Sartori's approach to parties: the systemic implications of counting

The distinctiveness of Sartori's approach was evident from the beginning. His first internationally impactful contribution to the party system literature already included a definition of the qualitative and quantitative parameters for classifying party systems and an articulated discussion of how the party system as a whole's characteristics impacted its parts and the parties (Sartori, Reference Sartori, La Palombara and Weiner1966). Sartori's argument began with a critique of Duverger's (Reference Duverger1964) “laws,” to which he was probably much more indebted than ever willing to admit. Here, his concern was establishing how many types of party systems exist in Western democracies. Sartori convincingly argued that there are three types of party systems, not two, as Duverger had positedFootnote ¹. This very simple distinction of what appear to be static empirical observations was based on an inherently dynamic four-dimensional scheme. Besides the number of parties already present in Duverger's classification, it consisted of the number of poles, the level of polarization, and the direction of competition. As such, it already included all the elements needed to develop a theory of party systems. At a minimum, it could be considered an analytical framework more than a classificatory template.

This important potential theoretical advancement was unnoticed because of the enormously positive reception of the innovation to the classification of party systems Sartori produced by including polarized pluralism in the framework. Yet, the rich discussion of this model had significant explanatory power and relevant theoretical value. To explain why the distinction between party systems cannot be only between two-partyism and multi-partyism, as well as why, in certain systems, there are three poles and not two, Sartori (Reference Sartori, La Palombara and Weiner1966: 138) introduced the concept of distance. He defined it as ideological in later works, especially when discussing Downs' spatial model of competition (Sartori, Reference Sartori1976: 342 passim). This simple innovation brought with it a cascade of implications of undoubted theoretical significance. In fact, distance is a factor that determines both the direction of competition between the parties and the level of basic consensus for the regime, which he derived from Easton (Reference Easton1965a, Reference Easton1965b). It has causal consequences that, when observed and measured, contribute to the explanation of party actions, but above all, of the effects that these produce on the moderation of the party system and ultimately also on the stability of the government.

As we will see in more detail, there is a relationship between the number of parties in the system and the number of poles (two or three) around which the system pivots. The significance and nature (moderate or extreme) of the system's polarity depends, in turn, on the distance between the poles. Sartori's argument is still descriptive and not yet theoretical because it does not clearly state what comes first and what causes what. Yet, adding distance to Duverger's scheme helps explain why moderation can prevail in some cases, even in multipolar systems.

From counting to theoretical thinking

Sartori's interest and attitude continued to be oriented towards refining his analytical framework rather than its theoretical deepening. However, the discussion that accompanied the innovations made to his classification (a term Sartori assertively claimed instead of typology) increased its complexity and augmented its conceptual and theoretical implications. The most obvious of these innovations was adding four classes to the three he had already defined (two-party systems, moderate pluralism, polarized pluralism): one-party systems, hegemonic party systems, predominant party systems, and atomized systems (Sartori, Reference Sartori1976: 125–129). These new classes define systems with uncommon or extreme characteristics. Therefore, they are not used often for classifying democratic systems. But the definition of these new categories suggested to Sartori the usefulness, if not the necessity, of introducing a new variable to the framework: the distribution of power, from total concentration (monopoly) to total dispersion, passing through various degrees of hierarchy, fragmentation and segmentation. The distribution of power is a function of the number of parties, of the format of the system, that is. The format, in turn,

This evolution of Sartori's approach to the study of parties and party systems made another step towards increasing its explanatory theoretical power. Again, his point of departure was the counting of parties. By now, however, the numerical count à-la-Duverger had been replaced by an indexation of power distribution among the parties. This was coupled with the notion that only relevant parties (with sufficient power in the system, that is) could be included in the count that determines, in Sartori's terminology, a system's format. This time, the target of his attention was Douglas Rae's (Reference Rae1967) index of electoral fractionalization, which was based on a weighted calculation of parties' shares of electoral votesFootnote ².

Sartori's answer to this idea was that the criteria to assess the relevance of parties should not be mathematical but nominal and determined by position value and not electoral strength. Rae's index measures party size in terms of the distribution of seats among the parties, but Sartori focuses instead on what he calls natural size thresholds. He does not provide a numerical measure of party size but provides descriptive nominal categories: “(i) a protracted 50 per cent or greater majority of one party (predominant system); (ii) two parties nearing the 50 per cent majority; (iii) no party nearing the absolute majority and, within this group, either relatively few parties (from three to five) or numerous parties (six and more)” (Sartori, Reference Sartori1976: 315).] By conceptualizing format and listing its classes, “predominance, twoparty near evenness, limited fragmentation, and extreme fragmentation” (Sartori, Reference Sartori1976: 316), Sartori, also provides the basic notions to define and operationalize, without saying so, the party system's structure as a potential independent variable in his embryonic theory. Sartori admits that “the mathematical properties of Rae's index are indeed attractive properties; but the ‘position value' of parties may well establish a more realistic correspondence between theory and fact” (Sartori, Reference Sartori1976: 316). And concludes by timidly hinting at the importance of theory: “Words alone beat numbers alone. Words with numbers beat words alone. And numbers make sense, or much greater sense, within verbal theory” (Sartori: 319, emphasis mine).

Coming short of a theory

In the next and final chapter of Parties and Party Systems, Sartori reverted to his critique/discussion of Downs' Economic Theory of Democracy (Downs, Reference Downs1957), which had been central to his first English language article on parties (1966). This effort, which elaborated and, in his own words, improved previous versions of the discussion, was and is probably the theoretically most advanced version of his reflection on party systemsFootnote ³. Sartori started out by taking a critical look at Downs' model. To a greater extent than in other cases, the better part of his discussion was aimed at an empirical contestation of Downs’ key assumptions, largely based on findings in the literature. The discussion touched on many features of the model. The most compelling criticism probably concerned the idea that parties can compete in a unidimensional spatial model. First, because, unavoidably, space is not neutral and requires policy and, more often, ideological positioning, detracting from the purity of the model. Moreover, ideology is not a fundamental factor in many polities and issue voting often prevails. Secondly, because more often than not, the space in which parties compete has more than one dimension. For example, the Netherlands and Israel's spaces are structured by two dimensions and Switzerland's by three.

Sartori was aware that this approach could dangerously lead to a complete dismissal of Downs' model, which, at some level, he still considered useful. He accepted that models are simplifications of reality and that the narrower the issue they address, the better they fit. As such, he downgraded the theory of democracy to a theory of voting behavior, where party positioning was the independent variable. His attempt to “simplify” the analysis while at the same time “saving” part of the Downsian model produced another, albeit positive and important, complication. This was in the form of introducing a new variable, the direction of competition. The unequal ideological intervals that are observable in the unidimensional competitive electoral space can produce pulls on party positioning that are incompatible with the centripetal orientation assumed by Downs. Sartori observed that if extreme parties are far enough from the center, they will be more interested in positioning themselves near secure extreme voters than trying to attract available ones who may be very far, possibly at the other end of the unidimensional space. They would probably lose more identifiers than attract available voters if they did. This being the case, the direction of competition becomes centrifugal.

In purely mechanical terms, the higher the number of parties, the higher the likelihood that the direction of competition will be centrifugal, as the amount of space covered by parties away from the center necessarily increases. This possibility becomes a certainty when, above a critical threshold in the number of parties that Sartori places at five, “polities display a strong ideological focus” (1976: 349). The two poles of the system move towards the extremes to leave room at the center for the creation of a third, central pole. This stabilizes the moderate electorate near the center that replaces the floating voters so dear to the Downsian bipolar model. The virtual monopoly the center parties acquire in their portion of the competitive space induces them to expand by attracting voters on both sides. The direction of competition thus becomes centrifugal because the pressure coming from the center pushes all the other parties towards the extremes. Thus, ideological distance and direction of competition are crucial variables for explaining the mechanics of party systems, even if the number of parties, once the criteria for their counting are agreed upon, is still most important. Less for Sartori, however, than for Duverger, Rae or Taagepera.

Party systems: complexities and analytical parameters

Sartori's conceptual and theoretical advancements notwithstanding, his framework is still valued mainly for the fit of its typology. Peter Mair remarked that the scheme developed by Sartori allows us to record the changes experienced by parties and party systems with precision. However, their explanation still appears difficult (Mair, Reference Mair, Katz and Crotty2006). This is because the description of systemic changes is often limited to the individual components, the parties. Furthermore, the explanation of change generally focuses on the rules and dynamics of competition in the electoral arena. It does not consider the complex system of other arenas where parties compete. More generally, the probability that a given political system will maintain more than one party system is increased by the presence of three other types of possible divisions within the political system itself: a) vertical divisions, more readily associated with political systems characterized by “pillar” structures (Lijphart, Reference Lijphart1968) such as Belgium; b) horizontal divisions, which can be associated with highly decentralized political systems in which the lower units enjoy considerable autonomy in the exercise of their political prerogatives; c) functional divisions, in which the modes of interaction between parties vary considerably in the different functional arenas (electoral, parliamentary, governmental).

A correct understanding of the parameters for studying party systems requires a comprehensive evaluation of vertically, horizontally, and functionally distinct arenas in which parties interact (Bardi and Mair, Reference Bardi and Mair2008). Primarily this relates to the relevance of the different arenas of potential competition between parties and then also to the institutional context, to the opening or closing of the system, to the models of alternation between government and opposition parties (Bardi, Reference Bardi and Bardi2006: 268). For each of them, the structural elements of the party system must be observed: the polarization or ideological distance between the extreme poles of the system and the number of relevant parties.

Moreover, parties differ functionally in different arenas. The electoral party system is quite closed, at least from an analytical point of view. Its functioning is usually quite well isolated from other (sub)systems. Only after obtaining their main electoral goal will the parties “begin to pursue the other objectives dictated by the different competitive logics (e.g. in the German system, favouring with second votes the survival of a weaker potential coalition partner, [….]). Moreover, even anti- system parties, as long as they compete, are part of the system in the electoral arena, even though they may be excluded or exclude themselves within the parliamentary arena (Bardi and Mair, Reference Bardi and Mair2008, 158-159). These undeniable conceptual and empirical obstacles are probably some of the reasons for Sartori's hesitancy in attempting to produce a more comprehensive theoretical construct and decision to only consider the national electoral arena.

Is Sartori's framework a party system theory after all?

So far, our discussion has emphasised the contradiction between Sartori’s undeniable propensity for conceptualisation and theoretical reflection and his coming short of formulating a fully-fledged party system theory. One of Sartori's most explicit theoretical discussions on the party system was in the article “The party‐effects of electoral systems” (Sartori, Reference Sartori1999). The article starts rather abruptly with the statement that “debate whether electoral systems are an independent or dependent variable is pointless.” The reason is, Sartori quickly points out, that the party system can be seen both as an independent and a dependent variable.

As such, he establishes a causal chain that starts with the electoral system and ends with the parties.

Sartori then asserts “that electoral systems ‘cause’ the number of [relevant] parties,” according to empirical laws that indicate how plurality or proportional representation (PR) electoral laws produce different types of party system formats. Having so determined how the structure of the system (the format) is created, he then discusses how “the format explains and predicts the mechanics, i.e., the systemic characteristics of distinctive types of party systems, identifying “three major systemic patterns: (1) two-party mechanics, i.e., bipolar single-party alternation in government; (2) moderate multi-partism, i.e., bipolar shifts between coalition governments; (3) polarized multi-partism, i.e., systems characterized by multi-polar competition, centre-located coalitions with peripheral turnover and anti-system parties” (Sartori, Reference Sartori1999: 16). Finally, Sartori goes as far as formulating on this basis three testable hypotheses respectively 1) on the polarization conditions under which the two-party format produces two-party mechanics or not; 2) on how the three-four party format engenders “the mechanics of moderate multi-partism if and only if, the polity does not display high polarization”; 3) on how five-to-seven party formats generated by PR can produce moderate multi-partism polarized multi-partism according to the level of polarization of the system.

Structurally, this formulation of Sartori's framework has all the elements of an empirical party system theory. All that is needed to recognize it as such is to separate analytically different parts of the causal chain that starts with the electoral system and ends with the party system's mechanics. Suppose the format of the system is taken as a given, as the starting point of a reduced causal chain, that is. In that case, the theory states that the number of parties and their positioning in the unidimensional competitive electoral space cause the system's mechanics. What we have is thus a very parsimonious theory of how the party system's format determines party interactions and their effects.

Why is this so rarely acknowledged? Probably for the very reasons Sartori himself never claimed he had formulated such a theory. The main one is that he never took the number of parties in the system as a given. His discussion always went back to how the electoral system affected their number and how it should be counted. Moreover, he felt uneasy about some empirical limits of its assumptions and explanatory power. On the former, Sartori himself had already convincingly dismissed potential criticism when he closed “Parties and Party Systems” by quoting Downs and stating that models “should be tested primarily by the accuracy of their predictions rather than by the reality of their assumptions” (Downs, Reference Downs1957: 21, cited in Sartori, Reference Sartori1976: 351). Indeed, assumptions must be useful, not necessarily empirically solid. The assumption of human rationality, which is at the basis of Down's theory, is highly questionable empirically, but it permits the prediction of the most likely forms of behavior. Irrationality, instead, would make all behavioral options equally possible and would thus be useless. On the latter, we could say that perhaps Sartori expected too much from his framework. The complexity of the phenomena he intends to explain with the whole length of his causal chain requires the consideration of dynamics in the political system at large and of the many arenas and different levels (national and subnational, but more recently also supranational) in and at which parties compete.

In conclusion, had Sartori limited himself to the electoral arena, which he did in developing the model but did not when discussing its implications for governmental stability, most of the theory's potential explanatory limits would not have manifested themselves. Party relations in the electoral arena, like state relations in the international system, are determined as an independent variable by the structure of the system (Waltz, Reference Waltz1979). Like states, parties have an identical function/interest: to maximize their ability to survive. States achieve their goals by maintaining their sovereignty and parties by obtaining votes. Thus, a systemic theory of party relations would not have to explain party behavior but, like Waltz’s theory does for states, only the combined effect of all party actions, in terms of the system’s mechanics. If limited to the electoral arena, Sartori's theory certainly does that. The difference with Waltz's theory is that, in Sartori's, the dependent variable is not dichotomous (system's stability/instability) but ordinal (quantitative and qualitative distribution of power among the parties). This makes for a less parsimonious theory, requiring other independent or intervening variables besides the number of parties: ideological distance and direction of competition. This difference notwithstanding, everybody considers Waltz's theory a veritable theory of international relations, and we should also accept Sartori's framework as a general party system theory. Appropriate mid or lower-range theories could supplement the general one in explaining sub-systemic phenomena. Irrespective of whether we call it a classification, a framework or a theory, Sartori's contribution remains a fundamental milestone in the study of parties and party systems.