Sartori's legacy and the classification of party systems

The classification of party systems has long been a central focus in political science. One of the earliest and most enduring classification is Duverger's (Reference Duverger1951), which identified three main categories: one-party systems, two-party systems, and multi-party systems. Years later, Blondel (Reference Blondel1968) advanced this framework by considering not only the number of parties but also their relative parliamentary strength. This approach provided a more nuanced understanding of party system formats, exemplified by the case of two-and-a-half party systems (such as Germany at the time) or multi-party systems with (e.g., Sweden) or without (e.g., the Netherlands) a dominant party. Similarly, Rokkan (Reference Rokkan1970) offered a more complex classification of party systems, using criteria such as the likelihood of achieving a one-party majority and the distribution of votes among competing political parties.

However, the most influential typology is undoubtedly Sartori's (Reference Sartori2005), which surpasses previous classifications in several respects. Sartori not only provided a method for ‘intelligently’ counting parties but also incorporated the criterion of ideological distance alongside the numerical count. This allowed for a more refined differentiation of multi-party systems, categorizing them not only as limited, extreme, and atomized but also as moderate or polarized. As Mair highlighted, Sartori's typology is ‘the most comprehensive of all the available typologies, both in terms of the care with which it is developed, as well as in the way it is applied to empirical cases’ (2005, XVI). This is because it focuses explicitly on patterns of party competition and interaction, directly addressing the functioning of party systems. Furthermore, ‘it underlines the influence exerted by systemic properties, and by the party system, on electoral behavior and patterns of government formation’ (Mair, Reference Mair and Sartori2005, XVII).

Over time, Sartori's typology has gradually become less effective in capturing and differentiating the evolving nature of party systems, which began to change after a prolonged period of stability. Specifically, the increasing fragmentation of political parties initially overextended the category of limited and moderate multiparty systems, subsequently leading many cases to be classified as extreme multiparty systems. However, this shift did not necessarily coincide with a rise in polarization.Footnote ¹ This development also challenged one of the fundamental assumptions of Sartori's theory: that the number of parties has ‘mechanical predispositions,’ meaning it inherently determines the nature of competition among parties. Indeed, Sartori himself acknowledges that one ‘cannot be happy with his classifications once and for all: his world being extremely dynamic, he must reclassify, and classify anew, all of the time’ (2005 [1976], 261–262).

Given these challenges, scholars have sought to address the evolving nature of party systems through various new classifications. Some of them have diverged from Sartori's approach, while others have sought to build upon it. For instance, Mair (Reference Mair1997, Reference Mair, Graziano and Vinck2007) introduces a distinction between ‘open’ and ‘closed’ party systems, based on factors such as the degree of government alternation, the familiarity or innovation of governing formulas, the presence of new parties in government, and the frequency of changes in the composition of ruling coalitions.

Conversely, several scholars have worked within Sartori's framework but have introduced modifications or enhancements. Siaroff (Reference Siaroff2019), for example, combines the numerical criterion with the relative strength of the parties, resulting in a typology that intersects with those proposed by Sartori and Blondel. Similarly, Wolinetz (Reference Wolinetz2004, Reference Wolinetz, Katz and Crotty2006) incorporates the number of parties and ideological polarization but adds two additional criteria: the form of competition and the presence of party clusters. Nwokora and Pelizzo (Reference Nwokora and Pelizzo2018) offer an updated version of Sartori's typology, redefining it based on both the number of parties and, more importantly, the patterns of power alternation. More recently, Casal Bértoa and Enyedi (Reference Casal Bértoa and Enyedi2021) have identified eight types of party systems, using classification criteria such as the number of poles, the degree of polarization, and the primary power configurations among parties or blocs.

It cannot be said that efforts to develop classifications more suited to the evolving conditions of party systems after Sartori have been particularly successful. None of the classifications discussed here have achieved a level of recognition within the academic community comparable to Sartori's typology. This prompts the question of why this is the case. One possible explanation may lie in the limitations of these newer classifications, perhaps due to the inadequacy or insufficiency of the criteria used to differentiate party system types. Another explanation could be related to shifts in scholarly interests, with researchers now less inclined to value the importance and explanatory power of classifications—particularly those of party systems, which some may regard as a somewhat outdated subject.

While acknowledging that both of the previously mentioned causes may have played a role, we argue that a third, more significant explanation exists for the marked decline in theoretical and empirical efforts to classify party systems in recent years. Even when new classifications have emerged, they have largely proven to be short-lived and idiosyncratic. This explanation is tied to the very nature of party systems themselves. According to Sartori's definition, party systems are ‘the system of interactions resulting from inter-party competition’ (2005 [1976], 39), and this system of interactions is becoming increasingly elusive. We believe that, even in the Western world—once known for the extraordinary stability of its party systems—many party systems are now characterized by increasingly unstable and unpredictable interactions among parties. In such cases, party systems lose their ‘systemness’ and transform into systemless ‘sets of parties’ (Bardi and Mair, Reference Bardi and Mair2008) or ‘party non-systems’ (Sanchez, Reference Sanchez2009).

In this context, classifying party systems becomes a complicated exercise—if not impossible or, worse, pointless and misleading. Therefore, this research investigates whether the transformations in Western European party systems have eroded their systemness to such an extent that the very feasibility and viability of traditional classifications are now questionable.

Classifying party systems: in search of systemness

As explained above, an empirical investigation of the ‘systemness’ entails party system change, originally defined by Mair (Reference Mair1997, 51–52) as the change that occurs ‘when a party system is transformed from one class or type of party system into another.’ Consequently, we do not have simply to look at shifts in parties’ electoral performances or their entry or exit from parliament or government across subsequent legislatures. Yet, the transformation of the party system from one type to another requires focusing on those shifts that truly affect the fundamental nature of a country's party system, such as moving from one to another pattern of inter-party competition. To do so, a preliminary requirement is to develop a classification that identifies meaningful ‘party system types’ — distinct patterns of inter-party competition (‘party system configurations’ emerging in each legislature) that remain stable over time (thereby providing systemness to party systems).

Since existing classifications of party systems are largely outdated, have lost empirical ground, or are nearly impossible to apply consistently across countries or time periods, here we propose a classification aimed specifically at investigating the systemness of party systems. This classification defines party system types based on the number of ‘poles’ that compete and the identity of the main political parties belonging to each pole. Therefore, poles, rather than individual parties, are the basic units of the classification. Poles are either individual parties or aggregations of parties that (are perceived to) present themselves in elections as government alternatives. Focusing on poles rather than parties is consistent with Casal Bértoa and Enyedi (Reference Casal Bértoa and Enyedi2021) and represents a conservative choice to assess party system change as poles are more stable than individual parties and tend to be perceived by voters as more longstanding entities. Moreover, poles also allow to take into account not just the format but also the mechanics of the system as it unfolds from the electoral to the governmental arena. Not by chance, Sartori himself focused on poles rather than parties when he had to outline the ‘structural configurations’ of party systems (2005, 254).

To identify a pole, we adopted both quantitative and qualitative criteria. The quantitative criteria provide, for each country in each legislature, the potential number of poles competing in the system, while the qualitative assessment helps us identify the specific composition of poles, namely the identity of the party or parties forming the pole. As for the former, poles must obtain a certain parliamentary strength — measured in terms of parliamentary seats — to be perceived and counted as a viable alternative for the national government. Under certain thresholds, the party or the group of parties potentially forming the pole is simply not relevant enough to be perceived as a viable alternative and cannot be counted as a pole. The debatable point is how to identify such thresholds of relevance. We let the latter vary from country to country depending on two main factors: the overall degree of party system fragmentation — calculated through the effective number of parliamentary parties, ENPP (Laakso and Taagepera, Reference Laakso and Taagepera1979) and volatility — calculated through the Pedersen's index (Reference Pedersen1979) on parliamentary seats, TPV (see Table 1).Footnote ²

Table 1. Threshold of relevance for a pole to be considered

Note: Cells represent the varying share of parliamentary seats a pole needs to get in order to be considered under different combinations of ENPP and TPV.

The two factors are fundamental party system properties (Pennings and Lane, Reference Pennings and Lane1998; Golosov, Reference Golosov2011) ‘exactly because they define the patterns of government alternation, register the emergence of new government alternatives, and reflect the availability of office access to new claimants’ (Golosov, Reference Golosov2011, 128). The underlying rationale is that an increase in both factors increases the entropy of the system and consequently relaxes the threshold of relevance a pole needs to pass to be perceived as a viable alternative. Therefore, the higher party fragmentation and volatility, the lower the parliamentary strength a party aggregation must have to obtain the status of a pole.

The party system literature provides guidance for the respective thresholds in the table. According to Siaroff (Reference Siaroff2019), the highly multi-party type, the most fragmented type in his typology, has more than six parties, and this is why we opted to include all instances above six effective parties in the same category of ENPP; as for the share of parliamentary seats needed at different levels of ENPP, we use 40% and 25%. According to Blondel (Reference Blondel1968), 40% is the minimum expected vote share for one of the two competitors in a two-party system and 25% is the typical distribution of votes for the two major parties in multiparty systems without dominant party. Then, such thresholds progressively relax as far as parliamentary volatility increases, with the anchor of 20 as a high-volatility threshold suggested by Mair (Reference Mair, Gallagher and Marsh2011).Footnote ³

Once obtained the potential number of poles competing in the system, we looked at the aggregations derived from explicit or implicit agreements. Poles can indeed be formed by a single party — strong enough to be perceived as a viable governing alternative — or can they take the form of pre-electoral coalitions (Golder, Reference Golder2006; Tillman, Reference Tillman2015), which are explicitly formalized by the partners. Alternatively, parties can cooperate by forming potential, though not formalized, coalitions based on ideological and programmatic proximity or habitual patterns of governing alliance, e.g., ‘clustering of parties’ (Wolinetz, Reference Wolinetz2004).Footnote ⁴ A detailed discussion with specific notes on each party system and the related sources can be found in the Appendix. As a result of this qualitative assessment, a parties’ configuration emerges in each legislature.

As shown in Table 2, a parties’ configuration is the combination of two key characteristics of the system in a given legislature: the first is the number of poles that may range from 0 to more than 2, namely from an Apolar to a Multipolar configuration; the second is the partisan identity of the pole, namely the main party making or leading the pole. A change in either one or both these characteristics means a change in the parties’ configuration (e.g., when the system moves from Unipolar to Bipolar or when, while staying Bipolar, the identity of the main party in a pole changes).Footnote ⁵ For the sake of this work, we deliberately neglect changes of minor importance — such as a change in the composition of a pole that does not affect the main partyFootnote ⁶ — that would further boost the instances of change.Footnote ⁷

Table 2. The basic scheme of our classification

As this study aims to assess the (lack of) systemness of Western European party systems, looking at changes in parties’ configuration across subsequent legislatures would not be sufficient. Indeed, parties’ configurations need time to stabilize and consolidate, thus becoming proper party system types. In this regard, we operationally define a party system type as a party configuration that persists for at least three consecutive legislatures. Three legislatures provide enough time to establish the systemness, namely the presence of a stable pattern of interparty competition.Footnote ⁸ Furthermore, three legislatures are also needed to observe the replacement of a party system type with another. However, the disappearance of a given party system type does not necessarily lead to the emergence of a new party system type as it may be replaced by a ‘party non-system’ (Sanchez, Reference Sanchez2009; Golosov, Reference Golosov2011).Footnote ⁹ In our definition, a party non-system arises when the number of relevant political poles and/or the identity of the main party associated with each pole remains in flux and fails to stabilize. Hence, all the legislatures where a party system type as defined above does not emerge are party non-systems. Party non-systems can be typical of political systems undergoing political transitions but can also occur in established democracies that experience a crisis of political representation or governance.

From party system types to party non-systems

The rules discussed above are applied to classify 20 Western European party systems after World War II.Footnote ¹⁰ Table 3 reports all instances of party system types with the respective timeframe.

Table 3. Party system types in Western Europe after World War II

Note: Parties indicated without parentheses are always part of the pole in the period considered; parties indicated in parentheses are sometimes part of the pole in the period considered; Oth = parties other than the main one that are part of the pole. The list of party abbreviations is available in the Appendix.

Overall, 32 party system types have characterized the post-WWII Western European party systems, with a prevalence of bipolar types (19) over unipolar ones (12). Conversely, multipolar configurations have rarely occurred but have always failed to stabilize as proper party system types. Finally, an apolar type has characterized Switzerland for as much as 60 years.

As Table 3 shows, the 32 party system types do not account for all the post-WWII years, indicating the presence of periods of party non-systems, that require further investigation. Analyzing party non-systems and their variations across time and countries may effectively address the question about the (lack of) systemness of contemporary Western European party systems.

Figure 1 shows the cross-time variation of party non-systems in Western Europe. It reports the predicted probabilities of a logistic regression model where the dichotomous variable ‘Party non-system’ in a given legislature is regressed on the categorical variable for decades. We have also added country-fixed effects to take into account country-specific unobserved heterogeneity (Beck and Katz Reference Beck and Katz1995).Footnote ¹¹

Figure 1. Predicted levels of party non-system by decade.

The contrast between the 1960s and the most recent decades is striking. While in the ‘golden age’ of party system stability (Janda and Colman, Reference Janda and Colman1998), the probability of having a party non-system was low (14% in the 1960s and under 20% in the 1980s), it has risen sharply since the 1990s, reaching 50% in the 2010s — the same as having a defined party system type.

The decline of systemness in Western Europe since 1989 can also be appreciated by looking at cross-country variation in Figure 2. Only Iceland and Switzerland consistently maintained a party system type (Unipolar IP in Iceland; Apolar in Switzerland until 2007, and, later on, Unipolar Swiss People's Party, SVP). In contrast, 11 countries have had party non-systems in at least half of their legislatures, and in Belgium, the Netherlands, and Spain, a stable party system has never consolidated. The other countries where party non-systems represent at least half of the legislatures after the fall of the Berlin Wall are countries like France and Italy, well-known for their low party system institutionalization (Chiaramonte and Emanuele, Reference Chiaramonte and Emanuele2022), Southern European, third-wave polities like Cyprus and Greece, Scandinavian countries like Norway and Finland, a small polity like Luxembourg, and eventually an idiosyncratic political context like Ireland. In a nutshell, the crisis of systemness in Western European party systems affects the region broadly, without regard to geography or institutional settings.

Figure 2. Share of party non-systems’ legislatures after 1989.

Conclusion: classifying party systems is still useful?

The results presented in this research highlight a significant decline in the systemness of Western European party systems over the past decades. The analysis reveals that the majority of such systems have experienced a ‘non-system’ type for at least 50% of their legislatures since 1989. Party systems are increasingly characterized by flux, with the number of relevant poles and the identity of major parties failing to stabilize. The fluid and dynamic nature of party configurations makes it difficult to identify enduring patterns of interparty competition. This raises the question of whether party system classifications remain useful for capturing variations across different contexts and time periods or if they have become obsolete, offering only fleeting snapshots of party landscapes.

In this regard, one could recall Sartori's lesson, who warned about the limitations of classifications for inchoate forms. Indeed, while classification is essential for unpacking concepts and facilitating comparative analysis (Sartori, Reference Sartori1970), attempts to classify party systems lacking ‘structural consolidation'Footnote ¹² are likely to be misleading. This is because ‘inchoate forms not only tend to defy classifications, but tend to be misrendered by classifications. The very act of assigning something to a type imposes upon it a definiteness, a fixity, a form. Therefore, with respect to a state of flux, classes and types can be more deceptive than informative’ (Sartori, Reference Sartori2005, 227).

However, first, identifying cases of party systems that lack ‘structural consolidation’ — or, as we have referred to them here, party non-systems — still requires a classificatory process that distinguishes between party systems and party non-systems based on their respective levels of systemness (however defined). Second, classifications remain essential for comparing and understanding party systems that do exhibit ‘structural consolidation.’ After all, Sartori himself built his famous typology of party systems by excluding precisely those lacking structural consolidation.

So, to sum up: yes, classifying party systems is still useful, even though true party systems are becoming increasingly rare. However, it is necessary to adopt a different, new approach — one that implicitly or explicitly incorporates party non-systems into the classifications themselves. This means adopting a taxonomic approach — first distinguishing between party systems and party non-systems and then differentiating between various classes or types of party systems. Only in this way is it possible to properly account for the current state of party systems in Western Europe, where we have observed that systemness has become the exception rather than the rule.