Scholarly article on topic 'Presidentialism, Parliamentarism and Semi-Presidentialism: Bringing Parties Back In'

Presidentialism, Parliamentarism and Semi-Presidentialism: Bringing Parties Back In Academic research paper on "Political Science"

0
0
Share paper
Academic journal
Gov. & oppos.
OECD Field of science
Keywords
{""}

Academic research paper on topic "Presidentialism, Parliamentarism and Semi-Presidentialism: Bringing Parties Back In"

GOVERNMENT AND OPPOSITION

AN INTERNATIONAL JOURNAL OF COMPARATIVE POLITICS

Government and Opposition, Vol. 46, No. 3, pp. 392-409, 2011 doi:10.1111/j.1477-7053.2011.01345.x

Review Article

Robert Elgie: Presidentialism, Parliamentarism and Semi-Presidentialism: Bringing Parties Back In

D. J. Samuels and M. S. Shugart, Presidents, Parties, Prime Ministers: How the Separation of Powers Affects Party Organization and Behavior, Cambridge, Cambridge University Press, 2010, 310pp., hardback, ISBN 978-0-521-86954-6, £55.00; paperback, ISBN 978-0-521-68968-7, £16.99.

The resurgence of interest in the effects of different regime types began with Juan Linz's critique of presidentialism in the early 1990s.1 Coming at a time when many countries were democratizing and adopting new constitutions, and being firmly rooted in the 'new institutionalise academic framework, the work of Linz and others was highly influential. Two decades on, the study of regime types remains in vogue, with José Cheibub recently devoting a whole book to a test of Linz's central proposition.2 Over time, though, the terms of the debate have changed. The focus on presidentialism and parliamentarism has been complemented by an interest in semi-presidentialism.3 The study of institutional variation has been extended beyond its effects on the collapse or survival of new democracies to incorporate a much broader range of concerns, including general indicators of political and economic development4 as well as

1 J.J. Linz, 'The Perils of Presidentialism', Journal of Democracy, 1 (1990), pp. 51-69; J.J. Linz, 'Presidential or Parliamentary Democracy: Does It Make a Difference?', in J.J. Linz and A. Valenzuela (eds), The Failure of Presidential Democracy: Comparative Perspectives, Baltimore, Johns Hopkins University Press, 1994, pp. 3-87.

2 J. A. Cheibub, Presidentialism, Parliamentarism, and Democracy, Cambridge, Cambridge University Press, 2007.

3 R. Elgie, 'Semi-Presidentialism: Concepts, Consequences and Contesting Explanations', Political Studies Review, 2 (2004), pp. 314-30.

4 J. Gerring, S. C. Thacker and C. Moreno, 'Are Parliamentary Systems Better?', Comparative Political Studies, 42 (2009), pp. 327-59.

© The Author 2011. Government and Opposition © 2011 Government and Opposition Ltd Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

the initiation of conflict.5 In addition, the use of different analytical methods, such as the veto-players approach6 and principal-agent analysis,7 has produced a new generation of regime-type studies.8

This review article focuses on the recent book by David Samuels and Matthew Shugart.9 The authors state that the 'intellectual impact' of Juan Linz is felt throughout the book,10 and they make it clear that their approach is rooted in a 'theory of how institutions shape politicians' behavior'.11 Thus, they place themselves resolutely within the institutionalist tradition, even if that tradition can no longer be labelled 'new', and they focus on the effects of variation in the separation of powers, comparing the impact of presidentialism, parliamentarism and two forms of semi-presidentialism - president-parliamentarism and premier-presidentialism. At the same time, though, apart from a brief discussion towards the end of the volume, they are not concerned with the impact of institutions on the collapse or survival of new democracies. Instead, the 'central question' of the book is: to what extent does 'the presence of constitutionally separate executive authority "presidentialize" political parties?'12 Using the language of the principal-agent approach, their answer is 'considerably' and various ways in which party organization and behaviour are shaped by regime type are identified over the course of the volume. Thus, unlike Cheibub's recent volume, Samuels and Shugart's work is part of what has been identified as the 'third wave' of regime-type studies.13

The Samuels and Shugart book is an important contribution to comparative politics and contemporary institutional analysis,

5 D. Reiter and E. R. Tilman, 'Public, Legislative, and Executive Constraints on the Democratic Initiation of Conflict', Journal of Politics, 64 (2002), pp. 810-26.

6 G. Tsebelis, Veto Players: How Political Institutions Work,, Princeton, Princeton University Press, 2002.

7 K. Str0m, W. C. Müller and T. Bergman (eds), Delegation and Accountability in Parliamentary Democracies, Oxford, Oxford University Press, 2004.

8 R. Elgie, 'From Linz to Tsebelis: Three Waves of Presidential/Parliamentary Studies?', Democratization, 12 (2005), pp. 106-22.

9 D. J. Samuels and M. S. Shugart, Presidents, Parties, Prime Ministers: How the Separation of Powers Affects Party Organization and Behavior, Cambridge, Cambridge University Press, 2010.

10 Ibid., p. xi.

11 Ibid., p. 22.

12 Ibid., p. 14.

13 R. Elgie, 'From Linz to Tsebelis'.

bringing together two previously separate literatures: one on the effect of regime types and the other on the determinants of party organization and behaviour. There are three parts to this article. The first part outlines the book's argument and the authors' main findings. The second identifies a small number of minor issues. The final part extends the implications of the book, demonstrating the agenda-setting potential of the theoretical framework.

THE SEPARATION OF POWERS AND PARTY ORGANIZATION AND BEHAVIOUR

The focus of the volume is very clear: the authors wish to explain variation in the organization and behaviour of political parties. 'At the core of this book', they state, 'is the question of how political parties organize and when they must bridge the gap between the executive and the legislature.'14 Samuels and Shugart position the book in a line of scholarship that dates back to the work of Maurice Duverger in the 1950s.15 That said, while the authors cite the main contributions to the study of party organization since this time, including Kirchheimer and more recently Katz and Mair, they do not provide a critique of this literature. They acknowledge their debt to the work of Leon Epstein, but the authors merely assert, quite rightly, that much of the literature on party organization has simply failed to take account of the impact of regime type.16 It is the 'missing variable'.17 In this regard, they make a very strong claim. They state: 'We do not claim that our argument supersedes or supplants the importance of social, economic, or cultural forces that drive party formation, evolution, and behavior . . . but analysis of these differences should begin with the difference in constitutional design.'18 Thus, Samuels and Shugart are interested in explaining variation in party organization and they believe that the design of executivelegislative relations is primary in accounting for such variation.

14 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 250.

15 Maurice Duverger, Political Parties: Their Organization and Activity in the Modern State, London, Methuen, 1954.

16 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 7.

17 Ibid.

18 Ibid., p. 18.

This belief is based on an equally strong theoretical claim. They state: 'to the extent that the constitutional structure separates executive and legislative origin and/or survival, parties will tend to be presidentialized. '19 Separate origin refers to the situation where there is a direct presidential election that is held separately from legislative elections, albeit perhaps simultaneously. Separate survival 'means that a party or legislative majority cannot remove a sitting president'.20 A presi-dentialized party has a leader with considerable independence in the electoral and governing arenas.21 In other words, the leader will stand for election on a personal platform rather than the party platform and, once elected, will choose cabinet members and propose policy reforms with minimal party intervention. By contrast, a parliamenta-rized party fuses the electoral and governing arenas. The party selects a leader who sticks to the party platform at the election, and the party can hold the leader accountable for their actions following the election. Samuels and Shugart make it clear that parties in presidential regimes can exhibit 'parliamentarized' characteristics and vice versa.22 However, they assert that such characteristics are likely to be 'ephemeral' because of the 'inescapable logic' of the regime's institutional foundations.23

The causal mechanism underpinning this theoretical claim is highly intuitive. Under presidentialism where executive and legislative origins are separate, parties face competing incentives. What it takes for the president to be elected is not necessarily what it takes for individual party candidates of the legislature to be elected or for the party as a whole to gain a majority there. Presidential candidates are likely to adopt a vote-seeking strategy that emphasizes public goods because they need to win a large proportion of the national electorate.24 By contrast, the party's legislative candidates may adopt a policy-seeking strategy that emphasizes constituency goods. These competing incentives create a dilemma for parties under presiden-tialism that is absent under parliamentarism. This dilemma is internal to each party in the system. A similar dilemma occurs under presi-dentialism where the survival of the executive and the legislature is

19 Ibid., p. 37, emphasis in the original.

20 Ibid., p. 15.

21 Ibid., p. 16.

22 Ibid., p. 17.

23 Ibid., p. 18.

24 Ibid., p. 50.

also separate. Here, presidents 'have little to fear from their own colleagues'.25 Thus, Samuels and Shugart state: 'to the extent that capture of a separately elected presidency is important for control over the distribution of the spoils of office and/or the policy process, party behavior and organization will tend to mimic constitutional structure, giving rise to "presi-dentialized"parties.'26 By contrast, under parliamentarism parties may dismiss their own leaders, meaning that leaders potentially have much to fear from their own colleagues, but the system ensures that the equivalent intra-party dilemma is essentially absent.

The theoretical structure and the causal mechanism are framed in the language of principal-agent analysis. In this analysis, the principal is the party and the agent is the party's leader. For Samuels and Shugart, the separate origin of the executive and the legislature means that systems 'with direct presidential elections exacerbate the problem of adverse selection'.27 In other words, the presidential candidate who is likely to be the most faithful agent of the party may not necessarily be the candidate who is best placed to win the election, thus generating the intra-party dilemma. Equally, the separate survival of the executive and the legislature creates a problem of moral hazard. They state: 'For political parties-as-principals, the danger is that leaders-as-agents might use their authority to advance their own personal goals rather than work toward their party's collective goals.'28 Thus, while there is always the potential for conflict between principals and agents, the likelihood of intra-party conflict is increased under certain types of executive-legislative relations.

A real strength of Samuels and Shugart's approach is that it allows the comparison of presidential, parliamentary and semi-presidential regimes. To date, most attention has focused on the first two systems. By emphasizing the dynamics caused by separate origin and separate survival, semi-presidential countries can be integrated into the analysis as well. Consistent with the current standard wisdom, semi-presidentialism is defined as the situation where there is a directly elected president and a prime minister and cabinet that can be dismissed by the legislature. This is a post-Duvergerian definition of semi-presidentialism, meaning that a country's president does not

25 Ibid., p. 15.

26 Ibid., pp. 15-16, emphasis in the original.

27 Ibid., p. 47.

28 Ibid., p. 48.

have to have a certain amount of power - however such power might be measured - in order for that country to be classed as semi-presidential. The definition is purely constitutional. In addition, Samuels and Shugart make the now equally standard distinction between premier-presidential and president-parliamentary forms of semi-presidentialism. In the former, the prime minister and cabinet can be dismissed solely by the legislature. In the latter, they can be dismissed by the president as well.

Samuels and Shugart assert that the problems of adverse selection and moral hazard are likely to be greater under presidentialism and both types of semi-presidentialism than under parliamentarism. They also assert, however, that 'party presidentialization is greater in president-parliamentary systems than in premier-presidential systems'.29 The direct election of the president, they say, 'tends to "contaminate" the parties, interfering in the principal-agent relationship between parties and their prime-ministerial agents in the legislature'.30 All the same, premier-presidentialism has the potential advantage that it can parliamentarize parties during periods of cohabitation - defined, again in the now standard way, as the situation where the president's party is not represented in the cabinet. Thus, Samuels and Shugart expect there to be different outcomes under president-parliamentary and premier-presidential forms of semi-presidentialism.

Having established their theoretical framework and the expectations that flow from it, the authors proceed to a series of empirical tests. They apply their arguments to democratic countries. To identify the set of democracies, they rely on the Polity IV scale, including in their data set all countries that have had a polity2 score of +5 or more for at least five consecutive years in the period 1945-2007. Using this sample, they engage in a good blend of quantitative analysis and qualitative case studies. Interestingly, the quantitative analysis usually takes the form of simple inferential statistics. This is not a book that relies on results from the latest econometric model. There is also a nice mix of qualitative analysis. Throughout the book, there are indicative one-page vignettes illustrating particular points. There are also two in-depth case studies of parties' 'presidential dilemmas' in Brazil and Mexico. There is a chapter that illustrates

29 Ibid., p. 108.

30 Ibid., p. 256.

the effect of constitutional change on intra-party development. This chapter takes the form of two in-depth studies of France and Israel, both of which countries changed their organization of executive-legislative relations and subsequently experienced different intra-party behaviour. Thus, this chapter describes the effect of a natural-like experiment.

To demonstrate the empirical validity of their argument, Samuels and Shugart show that political outsiders are least likely to hold office under parliamentary systems and most likely to do so under presiden-tialism. They also show that outsiders are more likely to be successful under presidentialism than president-parliamentarism and, in turn, more likely under president-parliamentarism than premier-presidentialism. In another test, they show that in pure parliamentary systems, 'about three in ten changes in prime minister result from purely intraparty politics'.31 A similar finding occurs under semi-presidentialism, but here presidents also have influence over prime ministerial appointments and dismissal, again indicating the presidential 'contamination' of intra-party relations under this system. By contrast, under presidentialism parties have almost no direct influence over the removal of presidents. In addition, they establish that there is an electoral separation of purpose - meaning the extent to which 'parties' candidates for executive and legislative office derive support from and respond to different sets of voters'32 - under both presidentialism and semi-presidentialism; by definition, this cannot be the case under parliamentarism. Finally, they show that regime type interacted with party system variables affect the likelihood of policy switching, with greater switching, or more violations of mandate representation, ' as we move away from the ideal-typical parliamentary chain of delegation' .33 Along with subsidiary findings - for example, that the incidence of cohabitation is greater under premier-presidentialism than president-parliamentarism, again because parties under the latter are more presidentialized - the authors provide compelling empirical evidence to back up their theoretical claims.

Generally, the organization of the book and the execution of the argument could serve as a blueprint for the work of any early-career academic. Samuels and Shugart have provided a lesson in how to

31 Ibid., p. 120, emphasis in the original.

32 Ibid., p. 123.

33 Ibid., p. 221, emphasis in the original.

present a piece of research as well as making a genuinely innovative contribution to comparative political analysis and, in particular, to the study of the effects of executive-legislative relations on party development and behaviour.

QUIBBLES, QUESTIONS AND QUERIES

Any large-scale research exercise raises various issues of research design. Authors have to make choices and those choices can always be contested. As long as the choices do not make a material difference to the findings of the project, then the issues are minor and do not threaten the validity of the overall results. This section identifies a number of such issues in Samuels and Shugart's book, but they are all of the 'relatively minor' type. None of them challenges the fundamental nature of the project.

As with any large-n study, the issue of case selection can be questioned. As noted previously, the authors rely on the Polity IV data set to identify the universe of democracies from 1945 to 2007. Recently, the methodology underpinning the Polity IV project has been chal-lenged.34 Some of these problems manifest themselves when Polity is used as a continuous variable, which is not the case in the Samuels and Shugart volume. However, it might be noted that the authors neither justify the choice of Polity, nor do they attempt to use any other way of identifying democracy as a check of the robustness of their results. Recently, the strategy of confirming any results by using more than one measure of democracy has been strongly recommended by Bayer and Bernhard.35 In addition, Samuels and Shugart use the +5 polity2 score as their threshold for democracy. However, even though Bogaards has shown that there is no consensus on where the threshold for democracy should be drawn using Polity,36 a score

34 For example, S. Treier and S. Jackman, 'Democracy as a Latent Variable', American Journal of Political Science, 52: 1 (2008), pp. 201-17; and J. R. Vreeland, 'The Effect of Political Regime on Civil War: Unpacking Anocracy', Journal of Conflict Resolution, 52: 3 (2008), pp. 401-25.

35 R. Bayer and M. Bernhard, 'The Operationalization of Democracy and the Strength of the Democratic Peace: A Test of the Relative Utility of Scalar and Dichoto-mous Measures', Conflict Management and Peace Science, 27: 1 (2010), pp. 85-101.

36 M. Bogaards, 'Measures of Democratization: From Degree to Type to War', Political Research Quarterly, 63: 2 (2010), pp. 475-88.

of +6 is perhaps the commonest threshold, not least because this is the one that is used by the authors of the Polity project themselves. Moreover, even though Samuels and Shugart have clear, if contestable, criteria for identifying democracies, they seem to break their own rules by ignoring short authoritarian intervals. This point seems to apply to Malawi and Zambia, both of which had a period of >+5 democracy followed by two and five years respectively of authoritarianism and then a further period of >+5 democracy. Rather than identifying separate periods of democracy in each country, Samuels and Shugart identify only one longer period. Again, the results are unlikely to be affected by such a decision, but it is a somewhat strange choice.37 Generally, the authors identify a set of countries that look like democracies and quack like democracies. So, they probably are democracies. However, the choice of Polity and the +5 threshold need to be justified. What is more, the authors might be a little more careful when subsequently they generalize from this set of democracies. For example, they state that Austria is 'the only obviously "par-liamentarized" president-parliamentary regime in the world today'.38 Well, Iceland is certainly a candidate for inclusion in such a list, but it is excluded from the Polity data set because it has a population of fewer than 500,000 people. So, the general statement is perhaps a little misleading. Equally, Samuels and Shugart state, 'no premier-presidential democracy has ever been replaced by an authoritarian regime.'39 They do acknowledge that this statement is driven by their case selection, but even so it is perhaps too bold a claim as citizens of Congo-Brazzaville in 1992, Niger in 1996, and Haiti in 1999 might wish to testify.40

Another aspect of research design in a study of this sort with which it is always possible to quibble is the classification of certain countries as examples of particular regime types. In this regard, the classification of Madagascar is genuinely puzzling. Here, it is classed as premier-presidential in the period 1991-93, president-parliamentary in 1993-97 and presidential in 1997-2007. However,

37 In addition, Samuels and Shugart include Papua New Guinea from 1975 to 2007, even though Polity has now reclassified this whole period as +4.

38 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 88.

39 Ibid., p. 260.

40 The collapse of democracy in Finland in 1931 is another example, but it lies outside the timeframe of Samuels and Shugart's study.

in October 1991 an interim constitution was passed that provided little detail about the nature of executive-legislative relations.41 The first democratic constitution was promulgated in August 1992 and was premier-presidential.42 In October 1995 the constitution was changed and Madagascar became president-parliamentary.43 The text was then amended again in April 1998, but no material change was made in terms of its provisions for censuring the government.44 Thus, the classification of Madagascar is serially contestable. A different issue altogether is raised when it comes to the classification of Austria and Ireland. Constitutionally, both countries are semi-presidential, but politically both operate as de facto parliamentary regimes. The authors are well aware of the exceptional nature of both countries,45 but they appear to make different operational decisions in each case. The Austrian case is discussed in some detail46 and the decision is taken to exclude Austria from the list of president-parliamentary cases for the purposes of the quantitative analysis. This is a perfectly reasonable decision and a clear justification is provided for it. However, Ireland seems to be systematically included in the list of premier-presidential cases for such purposes. In isolation, each decision is perfectly justifiable, but given that both countries are exceptional and for essentially the same reason it seems a little strange not to treat them in the same way. Since there are a large number of Ireland years in the data set,47 it would be interesting to report explicitly the difference it would make if Ireland were excluded from the list of premier-

41 This document is available in French at http://mjp.univ-perp.fr/constit/ mg1991.htm, accessed 17january 2011.

42 The text of the 1992 constitution is available in English at http://www. chr.up.ac.za/undp/domestic/docs/c_Madagascar.pdf, accessed 17 January 2011.

43 The changes in the wording of Articles 53 and 90 are key in this regard. The text is available in J. du Bois de Gaudusson, G. Conac and C. Desouches (eds), Les Constitutions africaines publiées en langue française. Tome 1, Paris, La Documentation française, 1998, pp. 390-408.

44 After the 1998 changes it became easier for the legislature to dismiss the government if the latter made a bill a matter of confidence. The text of the 1998 constitution is available in French at http://mjp.univ-perp.fr/constit/mg1998.htm, accessed 17 January 2011.

45 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 257.

46 Ibid., pp. 88-90.

47 On a separate issue, Samuels and Shugart include Ireland in their data set from 1952, but using their Polity criteria it would seem to require inclusion from 1945.

presidential cases, or if both Austria and Ireland were operational-ized as parliamentary regimes.

A great strength of the book is the way it undertakes a comprehensive set of empirical tests of the central theoretical proposition. That said, one or two of the tests are less fulfilling than they might at first appear and/or than they might otherwise have been. This is particularly the case with the chapter on the electoral separation of purpose. As defined above, this concept 'measures the degree to which the electoral process generates misalignment of the political incentives between a party's executive candidate and its median legislative candidate'.48 The expectation is that the intra-party dilemmas caused by separate origin and survival of the executive will generate the situation where this misalignment is relatively large. By contrast, there can be no electoral separation of purpose under parliamentarism, 'because there is no way for a citizen to vote for a party's prime ministerial candidate without also endorsing the party's legislative candidate or slate'.49 This is true. The reason, though, why this test is less fulfilling than the others in the book is because, as the authors state, by definition there can be no electoral separation of purpose under parliamentarism. Thus, Samuels and Shugart have created a test that can only be falsified if they were to find no evidence of electoral separation of purpose under presidentialism and semi-presidentialism. They do find such evidence; therefore, the proposition is true. This is a perfectly reasonable research strategy and the test for evidence of electoral separation of purpose under these two regimes is very sophisticated. All the same, it might have been more fulfilling to have devised a test that could, by definition, have generated some variation within the set of parliamentary countries and that, therefore, would have allowed a comparison of the full set of countries in the data set. The concept of the electoral separation of purpose cannot be operationalized in this way. Perhaps, though, another test might have been devised. For example, in the UK during the Blair premiership there were concerns within the ruling Labour Party that the prime minister was leading the party away from some of its core values. This was being done, Blair's supporters claimed, in order to maximize the party's chances of being re-elected. As a result, though, on certain issues there were backbench rebellions against the

48 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 123.

49 Ibid., pp. 123-4.

prime minister's policies.50 Such rebellions are not a manifestation of an electoral separation of purpose, but they may be an example of intra-party tensions caused by the executive candidate (the prime minister) being concerned with vote maximization and the parliamentary party (or elements of it) being concerned with policy maximization. This is very close to the dilemma that Samuels and Shugart are trying to capture in the electoral separation of purpose. If cross-national levels of governing party cohesion had been measured, then parliamentary countries could have been included fully in the test.

The same sense of a certain lack of fulfilment applies to the situation where Samuels and Shugart are unable to test for the effect of variation within semi-presidentialism. This occurs in the chapter on the electoral separation of purpose and also in the chapter on policy switching. The reason for all semi-presidential countries being lumped together in these chapters is that the country numbers are too small for the two types of semi-presidentialism to be operation-alized separately. Again, in both chapters the empirical tests are carefully designed and implemented. Moreover, the results are entirely consistent with the expectations of the theoretical claims. So, there is no fundamental problem in this regard with the findings in either of the chapters. That said, there is always a nagging concern about any results when semi-presidential countries are clustered together. Semi-presidentialism in Ireland is very different from semi-presidentialism in Peru. What is more, the low number of semi-presidential countries is, of course, a function of the case selection procedure. It would have been possible to increase the semi-presidential n and perhaps allow for the two types of semi-presidentialism to be tested separately had a different case selection procedure been applied. This could have been achieved by lowering the Polity threshold for democracy or by using the Freedom House and/or the Polyarchy data sets as either the main data source for the book or as an alternative data source for the hypotheses in these chapters and perhaps also as a robustness check for the hypotheses in other chapters. In fairness to Samuels and Shugart, though, life is short and data collection is potentially never-ending. The decision to use Polity IV as the data source and +5 as the democracy threshold does need to be justified, but both are justifiable choices. Whether or

50 In this regard, see the publications of Philip Cowley at: http://www. nottingham.ac.uk/politics/staff/philip.cowley, accessed 18 January 2011.

not the results change on the basis of a different data source and/or whether there is evidence of the effects of variation within semi-presidentialism as a function of using such a source are projects for another day, or for other researchers.

Finally, the book generates a disparate set of minor queries and comments. For example, how do the authors operationalize non-partisanship? In presidential and president-parliamentary regimes especially, presidents have been known to stand as independent candidates. Sometimes such non-partisanship is merely a front for a candidate who is clearly partisan. At other times the candidate may be genuinely non-partisan at the election but afterwards may receive the support of a party or bloc of parties in the legislature, thus, in effect, becoming partisan. It is not entirely clear how Samuels and Shugart treated any cases of non-partisanship and whether this issue has the potential to make any difference to their results. In addition, the very brief discussion towards the end of the book about the relationship between regime type and democratic survival is perhaps a little out of place, given the focus of the volume as a whole.51 Moreover, the failure to mention any control variables in this discussion means that the comments are even more speculative than they are presented.

Lastly, the country expert in me is compelled to point out one mistake and one highly contestable statement in the case study of France. Samuels and Shugart state that Mitterrand joined the Socialist Party (PS) 'after he had announced his presidential candidacy in 1974'.52 They cite a chapter by Ben Clift when making this point, but they misinterpret his text - as Clift is very well aware, Mitterrand joined the PS in 1971. Also, they state that 'by 2002 nearly all parties had become presidentialized' in that each 'saw advantage to presenting a presidential candidate'.53 At the 1981 presidential election all of the electorally competitive parties stood candidates and Jean-Marie Le Pen, whose National Front party was not competitive at that time, also tried to stand but failed to gain the requisite number of signatures. So, 2002 is a very late date for the presidentialization of parties in France, even on the basis of the authors' specific criterion in this regard.

51 Samuels and Shugart, Presidents, Parties, Prime Ministers, p. 260.

52 Ibid., p. 173, emphasis in the original.

53 Ibid., p. 177.

Overall, these points are not meant to cast doubt on the overall research design of the project or the empirical findings. It is rare to find authors being so transparent about their choice of research design, their definitions of key concepts and their classifications of particular countries. Too often, issues of such fundamental theoretical and empirical importance are buried away or taken for granted. This is not the case here. As a result, though, Samuels and Shugart actually make it easier for the reviewer to quibble with certain choices. I hope that such behaviour by the reviewer can be forgiven.

SETTING A RESEARCH AGENDA

As befits a major research volume, Samuels and Shugart end their study by identifying various ways in which the research agenda could be taken forward.54 For example, they urge variation in the separation of powers to be placed more centrally in explanations of party-system emergence and consolidation. They ask whether such variation may affect parties' organizational evolution and whether this might explain why there are so few shifts from presi-dentialism to parliamentarism. They raise the issue of how the separation of powers affects election campaigns, coalitional strategies, campaign organization and the allocation of resources. They also encourage scholars to explore presidential influence in representation more fully, notably whether voters hold presidents personally accountable.

In the rest of this review article, I would like to provide a brief demonstration of how Samuels and Shugart's work has the potential to set the agenda in other areas too. To do so, I will focus on one of their findings about cohabitation under semi-presidentialism. Their theoretical framework leads them to hypothesize that cohabitation is likely to be more frequent under premier-presidentialism than under president-parliamentarism. Indeed, they state that they 'expect cohabitation to almost never occur under president-parliamentarism'.55 The empirical evidence confirms this expectation. They found only one case of

54 Ibid., pp. 262-4.

55 Ibid., p. 45.

cohabitation under president-parliamentarism, namely in Sri Lanka from 2001 to 2004.56 In my work on semi-presidentialism I have frequently been struck by the commonplace argument that cohabitation is dangerous for the survival of new democracies - because it creates conflict within the executive between the president and prime minister - and yet there is a lack of evidence to support such an argument; only Niger has collapsed during a period of cohabitation.57 Even though Samuels and Shugart do not address the issue of democratic survival systematically, their theoretical framework helps to explain why there is so little evidence to support the argument about the perils of cohabitation.

To demonstrate this point, I also rely on the Polity IV data set. I identify all countries with semi-presidential constitutions since the first cases in Finland and Weimar Germany in 1919 through to 2008 inclusive. I consider a country to be democratic when it has a polity2 score of >+1. Thus, a country's democracy collapses if it first receives a score of >+1 and then at some later stage it receives a score of <0. Using the same definition of semi-presidentialism as Samuels and Shugart and the same definition of cohabitation, I find that there were 740 country years of semi-presidential democracy in the period 1919-2008 inclusive, during which time there were 100 years of cohabitation. However, I also find that 87 of the 100 cohabitation observations occurred in countries with a Polity IV score of +9 or +10. Bearing in mind that the highest score recorded by a semi-presidential democracy in the year prior to its collapse was +8 (which occurred both in Austria in 1932 before it collapsed in 1933, and in Niger in 1995 before it collapsed in 1996), what this shows is that cohabitation has overwhelmingly occurred in countries that are, in effect, consolidated democracies, where democracy is the only game in town and where it is extremely unlikely that a collapse will ever occur.

Thus, critics of semi-presidentialism who believe that cohabitation is dangerous for the survival of democracy have made this argument on the basis of a faulty assumption. They have assumed that the

56 Ibid., pp. 45-6. Again, their case selection means that they exclude periods of president-parliamentary cohabitation during the Weimar Republic as well as more recently in Sâo Tomé and Príncipe. In addition, there have been periods of cohabitation in de facto parliamentary Austria and Iceland.

57 R. Elgie, 'Semi-Presidentialism, Cohabitation and the Collapse of Electoral Democracies 1990-2008', Government and Opposition, 45: 1 (2010), pp. 29-49.

incidence of cohabitation will be distributed evenly across the range of semi-presidential democracies and that when it occurs in young or fragile democracies the political system may not be able to survive their impact. In fact, though, the figures demonstrate that the distribution of this potentially problematic situation is heavily skewed towards the set of democratically consolidated countries. Therefore, cohabitation is unlikely to be generally problematic for the survival of young democracies. Why, though, is the distribution of cohabitation and a divided executive skewed towards consolidated democracies? Samuels and Shugart's framework provides the answer. Using the Polity >+1 threshold for democracy, I find that only 13 of the 100 cohabitation observations occurred in countries with a president-parliamentary form of semi-presidentialism. What is more, 8 of these 13 observations occurred in Austria, which, everyone agrees, is a highly anomalous case. Thus, Samuels and Shugart's empirical intuition is confirmed even when a different threshold of Polity IV democracy is used and on the basis of a longer time period. Cohabitation is much more likely to occur under premier-presidentialism than under president-parliamentarism.

Together, the findings that cohabitation overwhelmingly occurs in consolidated democracies and in premier-presidential democracies are suggestive. They indicate that the critics of semi-presidentialism are undoubtedly correct to point out that countries with this type of constitution are likely to experience cohabitation. However, the critics miss the fact that cohabitation is least likely to occur in the countries that are most likely to collapse. This does not mean that semi-presidentialism is a good choice for young democracies. It simply means that when semi-presidential democracies collapse they are more likely to do so for reasons other than the standard problems of cohabitation. Whether a country has a president-parliamentary constitution or a premier-presidential constitution is likely to be one such reason. As Samuels and Shugart imply, under president-parliamentarism with presidentialized parties presidents have an incentive to govern against the legislature, including, if need be, representatives of their own party. Rather than cohabit, they are likely to try to form minority governments and/or to rule by decree. This is the real threat to young democracies. By contrast, under premier-presidentialism the president can govern only through the legislature and the government that it approves. This may mean that presidents have to experience cohabitation from time to time. Generally,

though, it also means that they are more likely to try to reach wideranging political deals with parties in the legislature so as to maximize their influence over the legislative process. The inclusiveness of premier-presidentialism relative to president-parliamentarism is the reason why countries with this form of semi-presidentialism are less likely to collapse than those with a president-parliamentary form. A controlled statistical test of this argument carried out with Petra Schleiter confirms this finding and also that cohabitation is a poor predictor of why some semi-presidential democracies collapse and others survive.58 It also shows that the intuitions at the heart of the Samuels and Shugart volume can be applied very widely.

CONCLUSION

Samuels and Shugart have rebooted a particular research agenda in comparative politics. In fact, they have done so in perhaps two regards. Most clearly, they have demonstrated that party behaviour and organization are in large part a function of variation in the separation of powers. This finding provides a direct challenge to those who argue that party organization is primarily the result of sociological cleavages or state organization. Instead, Samuels and Shugart show that regime type is the 'missing variable' when explaining variation in party organization. In addition, though, their work also suggests that party organization might be the 'missing variable' in studies of the effect of regime types. They argue that party organization mimics constitutional structure - presidentialized parties are the result of presidentialized constitutional systems. If this is the case, then parties also need to be integrated into studies of the effects of political regimes. Such studies need to pay attention to the direct effects of regime types on political outcomes, but Samuels and Shugart's work indicates that such studies also need to pay attention to the concomitant effects of party organization on such outcomes. Indeed, arguably, this aspect of their study is the one with the greater potential to shape the broader research agenda. The study of why political parties have different organizational features is long-

58 R. Elgie and P. Schleiter, 'Variation in the Durability of Semi-Presidential Democracies', in R. Elgie, S. Moestrup and Y-S. Wu (eds), Semi-Presidentialism and Democracy, London, Palgrave, 2011, pp. 42-60.

standing and ongoing, but it is relatively narrow. To put the point another way, there is a debate over whether there has been a shift to cartel parties. That debate, for those who follow it, is vibrant, but it has had little generalizable impact on the comparative politics research agenda. Samuels and Shugart have added to that debate. By contrast, the study of the effect of regime types has an extremely broad application in comparative politics, international political economy, political history and so on. If Samuels and Shugart have successfully demonstrated that the study of regime types needs to incorporate the study of political parties and their organizational features as well, then the impact of this book is likely to be even greater still.