A new political system model: Semi-parliamentary government

Introduction

In a seminal contribution to this journal – from which the main title is borrowed here – Maurice Duverger (1980) defined the concept of ‘semi-presidential government’ and analysed the diversity of practices in what was then only a small set of semi-presidential countries (see also Elgie 2011, 2016). This article carries out a similar analysis for the concept of ‘semi-parliamentary government’. The proposal of this article is to understand the latter as the mirror image of the former. A semi-presidential system divides the executive into two equally legitimate parts, only one of which – the prime minister – depends on assembly confidence for its survival in office. Conversely, a semi-parliamentary system divides the assembly into two equally legitimate parts, only one of which possesses the power to dismiss the prime minister in a no-confidence vote.1 Such a system establishes a formal separation of power between the executive and one part of the assembly. Semi-parliamentary government is a distinct, but neglected executive-legislative system and a ‘missing link’ in our typological thinking about democratic constitutions.

In this article, the constitutions of the Australian Commonwealth, the bicameral Australian states and Japan are identified as semi-parliamentary. Analysis of these cases will be connected with the comparative literature on executive-legislative systems and the Australian experience will be drawn upon to discuss new semi-parliamentary designs. These designs do not necessarily require formal bicameralism. For example, a simple way to implement semi-parliamentarism is a legal threshold of confidence authority: parties with a below-threshold vote share would be denied the right to participate in a no-confidence vote, but they would be allowed fair representation in the legislative and deliberative process in parliament. Semi-parliamentarism might also be an attractive option for democratising the European Union (EU).

Semi-parliamentary government deserves the attention of constitutional and democratic theorists because it may avoid important downsides of pure presidential and parliamentary systems, while maintaining many of their core strengths. In presidential systems, neither citizens nor representatives can remove an incompetent (rather than criminal or incapacitated) president from office – at least unless they are willing to stretch the constitutional rules (e.g., Marsteintredet et al. 2013).2 The concentration of executive power in a single individual also provides strong democratic reasons for constitutional limits on re-election, but these limits undermine the executive's electoral accountability and are often difficult to enforce (Ginsburg et al. 2013; Linz 1994: 12). Popularly elected chief executives weaken parties’ unity and programmatic capacities (Carey 2007; Samuels & Shugart 2010); and while they may not increase the overall risk of democratic breakdown (Cheibub 2007), they do seem to pose a persistent threat of authoritarian takeover by the incumbent president (Maeda 2010; Svolik 2015).

Semi-parliamentarism may offer potential solutions to these problems. It allows citizens to choose a prime minister in a way that mimics presidential elections, but the incumbent can be re-elected without limits and removed at any time by his or her party, or by the majority in one part of the assembly. Adequately designed, this part can function as a standing two-party ‘confidence college’ for the prime minister. Since the executive's survival is constitutionally independent from the other part of the assembly, there is still a form of branch-based separation of powers. Semi-parliamentarism can achieve power separation without presidents.

A core problem of pure parliamentarism is the unavoidable tension it creates between parliament's representative, deliberative and legislative roles, on the one hand, and its function as a ‘confidence college’ for the cabinet, on the other. The former roles may suggest highly proportional electoral systems, yet the resulting fragmentation of parliament raises worries about the identifiability, accountability and stability of cabinets. Political science has discussed this tension as one between the proportional and majoritarian ‘visions of democracy’ (Lijphart 2012; Powell 2000). Due to its separation of powers, semi-parliamentary government has the potential to balance these visions: the part of the assembly with confidence authority can achieve the ‘majoritarian’ values of identifiability, accountability and cabinet stability, whereas the separated part can allow for proportionality, a multidimensional party system and issue-specific deliberation on individual pieces of legislation (cf. Ward & Weale 2010).

The discussion proceeds as follows. I begin by locating semi-parliamentarism in a general typology of executive-legislative relations and providing a minimal definition of it. Then, I construct an ‘ideal-type’ of semi-parliamentary democracy and analyse how far the (minimally) semi-parliamentary systems in Australia and Japan approximate it. After elaborating on the theoretical argument, I go on to summarise the electoral designs in the bicameral Australian polities and empirically compare the balancing of the majoritarian and proportional visions of democracy in New South Wales and the Australian Commonwealth to the balancing achieved in 20 parliamentary and semi-presidential systems. Finally I discuss new semi-parliamentary designs before drawing my conclusions.

Typology and minimal definition

The dominant practice in political science has been to neglect upper houses in classifying a democracy's executive-legislative system. In my view, this practice cannot be justified if the upper house is directly elected and thus prima facie equally legitimate as the lower house.3 If two houses of parliament have an equal claim to represent the people, but only one of them can dismiss the prime minister, then a hybrid system is established. Once we understand the structure of this hybrid, we can also see that it does not require formal bicameralism. In developing the following typology, therefore, I will sometimes speak of the two ‘parts’ of an assembly rather than its two ‘houses’.

Table 1 distinguishes six types of executive-legislative systems based on how the executive comes into and survives in office (Shugart & Carey 1992). The types form three logical pairs. The first consists of the two pure types. A popularly elected fixed-term president, whose survival in office does not depend on the assembly, defines a presidential system. If two directly elected houses exist, the president must not depend on the confidence of either of them. In a parliamentary system the executive emerges from the assembly and its survival depends on the assembly's ongoing confidence. If two equal houses exist, it is dependent on both of them (as it continues to be in Italy after the failed constitutional referendum of December 2016).

The second pair consists of two hybrids that combine one parliamentary with one presidential feature. ‘Assembly-independent government’ (Shugart & Carey 1992: 26) describes a system in which the executive is voted into office by the assembly but, once in office, cannot be dismissed in a no-confidence vote. If there are two equally legitimate houses of parliament, as is the case in Switzerland, the survival of the cabinet must not depend on either of them. A popularly elected prime minister, who can be dismissed in a no-confidence vote of the assembly, defines prime-ministerial government. If the assembly consists of two equally legitimate houses, both must have this dismissal right. Israel established a unicameral version of this system in the mid-1990s and abandoned it again a few years later (Ottolenghi 2001).

The third pair also mixes elements of parliamentary and presidential government, but it does so by dividing either the executive or the assembly into two parts with equal democratic legitimacy. In semi-presidentialism a fixed-term president is legitimised through popular (direct) elections, but there is also a prime minister dependent on parliamentary confidence (Elgie 2011). If there are two equally legitimate houses, both must be able to dismiss the prime minister, as is the case in Romania. Finally, in semi-parliamentarism both parts of the assembly (most commonly: both houses) are legitimised though direct election, but the prime minister and his or her cabinet are dependent on the confidence of only one of them.4 Both hybrids establish a partial dependence of the executive on the assembly's confidence: either only a part of the executive is dependent on confidence (semi-presidentialism), or only a part of the assembly needs to provide this confidence (semi-parliamentarism).

Note that this definition does not include any requirements about the legislative power of the upper house, especially its veto power. This is in line with the most recent literature on executive-legislative systems. While earlier work on presidentialism and semi-presidentialism views the formal powers of the president as a defining attribute (Duverger 1980; Shugart & Carey 1992), more recent work does not (Cheibub et al. 2014; Elgie 2011). Yet legislative power is certainly important. I argue below that an ideal-type of semi-parliamentary democracy requires that at least the separated part of the assembly (the upper house in the bicameral case) have absolute veto power over all (non-budgetary) legislation. The minimal definition also does not rule out the possibility that the cabinet is partly drawn from the part of the assembly without confidence authority (the upper house). Upper house ministers are common in Australia. Finally, while the suggested minimal definition of pure semi-parliamentarism rules out popular executive elections, there can also be a ‘semi-parliamentary’ version of semi-presidentialism. This more complex hybrid exists in the Czech Republic, which has a directly elected, fixed-term president, a prime minister dependent on lower house confidence and a directly elected upper house without the power to dismiss the cabinet. However, the Czech constitution gives neither the president nor the upper house strong formal powers. The lower house can overrule their legislative vetoes with absolute majorities.

Based on the minimal definition, we can identify seven (pure) semi-parliamentary democracies: the Australian Commonwealth, five Australian states and Japan.

A semi-parliamentary ideal-type and its approximations

This section develops a richer definition of semi-parliamentary ‘democracy’. It is a more ‘ideal-typical’ definition (cf. Strøm 2000), although not in the strict sense that no real-world system could satisfy its conditions. One purpose is to clarify the logic of democratic delegation embodied in this constitutional regime. Another is to provide criteria for judging how closely the (minimally) semi-parliamentary cases approximate the ideal-type.

In the language of agency theory, the logic of democratic delegation in a semi-parliamentary regime is that voters, as the ultimate principal, select two separate but equally legitimate legislative agents, only one of which then becomes the principal of the prime minister and his or her cabinet. The cabinet is hierarchically subordinate to one part of the assembly, but not the other. This logic of delegation implies three desiderata. First, the democratic legitimacy of the two parts of the assembly ought to be equal. Second, the survival of the cabinet ought to be fully independent from one part of this assembly. Third, the legislative veto power of this separated part of the assembly ought to be absolute. These desiderata have six specific institutional implications, which I discuss in turn (see Table 2).

The first element to be considered is democratic legitimacy. By the minimal definition given above, all seven systems in Table 2 have directly elected upper houses. However, the relative (normative) legitimacy of the upper house is reduced if it is more malapportioned than the lower house – that is, if the apportionment of districts violates political equality to a greater extent.5 It is well-established that malapportionment also limits the perception of upper houses as legitimate (Russell 2013: 384). The former Australian Prime Minister Paul Keating famously referred to senators as ‘unrepresentative swill’. Most upper houses considered here have greatly reduced malapportionment. It is completely absent in the upper houses of New South Wales and South Australia due to their single state-wide districts. However, greater malapportionment than in the respective lower house exists in the Australian Senate (30 per cent), the Legislative Council of Western Australia (around 25 per cent) and the Japanese House of Councillors (16 per cent in the constituency tier of the electoral system).6

One might object that once we consider any deviations from equal legitimacy of the two houses, the upper houses of countries like Germany or the Netherlands might also be included in the analysis. Though not directly elected, their – formal and perceived – democratic legitimacy is quite high and might be on par with (some) directly elected but malapportioned upper houses. I am sympathetic towards extending the analysis in this way (cf. Ganghof 2014: 656, n. 10). However, since I contend that we must see a certain type of bicameralism as changing the executive-legislative system, consistency requires that I choose a restrictive operationalisation of upper house legitimacy. Analysing additional bicameral systems as semi-parliamentary might be analytically fruitful, but it is no logical necessity in terms of our standard criteria for classifying political regimes and executive-legislative systems.

The second element we will look at is upper house confidence. By the minimal definition of semi-parliamentarism, all upper houses considered here lack a no-confidence vote. However, an absolute veto over the budget can be used as a de facto no-confidence vote. New South Wales, Victoria and Japan lack an absolute budget veto and thus come closer to the ideal-type.7 The other cases have an absolute budget veto and deviate more from it.

Third, the upper house's absolute legislative veto power (with the exception of the budget) is not part of the above-proposed minimal definition of semi-parliamentarism, but an important part of the underlying ideal of democratic delegation. If the lower house possesses absolute veto power, as it typically does, and the upper house lacks it, as in the Czech hybrid, the constitution itself makes it clear that the two houses are not equal agents of the voters. In contrast, I think that the absolute veto power of the lower house is no necessary part of semi-parliamentary democracy. Since the lower house is the principal of the cabinet and may have substantial agenda-setting and dissolution powers, a lack of veto power would not necessarily undermine its equal status. I will return to this issue later in the article.

In all minimally semi-parliamentary cases, except Japan, upper houses do have absolute veto power. Even if a veto is formally absolute, though, it may be substantially weakened in the dispute resolution procedure between the two houses. The crucial question is whether dispute resolution can involve a joint session deciding a conflicted issue by simple or absolute majority and thus favouring the larger (lower) house. This is the case in the Commonwealth parliament and in Victoria. All other Australian states maintain the equality of veto power (Stone 2008). Tasmania and Western Australia have no provisions for dispute resolution; South Australia's constitution provides for a double dissolution election but no joint session; in New South Wales persistent deadlock can only be resolved in a popular referendum in which the lower house is the agenda-setter but both houses lose their veto power.

Table 1 is obviously a great simplification, but the overall picture is rather clear. New South Wales comes closest to the constitutional ideal-type of semi-parliamentarism, the Australian Commonwealth and Japan are farthest away; the other Australian states are in-between. Note that this rough ranking does not imply the hypothesis that upper houses will be more assertive, the closer they are to the semi-parliamentary ideal-type. In all Australian polities, Westminster norms and conventions remain strong, and the legitimacy of the lower house is widely perceived as being superior (e.g., Reynolds 2011). This is partly explained by the fact that Australian upper houses at the state level had a long anti-democratic history and were only democratised rather recently – 1978 in the case of New South Wales (Clune & Griffith 2006). Another reason for the greater perceived legitimacy of the lower house is precisely its exclusive authority to dismiss the cabinet. This, I have argued, is a public misconception that political science should not replicate. After all, we do not question the democratic legitimacy of the assemblies in the United States, Uruguay or Switzerland because they lack a no-confidence vote. The existence or lack of confidence authority does not matter for democratic legitimacy but for the executive-legislative system.

Trade-offs in the design of executive-legislative systems

Semi-parliamentarism is a particular instance of what constitutional theorists have called the ‘new separation of powers’ (Ackerman 2000; see also Albert 2010). Proponents of presidential systems often conflate the popular election of the chief executive with the branch-based separation of powers, as if one could not be had without the other (Calabresi 2001: 54–55). The analysis of semi-parliamentarism makes clear that this is a mistake. This executive-legislative system can achieve the main advantages associated with presidentialism, while avoiding some of its main downsides. I start with the balancing of the ‘majoritarian’ and ‘proportional’ visions of democracy and then move on to the independence of (one part of) the assembly on matters of legislation.

Normative balancing across executive-legislative systems

Let us start by summarising the electoral systems in Australia's bicameral systems. As Table 3 shows, Tasmania differs from the other Australian polities. It uses semi-parliamentarism to achieve local, non-partisan representation in the upper house (Sharman 2013). Alternative vote in single-member districts (SMD) is one important element of this design; staggered yearly elections and a small assembly size are others (see Table 3). Partly due to these features, the Tasmanian upper house is dominated by independents. The resulting normative balance is that ‘[p]rogrammatic choices can be made through parties at lower-house elections, supplemented with local representation through Independents in the upper house’ (Sharman 2013: 344).

The semi-parliamentary constitution is crucial for this equilibrium: if the upper house possessed the right to dismiss the prime minister, it would be democratically unacceptable that voters can never hold the upper house as a whole accountable for its actions and that these actions are not organised in terms of programmatic choices (cf. Fewkes 2011: 91).

As to the trade-off between ‘majoritarian’ and ‘proportional’ goals, Tasmania adopts the above-mentioned ‘sweet spot’ (Carey & Hix 2011) solution of using proportional representation (PR) in small districts in the lower house (all M = 5 since 1998). Indeed, it manages to have both a low effective number of legislative parties and low empirical disproportionality at the same time (see Table 3).

The more prevalent approach of using the semi-parliamentary constitution exists in the other bicameral polities in Australia. It uses upper houses to achieve greater ideological, policy-based representativeness. This strategy combines Alternative vote in SMD in the lower houses with Single transferable vote in upper houses. The degree of proportionality achieved in the upper houses varies due to the variation of assembly sizes and district magnitudes (Farrell & McAllister 2006). Again, New South Wales stands out with the highest district magnitude and lowest disproportionality.

To map the normative balance achieved by these semi-parliamentary systems more precisely, I focus on New South Wales and the Australian Commonwealth (for which expert estimates of party positions are available) and compare them to parliamentary and semi-presidential systems. The results will also inform the comparison to presidential systems, as explained below. I compare these systems’ capability of simultaneously achieving six goals, three for each ‘vision’ of democracy (Powell 2000).9 The details of the six resulting indicators are provided in the Appendix. All of them take account of directly elected upper houses.

The three ‘majoritarian’ indicators are as follows. (1) Party-based Identifiability (Ident) combines information on how much votes are concentrated on two competing pre-electoral blocs, on whether the government consists of a single bloc and whether it has majority status (cf. Ganghof et al. 2015; Shugart 2001). Since the focus is on party-based identifiability, the indicator neglects presidential elections. (2) Clarity of responsibility (CoR) measures the ability of voters to determine which parties are responsible for past policies. The ranking of cabinet types is similar to the one proposed by Powell (2000: 53). (3) Cabinet stability (Stab) relates the average term length of cabinets to the constitutional maximum.

The three ‘proportional’ goals are as follows. (1) Proportionality (Prop) is simply the inverse of Gallagher's (1991) Least Squares index for the more proportional house.10 (2) Dimensionality (Dim) is the effective number of dimensions (Ganghof et al. 2015) in the house with higher dimensionality, based on expert survey data by Benoit and Laver (2006) as well as Pörschke (2014). This indicator captures the idea that if voters’ preferences are latently or potentially multidimensional – if their positions are ‘right’ on some issues and ‘left’ on others – then representativeness requires that this multidimensionality be reflected or constructed in the assembly (cf. Stecker & Tausendpfund 2016).11 Finally, one version of the proportional vision wants to potentially allow all parties in the assembly to participate in decision making (Powell 2000: 256, n.9). Hence, (3) legislative flexibility (Flex) measures – very roughly – the extent to which governing parties commit themselves to a fixed majority coalition or remain free to choose between different support parties.

Figure 1 presents the trade-off profiles of six countries on these six variables for the period 1995–2015. All variables are period averages standardised between 0 and 1. This standardisation is based on the minimum and maximum values in a sample of 21 advanced democratic nation-states plus New South Wales (see Appendix).12 The figure also plots the distribution of values in this sample along the six dimensions in Figure 1.

The United Kingdom and Denmark exemplify the polar ‘visions’ of pure parliamentarism. They realise one vision at the expense of the other (cf. Powell 2000). Ireland and Italy are examples of semi-presidential and parliamentary systems, respectively, that try to balance the competing visions. The Irish strategy has been based on PR in moderately sized districts; the Italian on electoral systems that – until recently – encouraged the formation of competing pre-electoral alliances. We see that both cases do have somewhat more balanced trade-off profiles, but they are neither fully balanced nor do they achieve very high values on any variable (perhaps with the exception of cabinet stability in Ireland). New South Wales and the Australian Commonwealth represent semi-parliamentarism. Both cases balance the two visions of democracy in that they realise the majoritarian goals about as well as the United Kingdom, but also reach high levels of proportionality and flexibility in their upper houses. They may only fall short, in comparison to PR-parliamentary systems like Denmark, in representing the (potential) multidimensionality of voter preferences. However, this is not an inherent limitation of the semi-parliamentary constitution but most likely a result of the path-dependent designs that evolved in Australia. Most notably, the small sizes and/or district magnitudes of upper houses are likely to limit the effective number of parties and dimensions (Li & Shugart 2016). The next section will discuss ways to reduce the constraint on dimensionality, if desired.

The empirical analysis also speaks to the comparison with presidential systems, even though they are not explicitly included. Students of presidentialism emphasise that in presidential systems identifiability is ‘institutionally guaranteed’ and that ‘presidential institutions provide a context of more “clarity of responsibility”’ (Cheibub 2006: 361). Here we see that semi-parliamentary systems can achieve equally high levels of identifiability and clarity of responsibility in a party-based manner. Moreover, the next section shows that adequately designed semi-parliamentary systems can also institutionally guarantee identifiability.

The design of semi-parliamentary democracy

The bicameral systems in Australia have evolved in a long political process that often reflected the self-interest of, and compromises between, partisan elites rather than some grand democratic design (e.g., Clune & Griffith 2006). If constitutional designers in other systems were to consider semi-parliamentarism, which designs would be worth keeping, or changing? This section provides some preliminary answers.

Conclusion

Semi-parliamentary government is a distinct and neglected type of executive-legislative system that deserves consideration by scholars and constitutional designers alike. Parliamentary systems struggle with the tension between the proportional and majoritarian visions of democracy, and semi-parliamentary government is one approach to balancing these visions. Presidential systems struggle with the various negative effects of a popularly elected, fixed-term president, and semi-parliamentarism is one approach to avoiding these effects while maintaining or strengthening the advantages of a branch-based separation of powers highlighted in the literature.

One question for further research is how, and how successfully, legislative coalitions form under semi-parliamentarism. Another is how well semi-parliamentary systems perform in the sense of producing good outcomes. A recent study by Schwindt-Bayer and Tavits (2016: 56–57) argues that high clarity of responsibility reduces corruption, and they single out Australia as a democracy that was more effective in fighting corruption when it was governed by single-party majority cabinets – despite the fact that these cabinets usually lacked a majority in the proportionally elected Senate. While this is just one example, and perhaps a contestable one, it suggests the possibility that semi-parliamentary systems may combine certain performance advantages of ‘majoritarian’ and ‘proportional’ democracies. This possibility deserves further investigation.