. Thus, the two concepts could constitute two sub-types of semi-presidentialism right from the off. This is the interpretation that Shugart later privileged. In 2005, he referred for the first time to the ‘. . . subtypes [of semi-presidentialism] that Shugart and Carey introduced . . .’ (Shugart, 2005: 333). This soon became Shugart’s standard way of presenting the two concepts. In 2006, he stated that ‘Under the rubric of semi-presidentialism, there is much variation in formal powers, leading Shugart and Carey (1992) to propose a further subdivision of the concept into premier-presidential and president-parliamentary subtypes’ (Shugart, 2006: 357). This point was re-emphasised in what we consider to be Shugart’s ‘mature’ interpretation of the two concepts in Samuels and Shugart (2010). There, they state ‘Given the vagueness of the role of the presidency in early definitions of semi-presidentialism, Shugart and Carey (1992) introduced the concepts of premier-presidential and president-parliamentary subtypes of semi-presidentialism’ (Shugart and Carey, 1992: 30). This is certainly the way in which the two concepts are now understood. There is an overarching concept of semi-presidentialism, which has two sub-types, premier-presidentialism and president-parliamentarism. Arguably, this interpretation can be traced back to the original formulation of the schema in 1992.

It is not clear, though, that this interpretation is entirely accurate at least from the written record. In Presidents and Assemblies, there is a sense in which semi-presidentialism is synonymous with premier-presidentialism and that president-parliamentarism is something quite different. Making it clear that they do not like the term ‘semi-presidential’, Shugart and Carey (1992: 23) state that ‘what Duverger refers to as semipresidential, we designate premier-presidential’, suggesting the two concepts are the same but with a different name. In the next paragraph, as noted above, they adopt Duverger’s definition of semi-presidentialism as their definition of premier-presidentialism, further suggesting they are synonymous and implying that president-parliamentarism must be different not only from premier-presidentialism but also from semi-presidentialism. At another point in the book, they note that some ‘students of democratic institutions have spoken of premier-presidentialism (what others have called “semipresidentialism”)’ (Shugart and Carey, 1992: 55), again implying that premier-presidentialism and semi-presidentialism are equivalent. At no point in the book do they say that president-parliamentarism is a sub-type of semi-presidentialism. At no point do they say that there are two sub-types of semi-presidentialism. Indeed, in work that followed relatively soon after the idea that there are sub-types of semi-presidentialism is not only absent, it seems to be rejected. For example, in 1999, Shugart wrote that president-parliamentary systems are sometimes ‘called semipresidential in some of the literature’, but then immediately states that ‘[t]he latter term, however, is somewhat misleading’, which leads him to ‘eschew the term semipresidential entirely’ (Shugart, 1999: 57). In a 2001 book chapter, there was no mention of semi-presidentialism or any sub-types either (Shugart and Haggard, 2001: 69). As noted above, it is only in 2005 that the idea of sub-types was explicitly articulated for the first time. In short, while there are some grounds for thinking that premier-presidentialism and president-parliamentarism should be considered as sub-types of semi-presidentialism as far back as 1992 (not least through the words of Shugart himself!), it is not until 2005 that this idea is first expressed in print.

The second early conceptual change was whether the presence of a prime minister and collective cabinet responsibility is a necessary requirement for a country to be classed as president-parliamentary. As we have seen, in Presidents and Assemblies, Shugart and Carey (1992: 23) take Duverger’s definition of semi-presidentialism as their definition of premier-presidentialism. This definition includes the stipulation that there exists ‘a premier and cabinet, subject to assembly confidence’. Thus, the presence of a prime minister and collective cabinet responsibility is a necessary condition of premier-presidentialism from the start. However, the definition of president-parliamentarism makes no reference to a prime minister or cabinet. In this regime, ‘the president appoints and dismisses cabinet ministers’ and ‘cabinet ministers are subject to parliamentary confidence’ (Shugart and Carey, 1992: 24). Here, then, ministerial responsibility seems to be purely individual. What is more, at another point in the book, they state that they regard Peru and Chile as president-parliamentary ‘for reasons more profound than simply the lack of a post of premier in the Chilean case’ (Shugart and Carey, 1992: 117) and perhaps because of the fourth element of original definition – ‘the president has the power to dissolve parliament or legislative powers, or both’ (Shugart and Carey, 1992: 24). Thus, at this point, the presence of a prime minister does not seem to be a necessary condition of president-parliamentarism, reaffirming the idea that president-parliamentarism was not seen as a sub-type of semi-presidentialism at this stage.

Over time, though, this way of understanding president-parliamentarism seems to have evolved. The fourth element of the original definition soon disappeared. More than that, in his 2005 article, Shugart (2005) expressly states that:

one should not, in my view, label as semi-presidential those systems in which: individual ministers may be removed from office by the assembly but there is no prime minister to coordinate assembly control over the cabinet (e.g. Colombia); or where a prime minister, but not the cabinet as a whole, can be removed by the assembly (e.g. Argentina); or where there is a prime minister but no provision for a binding vote of no confidence (e.g. South Korea) (Shugart, 2005: 327).

Indeed, Shugart (2005: 333) makes the point explicit, defining the two concepts as follows: ‘Under premier-presidentialism, the prime minister and cabinet are exclusively accountable to the assembly majority, while under president-parliamentarism, the prime minister and cabinet are dually accountable to the president and the assembly majority’ [my emphasis]. These definitions are repeated almost word-for-word in Samuels and Shugart (2010: 30). There, they further underline the change of definitional emphasis since 1992, stating that ‘Bolivia is most definitely not semi-presidential, since it has no prime minister accountable to the assembly majority’ (Samuels and Shugart, 2010: 28). There is, thus, a clear change in the definition and understanding of the concept of president-parliamentarism between its original formulation in 1992 and its ‘mature’ reformulation in 2005 that was reiterated in 2010.

The final early conceptual change concerned the question of whether a country should be classified as premier-presidential or president-parliamentary on the basis of the strict application of a particular constitutional rule, or a more general consideration of the constitutional situation. In Presidents and Assemblies, there is a sense that the general constitutional situation is what matters. For example, Shugart and Carey (1992: 41) place Ireland in their list of parliamentary regimes. This is despite the fact that there are direct presidential elections and, therefore, Ireland would seem to be premier-presidential. In the text, they are explicit about the Irish case. They state that Ireland is parliamentary (Shugart and Carey, 1992: 275). They make this claim on the basis that the president ‘has no constitutional powers’ (Shugart and Carey, 1992: 71). This statement is factually incorrect, even if it is certainly true that the President of Ireland has very few constitutional powers. The point is that Shugart and Carey seem to require a president to be endowed with at least a certain amount of power in order for a country to be classed as premier-presidential. At the same time, though, they state that ‘[o]ur definition of premier-presidentialism requires that a prime minister direct the cabinet . . .’ (Shugart and Carey, 1992: 74), implying that if a president directs the cabinet, then a country might not be classed as premier-presidential either. In other words, if a president has too many powers, then it should not be classed as premier-presidential. In this regard, they say that ‘Finland and Iceland appear to be intermediate regimes between president-parliamentary and premier-presidential types’ (Shugart and Carey, 1992:162), because of the power of their presidents. In sum, it is easy to read Shugart and Carey in 1992 as believing that a president needs some power, but not too much power in order for a country to be classed as premier-presidential. This issue is important because if Shugart and Carey are classifying countries on the basis of a general assessment of the power of the president, then they risk falling foul of the ‘Duverger problem’ in relation to his 1980 definition of semi-presidentialism. That is to say, if a country is premier-presidential only when a president has some powers but not too many, then who is to say where that threshold lies? Here, the potential for countries to be classified differently by different people becomes embedded in the classificatory schema.

Whether or not Shugart and Carey are guilty of the ‘Duverger problem’ in 1992, once again the issue was addressed in 2005. Here, Shugart (2005: 333) explicitly acknowledges the ‘Duverger problem’, noting that Duverger’s definition of semi-presidentialism is vague ‘because it does not define what the “considerable” powers of the president are’. At this point, as we have seen, he makes it clear that the application of a single constitutional rule – whether the cabinet is exclusively responsible to the legislature or dually responsible to both the president and the legislature – is sufficient to determine whether a country should be classed as premier-presidential or president-parliamentary. Indeed, in his 2005 article, he classifies a set of countries on this basis and he includes at least one country – Bulgaria – with a weak president in the premier-presidential category, though Ireland is not included in the classification. As before, we see Samuels and Shugart (2010: 32) as the ‘mature’ statement of Shugart’s position in this regard. Here, the simple constitutional classificatory criterion is again rigorously applied, and by this time, there is no doubt that countries with directly elected but very weak presidents, including Ireland, are to be classed as premier-presidential rather than parliamentary.

To be fair to Shugart and Carey, it is moot as to whether they were ever guilty of the ‘Duverger problem’ in Presidents and Assemblies. All the same, there was a certain conceptual clarification in this regard over time. This clarification led to Shugart changing at least one country classification. Arguably, though, it affected other countries too. For example, both Austria and Iceland were classified as premier-presidential in Presidents and Assemblies, whereas both are constitutionally president-parliamentary, even though they have very weak presidents in practice. In 2005, Austria was reclassified by Shugart as president-parliamentary (Shugart, 2005: 336). That said, it is possible that this change in classification was a function of a misreading of Austrian and Icelandic constitutional texts in 1992 rather than a subsequent change from a general consideration of the constitutional situation to the strict application of a single classificatory rule.

To sum up, the concepts of premier-presidentialism and president-parliamentarism have been present in the scholarly consciousness since 1992. Since that time, they have been subjected to a small number of conceptual changes or clarifications that have had certain classificatory consequences. This is not a criticism. In fact, it is remarkable that there has not been more change over time. What we can say, though, is that for the most part by 2005 and certainly by 2010, both the concepts and their classificatory implications were finalised. This means that we now understand premier-presidentialism and president-parliamentarism to be two sub-types of semi-presidentialism. In this context, the standard definitions of the two terms are now as follows (Samuels and Shugart, 2010: 30):

On the basis of these definitions, we can identify a definitive list of premier-presidential and president-parliamentary countries simply by reading a country’s constitution and applying a parsimonious classificatory rule. In other words, we now have an essentially reliable way of identifying the set of premier-presidential and president-parliamentary countries. This is now the state of the art when it comes to identifying semi-presidential regimes and classifying countries as examples of one or other of the two sub-types. For sure, the precise wording of constitutions sometimes has to be interpreted. For the most part, though, constitutions make it very clear regarding whether or not the prime minister and cabinet are accountable to the president. In the meta-analysis that follows, we assume that a correct country classification is one that corresponds to the 2010 list in Samuels and Shugart (2010: 32).

Overall, therefore, we have two concepts that can reliably capture two constitutionally distinct situations. In 1992, Shugart and Carey made the case that this distinction was consequential. There were different political consequences depending on whether a country adopted a premier-presidential or a president-parliamentary regime. In so doing, they established a research agenda. Since this time, many scholars have pursued this agenda, including Shugart himself. Over time, he developed what he calls a ‘neo-Madisonian’ perspective on transactional and hierarchical authority patterns that correspond to different regime types, including ‘premier-presidentialism’ and ‘president-parliamentarism’ (Shugart, 2005). In the next section, we identify the areas of study that have been pursued and the conclusions that have been reached through a rigorous meta-analysis of the work on the two sub-types of semi-presidentialism.

A Meta-Analysis of the Two Concepts

In July 2017, we conducted a search for references to the terms ‘premier-presidential’, ‘premier-presidentialism’, ‘president-parliamentary’ and ‘president-parliamentarism’ in a large number of journals, journal databases and search engines. Specifically, we conducted a search in Google Scholar (examining the first 25 pages of returns), Google Books, JSTOR, EBSCO, ProQuest, Project Muse and Sage journals online, plus many individual publications, including but not confined to West European Politics, Democratization, Electoral Studies, Party Politics, Journal of Legislative Studies, Comparative Political Studies, British Journal of Political Science and American Political Science Review. We recorded any reference to the search terms in English-language publications. We included references in books and book chapters. However, we excluded conference papers, working papers, PhD or MA theses and short book reviews in journals. We also excluded references that were clearly unrelated to Shugart and Carey’s concepts, such as a book about the ‘premier-presidential’ New Hampshire primary in the United States. Undoubtedly, we failed to pick up some references. However, given the amount of repetition that we encountered searching across the different platforms, we suspect that the missing items are likely to be brief one-off mentions. Thus, we are confident that we have not missed any major studies. What is more, given that we did record many one-off mentions, we are also confident that the data are not systematically skewed against this type of reference.

In total, we found 276 references to premier-presidentialism and president-parliamentarism since 1992. Over time, the total number of references per year has tended to increase (see Figure 1). This has the effect that the 50th percentile was reached in 2009. There were 146 references from 1992 to 2009 inclusive and 130 references from 2010 onwards, noting that the coverage only includes part of 2017. As we have already seen, the ‘mature’ definition of the two concepts and the final list of country classifications were articulated in 2010. Given the total set of references is also distributed around this date, we use these two periods as a heuristic analytical device in the rest of this article. We label the first (1992–2009) as the ‘early’ period and the second (2010–) as the ‘contemporary’ period.

OPEN IN VIEWER

Figure 1.

Annual Number of References to Premier-Presidentialism and President-Parliamentarism, 1992–2017.

Having identified a reference to the concepts, we made a basic coding distinction between a brief mention and a more substantive reference. Only one coder was involved. A brief mention was typically just an isolated mention of the term with or without a citation to the original Shugart and Carey schema or a subsequent Shugart version of it. A substantive reference involved something more discursive, ranging from a short overview of one or other of the concepts in a paragraph of an article, book, or chapter, to whole books that focused on the concepts. A close reading of all the substantive references was undertaken.

Of the 276 references in total, 137 were coded as substantive references and 139 were brief mentions. Up to around 2004, there were typically more brief mentions annually than substantive references. Since then, the trend has been reversed. In total, there were 80 brief mentions in the early period and 59 in the contemporary period, whereas there were 66 substantive references in the former and 71 in the latter. Overall, we can say that the frequency of references has increased over time and the ratio of substantive references to brief mentions has also increased. This suggests that there is continuing, indeed growing substantive interest in Shugart and Carey’s typology.

We then analysed the brief mentions. Of the 139 brief mentions in total, 87 actually cited the basic Shugart and Carey typology, noting that only five did so critically. This perhaps implies that those who reject the typology tend to ignore it rather than mentioning it negatively. Of these 87 brief mentions, 13 were made with a view to classifying a country as an example of either premier-presidentialism or president-parliamentarism. In addition, a further 44 brief mentions referred to one or other of the two terms, but did so without explicit reference to the typology. This indicates that the terms have perhaps so thoroughly entered the language of political science that citation back to the original source is no longer deemed to be necessary.

Of these 44 brief mentions, 22 did so to classify a country. The remaining eight brief mentions were econometric studies that built on Persson and Tabellini’s (2003) work. These studies all briefly mentioned the existence of the two sub-types of semi-presidentialism in the Appendix to their studies, but simply recoded president-parliamentary countries as presidential and premier-presidential countries as parliamentary and estimated the relative effect of each. Thus, the typology made no independent contribution to this econometric work, but there was a brief mention of the typology. These eight cases did not make any citation back to the original source. Therefore, they could be added to the other 45 brief mentions of this sort. In sum, we have found that reference to Shugart and Carey’s typology is often briefly made in passing with or without a citation back to Presidents and Assemblies. We have also found that around 25% of the time when there is a brief mention, it is with a view to classifying a single-country as an example of either premier-presidentialism or president-parliamentarism.

We wished to know the reason why scholars were referring to Shugart and Carey’s typology. To do so, we coded the references into five types of studies: a long review of the concepts; a single-country or comparative country-classification exercise; a more general typological exercise; a study of institutional choice; and an empirical study of the effects of two regimes on some or other outcome. References could be classed as contributing to more than one type of study. It is worth noting that a brief mention could be coded as a country classification, for example, the simple statement ‘Poland is a premier-presidential regime’ with no other reference to the typology or even a citation to Shugart and Carey counted as a country classification, but a brief mention could not be attributed to any of the other four types of study, which required a more substantive contribution.

On this basis, we found that scholars were predominantly referring to the concepts to engage in a single-country or comparative country-classification exercise with empirical studies being by far the next most important focus of inquiry (see Figure 2). We found that the country-classification exercise is ongoing, with around the same number of examples in both the early and contemporary periods. However, we found some variation in the frequency of other types of studies over the two periods. There were no studies of institutional choice in the second period. There was, though, an increase in the number of review articles that included substantive reference to the typology. An example is Schleiter and Morgan Jones’ (2009) review of work on semi-presidentialism that incorporates the work on the two sub-types of semi-presidentialism in its remit. This increase in the frequency of review articles makes sense, given there is now a substantive body of work focusing on Shugart and Carey’s typology. By contrast, we found a sharp decline in articles providing a new typology that either discussed or incorporated Shugart and Carey’s schema in such an exercise. This is interesting, because it perhaps suggests that the scientific community has come to a well-settled belief about the validity of existing typologies, including Shugart and Carey’s. In other words, the desire to establish a new typology because of the perceived problems with existing ones has seemingly diminished. It is also worth noting in this regard that the frequency of empirical studies has increased over the two periods. This is consistent with the previous point. If scholars have indeed settled on the validity of existing typologies, including Shugart and Carey’s, then the obvious question to ask is ‘what difference does the institutional variation in the typology make?’, thus encouraging empirical studies to provide an answer to this question.

OPEN IN VIEWER

Figure 2.

Distribution of References to Premier-Presidentialism and President-Parliamentarism by Subject.

One of the strengths of the Shugart and Carey’s schema in its mature form is that it provides the opportunity for the reliable classification of countries as either premier-presidential or president-parliamentary (Elgie, 2016). As we have seen, all that is required is a reading of a country’s constitution and the strict application of a single constitutional classificatory rule. This is no doubt one of the main reasons why the typology has become well-settled generally and also why we have found that so many scholars have used it to engage in a country-classification exercise. However, scholars can make mistakes. In total, we identified 80 single-country classifications using Shugart and Carey’s schema and 52 comparative classifications. The latter ranged from a two-country comparison, to regional comparisons of Eastern Europe or elsewhere, to world-wide classifications. On the basis of the 2010 list cited above, we found that 28 (21.2%) of the 132 classification exercises were either incorrect in the case of a single-country classification or included at least one incorrect country classification in the case of a comparative exercise. The distribution of these errors over time, though, was startling. In the early period, 22 (34.9%) of the 63 such exercises were incorrect or included an incorrect classification, whereas in the contemporary period, the figure fell to just 6 (8.7%) of the 69. The decline in error is stark. However, we have already shown that even Shugart and Carey themselves were guilty of a small number of classification errors in 1992. What is more, to the extent that their 1992 schema ran the risk of the ‘Duverger problem’, then there was greater potential for misclassification in the early period than the contemporary period, which coincides with the 2010 schema where there is a reliable list of country classifications to which scholars can refer. For these reasons, even if the decline in error is great, it is not entirely surprising that the early period includes more incorrect country classifications than the contemporary period.

Focussing on single-country classifications, 13 (16.3%) of the 80 were incorrect over the whole time, comprising 8 (21.1%) of the 38 in the first period and 5 (11.9%) of the 42 in the second. In terms of the distribution of errors, we should not place too much emphasis on the total number of classification errors per country. This is because some countries simply receive more scholarly attention than others and so there is more opportunity for scholars to make a mistake. Nonetheless, we do find that the classification of Russia and Taiwan has been particularly problematic. There have been 11 single-country classifications of president-parliamentary Russia since 1992, 4 of which are incorrect. There have also been seven classifications of president-parliamentary Taiwan in the same period, three of which are incorrect. Specifically, they have both been classed as presidential, rather than president-parliamentary.

The problems with Russia and Taiwan might suggest there is a particular issue classifying president-parliamentary countries. However, there have been 15 single-country classifications of Ukraine, which has been president-parliamentary for much of its post-independence period, but only one of those classifications is incorrect. We also identified seven single-country classifications of Portugal using Shugart and Carey’s typology. Portugal experienced a president-parliamentary period from 1976 to 1982, yet all the Portuguese country classifications are correct. We should note that the ‘incorrect’ classifications we record do include a few cases where scholars have deliberately challenged Shugart and Carey’s classificatory label for a particular country (see below).

In terms of comparative country classifications, the difference in errors between the two periods is very pronounced. In total, 15 (28.8%) of the 52 comparative classification exercises included at least one country-classification error. In the early period, there were errors in 14 (56.0%) of the 25 such exercises, but in the contemporary period, only 1 (3.7%) of the 27. We record the original Shugart and Carey (1992) classification as an error in this regard, notably regarding Austria, Iceland and Ireland as discussed previously. In a 1996 paper, Shugart (1996: 9) still classed Austria as premier-presidential. In his 2005 paper, Shugart (2005: 336) rectified the Austria error, but made a mistake regarding Sri Lanka, which was classed as premier-presidential, rather than president-parliamentary.

It is worth noting that Roper’s (2002: 259) influential study of variation under premier-presidentialism repeats the 1992 misclassification of Austria and Iceland, citing Shugart and Carey as the reason why these countries were included in a study that focused solely on this regime type. Roper also fails to include Ireland in his study of premier-presidential regimes, presumably because at this point it is still being classed as parliamentary according to the Shugart and Carey typology. In a footnote, though, Roper (2002: 270) does note that Iceland ‘technically should not be considered a premier-presidential regime because the president can dismiss any minister of government’. However, he goes on to state, ‘because this dismissal power has never been used, most scholars consider the country to have a premier-presidential regime’ (Roper, 2002). This confirms both that president-parliamentarism is being thought of in terms of individual ministerial responsibility rather than collective prime ministerial and cabinet responsibility at this point and that scholars are thinking about applying the Shugart and Carey schema in terms of general assessment of presidential power rather than the strict application of a particular constitutional rule.

Unsurprisingly, the classification errors in the comparative exercises concern cases where the president is either very weak or very strong. In the former category, Bulgaria and Slovenia are misclassified as parliamentary three times and Slovakia twice. In the latter, errors include the classification of Mozambique as presidential. The only comparative misclassification in the contemporary period occurred right at the start of this period. Here, Pasynkova (2010: 289–290) cites Shugart and Carey when correctly classifying Poland as premier-presidential and Russia as president-parliamentary, but makes a mistake when classifying Bulgaria and Slovakia as parliamentary (Pasynkova, 2010: 294). In general, we can say that since the mature classification of country cases in Samuels and Shugart (2010), there is now no reason why scholars who wish to apply Shugart and Carey’s schema should make a country-classification error. Indeed, we have seen that the number of such errors has been all but eliminated.

We now turn to the empirical studies about the effects of institutional variation. To begin, we identified when the concepts of premier-presidentialism and president-parliamentarism were operationalised quantitatively as an explanatory or control variable, meaning that the relative effect of the two concepts was explicitly estimated in a statistical model. We also identified when the concepts were employed qualitatively, meaning that there was some general discussion of the effect of one or other, or both, of the two sub-types of semi-presidentialism on some outcome of interest, but where a statistical model was not the evidential basis for a general conclusion. In this latter category, we included quantitative studies that placed their empirical analysis in the context of one of the two sub-types, but where the variation was not included in the statistical model. These were typically single-country studies where there was no sub-type variation to operationalise, but where the results were considered to be generalisable to the particular sub-type in question. An example is Magalhães and Aguiar-Conraria’s (2009) study, which examines voting behaviour in Portugal in the context of the country’s premier-presidential system. They argue that such systems generate executive conflict that can affect voter choice. Here, the econometric model operationalises variation in conflict to explain voting, rather than sub-type variation. This is, thus, a quantitative study, but we have coded it as applying the concept of premier-presidentialism qualitatively, because the sub-type simply provides the context in which the study is carried out. In total, we identified 87 studies that operationalised either premier-presidentialism or president-parliamentarism or both as an explanatory variable, of which 71 were studies where it was used qualitatively and 16 quantitatively. There has been an increase in both over time, but notably in quantitative studies. In the early period, there were 31 qualitative and only 4 quantitative studies. In the contemporary period, there were 40 qualitative and 12 quantitative studies. This trend perhaps reflects an increased emphasis on quantitative studies in the discipline more generally.

Whether the concepts were operationalised quantitatively or qualitatively, we then identified the main subject of the empirical inquiry. To this end, we distinguished between studies of five general types of outcome variables: democratisation, including the quality of democracy; domestic policy, including accountability; foreign and defence policy; parties and party systems; and a government category that included formation, survival and presidential/prime ministerial conflict, incorporating cohabitation. Note that more than one outcome can be the subject of an individual study. So, the number of studies is greater than the number of times the sub-types are employed as an explanatory variable. As noted above, we identified 96 empirical studies in total. The distribution of each type of study is recorded in Figure 3.

OPEN IN VIEWER

Figure 3.

The Outcome Variable in Empirical Studies of Premier-Presidentialism and President-Parliamentarism.

We can see that studies of democratisation continue to be the main focus of inquiry. Indeed, whereas we might have thought that the early period would have seen a greater emphasis on such studies as newly independent and newly democratic states adopted institutional rules, we find that studies of democratisation have increased considerably in the contemporary period. This perhaps suggests that scholars now feel in a position to take stock of the institutional determinants of democracy more dispassionately. We can also see, though, that studies of the effect of the two sub-types of semi-presidentialism on different aspects of government performance remain very popular. In some respects, the boundaries between these two categories of study are difficult to police. Often, scholars argue qualitatively that institutional conflict within the executive or government is caused by the sub-types of semi-presidentialism and that in turn such conflict causes problems for democratisation. For this reason, some of the studies dealing primarily with executive relations have implications for the study of democratisation too (Beuman, 2016).

Many of these topics are the subject of articles in this special issue. Therefore, we do not discuss the substantive findings of these studies. However, it is worth reporting that of the 16 quantitative studies, 9 returned good or strong support for a significant association between premier-presidentialism and president-parliamentarism and their outcome of interest. Four of these studies were concerned with the executive, including cohabitation, executive stability and president/cabinet conflict (Elgie and McMenamin, 2011; Fernandes and Magalhães, 2016; Kim, 2015; Sedelius and Ekman, 2010); three with democratisation (Elgie, 2011; Elgie and Schleiter, 2011; Sedelius and Linde, 2018); and two with parties and elections (Samuels and Shugart, 2010, 2014).

We have already noted an increase in the number of quantitative studies over time, but it is worth also noting that quantitative studies returning significant associations comprise some of the most recent studies of this sort. This suggests that there is good reason to engage in more quantitative studies of this sort. In this regard, Samuels and Shugart’s (2010) work is a good example. They argue that the variation in constitutional regimes shapes party organisation, which in turn affects presidential-legislative relations. For example, using quantitative methods, they show that the extent to which presidents follow through on their election promises varies as a function of whether there is a presidential, president-parliamentary, premier-presidential or parliamentary regime.

There is still a gap in our knowledge for all that. Samuels and Shugart’s privilege cross-national variation in outcomes dependent on the separation of powers, but what about within-country variation? Here, Passarelli (2015) provides a way forward. He suggests that the effect of the variation in the separation of powers is mediated by the genetic origins of political parties. This provides us with the potential to test whether the interaction of regime type and a party’s genetic features explains within-country variation too. This is an example of an area where future quantitative studies could usefully be applied, even if, as we note below, there are potential limitations to such studies.

Overall, we have found that the frequency of references to premier-presidentialism and president-parliamentarism has been increasing over the years; that these references have become more substantive; that substantive references have tended to take the form of country classifications and empirical studies of the effects of the two sub-types of semi-presidentialism; that there has been a notable decrease in the misclassification of country cases over time; that studies of the impact of the two sub-types of semi-presidentialism on democratisation and executive politics remain the main focus of empirical studies; and that the number of quantitative empirical studies has been increasing with some significant associations between the two sub-types and different outcome variables being identified in the recent period.

Limitations to the Future Study of the Two Concepts

We have no reason to believe that the increasing trend in references to Shugart and Carey’s typology is likely to be reversed any time soon. However, is it worth reflecting on the potential limitations to the study of the two sub-types of semi-presidentialism? How are references to the two concepts likely to evolve over time?

We found that only a very small number of explicitly critical references have been made of Shugart and Carey’s typology. Most of these were made in the early period in the context of single-country classifications. For example, in a study of Central and Eastern Europe, Frances Millard (2004: 4) refers to the presence of direct presidential elections in Poland, Lithuania, Romania, Bulgaria, Slovenia and Slovakia after 1999. She then states that ‘the first three are often regarded rather misleadingly as semi-presidential or premier-presidential’ (Millard, 2004). This reference is interesting because it is written at a time when, as discussed above, Shugart and Carey could be interpreted as implying that semi-presidentialism and premier-presidentialism were synonymous. The Millard reference indicates that this interpretation was taken on board by members of the scientific community at the time. This reference also shows that the scientific community was willing to think of country classifications on the basis of general constitutional features rather than the application of a strict constitutional rule at this time. After all, why does Millard state that only Poland, Lithuania and Romania are ‘misleadingly’ referred to as premier-presidential? The answer is because in the context of Shugart and Carey’s (1992) schema, Bulgaria, Slovenia and Slovakia could be interpreted, like Ireland, as parliamentary, given they all have presidents with very few powers. While misclassifications of this sort have almost disappeared in the contemporary period and while we do not know where Millard would now stand on the issue, some scholars do remain critical of the application of Shugart and Carey’s typology to particular countries. For example, writing in the contemporary period, Munkh-Erdene (2010) rejects the classification of Mongolia as an example of the premier-presidential sub-type of semi-presidentialism. Given those who do not subscribe to Shugart and Carey’s typology tend to ignore it, some scholars will simply choose to engage in a country-classification exercise without reference to the typology. This will somewhat limit the amount of future references to the concepts.

This basic point applies more broadly too. Since 2005 and certainly since 2010, Shugart has made it clear that premier-presidentialism and president-parliamentarism should be understood as two sub-types of semi-presidentialism. While the concept of semi-presidentialism has itself been the subject of increasing attention over time (Schleiter and Morgan Jones, 2009), it remains contested and not merely by certain country specialists in relation to the classification of their country of interest, but also by some comparativists. For example, José Cheibub rejects the idea of semi-presidentialism as a classificatory category, preferring instead to talk of mixed regimes. Indeed, Cheibub is one of the handful of comparativists who has specifically referred to Shugart and Carey’s typology critically, albeit only in a brief mention. He has written that ‘it is unclear . . . whether the distinction really matters’ (Cheibub, 2007: 38). In short, comparativists who reject the concept of semi-presidentialism are now more likely to reject the concepts of premier-presidentialism and president-parliamentarism too. This will also limit the amount of future references to the concepts.

More than that, since 2005/2010, Shugart and Carey’s typology has been firmly identified with a post-Duvergerian (Elgie, 2016) interpretation of semi-presidentialism, precisely because of the desire to avoid the ‘Duverger problem’. Yet, some scholars still wish to employ similar concepts in a Duvergerian-compatible way. For instance,Henry Hale (2015: 78) has rejected the use of ‘premier-presidentialism’ in favour of the term a ‘divided-executive constitution’. He defines this type of constitution as one that ‘formally enshrine[s] a balance between parliamentary and presidential power, assigning formally independent and roughly equal counterbalancing executive authority to each’ (Hale, 2015: 77 [emphasis in the original]). Here, the ‘Duverger problem’ is embedded in the definition of the concept of interest. Who is to say what is meant by ‘a balance’? Who determines what counts as ‘roughly equal’? Is there a roughly equal balance of this sort in Armenia? Some might say there is. Some might say there is not. We do not mean to be critical of Hale’s influential work of patronal politics. We are simply pointing out that some comparativists will wish to work with a Duvergerian-type definition of a certain concept rather than the post-Duvergerian definition of the sort that Shugart and Carey’s typology now corresponds to. This will limit the amount of future references to the typology.

There are also likely to be limits to the application of the typology even among the set of scholars who are willing to work with a post-Duvergerian understanding of semi-presidentialism. Without engaging in a similar meta-review of semi-presidentialism, we suspect that we would find a much greater number of references to semi-presidentialism in the period since 1992 than we found to premier-presidentialism and president-parliamentarism in the same period. This is at least partly because it is not necessary to refer to Shugart and Carey’s typology when working on the concept of semi-presidentialism itself. Put differently, we can think of the study of premier-presidentialism and president-parliamentarism as a sub-set of the study of semi-presidentialism generally. This does not necessarily make the former a niche academic product. After all, we have found an increasing number of references to the typology over time. All the same, it does restrict interest in the topic even among the set of semi-presidential scholars.

Furthermore, within this set of scholars, there are also likely to be limits to the number of purely quantitative studies that we are likely to find. We have shown that the number of such studies has been increasing, but the figure remains relatively small. This is at least partly because of the inherent limitations of the quantitative modelling that can be undertaken with Shugart and Carey’s typology. In practice, there are only two basic ways in which we can operationalise the difference between premier-presidentialism and president-parliamentarism quantitatively. If the research question focuses on the effects of the two sub-types relative to each other, then we can code them as a binary variable (Elgie, 2011; Elgie and Schleiter, 2011; Kim, 2015). If the research question focuses on the effects of the two sub-types relative to presidentialism, on the one hand, and parliamentarism, on the other, then we can code the set of regimes as a four-category ordinal variable (Samuels and Shugart, 2010; Sedelius and Linde, 2018). A number of quantitative studies have asked such questions, coded the sub-types accordingly, and, as we have seen, have returned significant results. Equally, as we have also seen, some scholars are willing to estimate the effect of a dichotomous institutional variable, but wish to include more than the set of semi-presidential countries in their sample. This means that they tend to fold president-parliamentary countries into the presidential category and place premier-presidential countries into the parliamentary category. This is a reductive strategy conceptually, but it increases the statistical n. In short, there are limits to the type of quantitative analysis that can be undertaken using Shugart and Carey’s typology, suggesting also that there are limits to the number of quantitative studies we are likely to come across in the future.

Finally, it is worth noting two developments that might pose a more profound challenge to the future application of quantitative studies using Shugart and Carey’s typology. The first concerns the study of presidential power. While some studies, notably Shugart and Carey’s (1992) study, identify the specific causal mechanisms associated with the two sub-types of semi-presidentialism, scholars often use constitutional regimes, including the two semi-presidential sub-types, as a simple proxy for variation in presidential power. Here, parliamentary systems have weak presidents, premier-presidential systems have somewhat stronger presidents, president-parliamentary presidents have stronger presidents still and presidential countries have the strongest presidents of all. If a study wishes to estimate the impact of presidential power on outcomes in this way, then Shugart and Carey’s typology can be operationalised to that end as a four-point ordinal scale as above. However, scholars are increasingly estimating the effect of more fine-grained variations in presidential power (Doyle and Elgie, 2016). These studies require presidential power to be operationalised as a continuous variable. Here, Shugart and Carey’s (1992) own presidential power scores can still be very useful, but their typology is not. If we are witnessing the beginning of a move from the study of the two sub-types of semi-presidentialism to the study of presidential power measured as a continuous variable, then we may see a concomitant decline in the number of studies that operationalise Shugart and Carey’s typology in the future.

The second challenge is more general. There has been an identificationist revolution among positivist political scientists in recent years. This revolution involves the rejection of conclusions based on observational and essentially correlational studies, including most of the statistical methods that have been employed in quantitative work over the last few decades. Instead, identificationists’ privilege conclusions are based on an experimental or at least a quasi-experimental research design. For identificationists, this is the only way to establish causal effects. For sure, there are statistical methods that mimic the experimental research design – Difference-in-Differences, Regression Discontinuity and so on. However, there are fundamental challenges to applying these statistical methods to the study of institutions. This is because institutions are chosen endogenously. Given an experimental design relies on the random assignment of the causal variable of interest (‘the treatment’ in the language of experiments), the endogenous or non-random selection of institutions means that the study of institutions generally, including the study of the effect of premier-presidential and president-parliamentary institutions, is increasingly under threat. This is a general challenge to the analysis of political institutions. There is no specific solution to this challenge. For sure, some researchers will continue to rely on observational and correlational studies. Other researchers will seek to establish the causal effect of institutions through qualitative methods, such as process-tracing. There is also the potential for the causal effect of institutions to be studied experimentally under laboratory or laboratory-like conditions. The chapter by Doyle and Elgie in Elgie (2018) is one such example, though it is confined to the study of presidentialism and parliamentarism rather than semi-presidentialism. Nonetheless, if political science moves even further in an identificationist direction, then we are likely to see fewer references to Shugart and Carey’s typology in the years to come.

Conclusion

Shugart and Carey’s (1992) volume, Presidents and Assemblies, was remarkable in many ways. Coming out only a few months after the fall of the Soviet Union and in the context of the nascent democratisation movement in Africa, Asia and Latin America generally, it was a remarkably advanced piece of scholarship that set the scientific agenda for years to come. Most notably, Shugart and Carey changed the terms of the scholarly debate, expanding the study of executive–legislative relations from a narrow focus on presidentialism and parliamentarism and to a lesser degree semi-presidentialism to a much broader perspective on variation in comparative political institutions generally, incorporating the executive and legislative powers of presidents and the role of party systems. In this article, we have concentrated on the introduction of the concepts of premier-presidentialism and president-parliamentarism. Although the terms hardly trip off the tongue, they captured an essential difference between two types of constitutional systems. Since then, the concepts have been slightly reformulated and some of the classifications that flow from them have been clarified. Nonetheless, they remain essentially the same as when they were first introduced to the world. The main advantage of the mature definition of the two concepts is that they allow the reliable classification of countries as examples of the two sub-types of semi-presidentialism. In turn, this allows the effects of institutional variation to be investigated more precisely. Based on a meta-analysis of references to Shugart and Carey’s typology, we have found that the frequency of references to the two concepts has increased over the years. Nonetheless, we have distinguished between an early and a contemporary period of references to the terms. In the contemporary period, there is much less misclassification of countries than in the early period. There are also many more empirical studies of the effects of the two sub-types of semi-presidential now than previously. For sure, we have indicated that there are limits to the extent to which scholars are likely to apply the typology. There are also limits to the way in which the concepts can be operationalised statistically. All the same, there is good reason to believe that the typology will continue to be a primary point of reference for scholars of semi-presidentialism and indeed for students of comparative political institutions generally.