0.0 2.5 5.0 7.5
2.5 5.0 7.5 10.0
Fragmentation
Polarisation, alternative PW measure
−1 0 1 2
1 2 3 4
Additional parties
Change in polarisation, alternative PW measure
Figure 5.3: Polarisation, Fragmentation and an Alternative Pairwise Measure.The figure shows how an alternative pairwise measure that is suggested here performs in the above mentioned simulations. The box-plot shows the inter-quartile range (median in the middle) and the whiskers the largest/smallest value no further than 1.5 of the inter-inter-quartile range from the edge of the box.
clearly has more desirable properties than the alternatives. The estimate of polarisation does seem to go up as additional parties are added to the system, but this increase is much weaker than for the other two indices. And it should be noted that it does have a problem with heteroskedasticity – the possible values of polarisation have a much higher variance, especially if fragmentation is at or below two parties. This is also a problem for the other two measures. However, such low fragmentation rarely exists in the party systems that we are looking and thus this cannot be a major problem here. Furthermore, unlike the ideological standard deviation and the Esteban and Ray measures, this proposed measure has a rather simple interpretation – it is the weighted average of the mean pairwise distance of each party from all other parties.2
with an emphasis on supply side phenomena – other aspects of parties and party interaction, as well as the broader socio-economic context. There is a whole strand of research, which has studied po-larisation in the context of individual level voting behaviour. The following will give a brief overview of the main findings from that research direction, but the focus in this comparison will be on other aspects of the party arena (supply side) and its socio-political context.
Much of what follows below are echoes of Downs and Sartori. In their account of political spaces and party competition one can find, in addition to the quite straightforward associations with the number of parties and types electoral systems, hints about the political or ideological distance between parties being related to certain nuances of voting behaviour, as well as more broader phenomena, like social inequality or the quality and viability of democracy. And indeed, many of the general associations that they were either non- (Downs) or half-empirically (Sartori) speculating about, ended up being borne out by some systematic empirical evidence (but not by others).
5.2.1 Fragmentation and the Electoral System
I will not repeat the non-empirical arguments of Downs (see Downs 1957, pp. 115, 124) and Sartori (see Sartori 2005, pp. 307-312) with regard to fragmentation, the electoral system and polarisation that were already mentioned above. Suffice to say here that this idea – different electoral systems condition the number of parties and their political dispersion in different ways – was recurring also in other early contemporary political science writings (see e.g. Duverger 1967). The basic idea was always the same: proportional as opposed to majoritarian electoral systems go together with a higher number of parties and higher amount of aggregate distance between them. This association has also been illustrated by formal modelling (Cox 1990; McGann, Koetzle, and Grofman 2002; Merrill and Adams 2002), suggesting that ideological dispersion is associated with district magnitude and/or the number of parties in the system.
Empirical research into these questions has provided some clarity, but with qualifications and without ultimate certainty. In one of the earliest empirical studies into the topic of the number of parties and polarisation, Gross and Sigelman (1984), using their pairwise measure, suggested that even though conceptually fractionalisation and polarisation are separate dimensions, they tend to be empirically highly correlated.3 In a later study Andrews and Money (2009), using a measure based on the distances between the extreme parties in a two dimensional political space, have also
3 Which is not surprising as their measure, akin to the Esteban and Ray measure, is also by definition correlated with the number of parties.
CEUeTDCollection
suggested an empirical association between fragmentation and polarisation, but also indicating that the association might top out at a certain level of fragmentation.
Closely related to the concept of fragmentation is the proportionality of electoral systems. It is usually the case that the latter influences how many parties can exist in a party system (see Taagepera 2007). Therefore, proportionality has recently received perhaps even more attention in polarisation research than fragmentation and despite the fact that these two are related, they have often been considered separately. For example Dalton (2008), using a version of the ideological standard devia-tion measure, claims that there is no bivariate associadevia-tion between polarisadevia-tion and fragmentadevia-tion, but that there is one with district magnitude, an important indicator of disproportionality.4 Also Dow (2011), using the measure of party system compactness by Alvarez and Nagler (2004), has shown that ideological dispersion and disproportionality (district magnitude and effective threshold) are related to each other.
In contrast, Ezrow (2008), using various kinds of data sources for the political positions of parties and the party system compactness measure, reports no association between the proportionality of the electoral system and policy extremism as well as no association with the number of parties. No association between polarisation measured as ideological standard deviation and fragmentation or the number of parties has recently been claimed also in the Latin-American context (Singer 2016).
Perhaps the two most thorough studies focussing on fragmentation or disproportionality and polarisation have been by Curini and Hino (2012) and Matakos, Troumpounis, and Xefteris (2015).
Curini and Hino (2012) use the Comparative Study of Electoral Systems survey data to test if there is a relationship between polarisation (using the ideological standard deviation measure) and the number of parties that is dependent on coalition expectations. The logic of this association is the following. A larger number of parties is related to the likelihood of forming a coalition government. In such conditions, parties are incentivised to be more moderate in order to increase their coalition potential. They postulate an interaction with coalition/minority cabinet formation and party system fragmentation, measured by the effective number of parties (Laakso and Taagepera 1979). They control for disproportionality using the logarithm of district magnitude and measure coalition habits with a dummy variable indicating whether coalitions were formed in the past. They also look into whether other elections and certain characteristics of the electorate are associated
4 District magnitude, the number of representatives elected from a district, determines what is called the effective threshold (Taagepera 1998). The lower the district magnitude, the higher the proportion of votes a party would need to get elected and the higher the likelihood that the votes given to smaller parties will be “wasted” because they will fail to get elected. Thus, the higher the discord between vote and seat shares, an important symptom of disproportionality (Gallagher 1991).
CEUeTDCollection
with polarisation. Their results show, among other things, that district magnitude and party system fragmentation by themselves are not related to party system polarisation, but associations emerge between fragmentation and polarisation if the model is supplemented with other variables. Most importantly, fragmentation has a positive association with polarisation if there are no coalition habits and a negative association if there are.
Matakos, Troumpounis, and Xefteris (2015) also look into the association between polarisation (using both the standard deviation measure and the distance between extreme parties measure), fragmentation, and disproportionality and add a further nuance to this possible association. They use the RILE index of the manifesto dataset to claim that polarisation is decreasing with disproportinality and increasing with the number of parties, but that the latter effect only holds if polarisation is measured on the basis of the two most extreme parties and not Dalton’s ideological standard deviation measure. In their results, the disproportionality of the electoral system is an important determinant of party system polarisation – the higher the former, the lower the latter. Their study suggests the impact of how polarisation is measured – some fundamental associations might be there if we use one measure and not there if we use another. By implication, this also suggests that looking at the distance between the two most extreme parties and looking at ideological dispersion might mean looking at different phenomena, at least to some extent.
5.2.2 Government Stability
Next to associations with fragmentation, a considerable amount of effort has been devoted to look-ing into the association between polarisation and government stability. The logic here is that more polarised party systems are supposed to be a sign of a more complicated bargaining environment due to political or ideological incompatibilities between parties. This, in turn, undermines the extent to which parties are able to govern together as being in a coalition means having to make political compromises. Indeed, one of the first empirical analyses involving polarisation chose to look system-atically into this very issue. Taylor and Herman (1971) found that ideological divisions, especially the proportion of seats held by anti-system parties, have important implications for government stability.
Numerous later studies have added evidence in favour of this association (Laver and Schofield 1990;
King et al. 1990; Warwick 1994).
CEUeTDCollection
5.2.3 Turnout, Volatility and Ideological Voting
There is an entire tradition of research that has been looking into the associations between voting behaviour and polarisation. To some extent these analyses venture into the domain of individual level data, which is out of bounds for the current comparison. However, many of the aspects considered here are also relevant for the following analysis.
Several analyses suggest an association with turnout, employing the following explanatory nar-rative (see e.g. Crepaz 1990; Dalton 2008; Steiner and Martin 2012). The more parties are distinct from each other in terms of their policies, there more is at stake at elections in terms of who might get the chance to rule. It should thus matter more for citizens who will form the next government and who will not. And thus there should be more motivation for people to participate in elections. But some of the empirical results in this respect are also mixed – for example Aarts and Wessels (2005) in their analysis of six countries, conclude that there are opposite bivariate associations between turnout and polarisation. In some countries it seems to be positive and in some negative. Other empirical evidence about this particular association is scarce.
One can also conceive of an association with polarisation and changing votes between parties on the individual level or electoral volatility on the aggregate level. If parties are closer together, then it should be easier for voters to switch from one party to another as doing so they do not have to jump over an ideological chasm. Indeed, there is research looking at either individual level party switching or aggregate level volatility that has confirmed these associations. Roberts and Wibbels (1999) have shown that polarisation (measured as the dispersion of the voters away from the centre) is negatively associated with electoral volatility in Latin-American parliamentary elections. Tavits (2005) looks at the relationship in the case of Eastern European countries, showing that polarisation (measured as the maximum distance between the government and opposition) is indeed negatively related to volatility also among that set of countries. Dejaeghere and Dassonneville (2015) also look into the individual level aspects of this phenomenon (party switching as something distinct from volatility) across party systems and show that party system polarisation is an important predictor for changing one’s vote across elections.
Furthermore, there is a whole range of research that looks into the relationship between polarisa-tion and ideological voting (van der Eijk, Schmitt, and Binder 2005; Lachat 2008; Singer 2016, e.g.) – the association between voters’ ideological orientations and their party choice. This research seems to be rather unequivocal about the fact that higher polarisation is related to a stronger association between voters’ ideological orientation and their party choice.
CEUeTDCollection
5.2.4 Social Inequality
What has also found more attention in recent analyses into party system polarisation is the possible association with the distribution of wealth in a society. The latter is supposed to determine the distribution of preferences and the latter, in turn, can be related to polarisation as parties adjust their political profiles to this distribution. Increased societal divisions are translated into larger divisions between parties. This relates back directly to what Downs was writing about social heterogeneity, the distribution of voter preferences and ideological variance (Downs 1957, pp. 100-118).
In a study of over sixty elections in 12 OECD countries, Pontusson and Rueda (2008) test the association between social inequality and the positions of left and right parties using the RILE index from the manifesto data set. Their hypothesis is a familiar one – greater inequality changes the preferences of voters, which in turn moves the ideological profiles of parties. However, they add an important qualification. We know that economically disadvantaged people are also less politically mobilised. Thus, the effect of inequality, particularly on parties on the left, should be the greatest when lower income people are more highly mobilised. Using voter turnout and unionisation as proxy measures for lower-class mobilisation, and both wage and overall household inequality (which includes wages, but also government transfers and returns on financial assets) as measures for the distribution of wealth, their results show that parties do react to inequality. At high levels of low income mobilisation, wage inequality is related to left skewed polarisation and at low levels of low income mobilisation household inequality is related to polarisation more towards the right (ibid., p. 346). It should be noted that they do not test the association between inequality and polarisation as such, but between inequality and the locations of different kinds of parties (left or right) on the left-right dimension. Yet, this is very close to the idea of polarisation and is thus also relevant in the context of the current comparison as evidence that polarisation and inequality can be related under certain conditions.
Another recent study, however, looks directly at the association between inequality and polari-sation. Han (2015) probes into the effect of inequality (top income shares) on polarisation (relying primarily on the ideological standard deviation measure, but also testing the results with the Esteban and Ray measure) through its interaction with disproportionality. The overall logic here is the same – inequality causes social polarisation and parties might want to respond to that. But the crucial element here is how permissive the electoral system is. In highly disproportional systems, where a slight loss in votes might induce a major loss in parliamentary representation, parties might be more reluctant to move away from the median voter as that would carry a risk. In their analysis,
dispropor-CEUeTDCollection
tionality is measured by district magnitude and polarisation with the RILE index from the manifesto data set. A range of variables are used as controls – among others, economic growth, unemployment, inflation as well as turnout (as a proxy for low income mobilisation), ethnic fractionalisation, regime type and coalition habits. The results show no direct association between polarisation and income inequality, as well as no effect of disproportionality when interactions are not considered. Supporting the results of Curini and Hino (2012), he reports an interaction between coalition habits and frag-mentation. Most importantly, however, the effect of inequality on polarisation seems to be dependent on district magnitude (disproportionality) and is stronger in more proportional systems. The results therefore seem to indeed show that more permissive electoral systems lead to more polarisation with increasing inequality.
5.2.5 Democracy and Affluence
Next to more or less established empirical associations between polarisation and other macro char-acteristics of political systems there are also a few associations which have been implied and which seem plausible in the context of what we know about the general development of Western political regimes. For example, Klingemann (2005, p. 33) has brought out an argument called “the end of ideology hypothesis”, which relates polarisation to broader macro-sociological developments in Western societies. As the traditional social cleavages have lost their role in structuring the political landscape, a factor that should keep up the distinctness between parties as they are trying to appeal to their core constituencies, polarisation has also eroded. If this is true, then over the last half of the 20th century we should be observing a gradual decrease in the polarisation of party systems.
An association with time, however, would conflict with an other association that has been sug-gested – that with the general level of affluence in societies, as the latter also tends to increase with time and thus its independent effect would be difficult to discern. With regard to the association with affluence, Sigelman and Yough (1978) looked at two contradictory hypothesis about political diversity and affluence. They hypothesised that on the one hand, social diversity and thus political polarisation might be muted in an affluent society. However, on the other hand, it can also be the case that affluence implies tolerance towards diverging points of view and thus allows for higher polarisation. They find support for the latter hypothesis, but not the former. Also, with regard to overall affluence, Pelizzo and Babones (2007) claim, using the historical examples of France and Italy, that worsening economic conditions can be related to increased levels of party polarisation. These two findings are contradictory to each other and there has not been much further research into the
CEUeTDCollection
association between the general wealth of a country and the polarisation of its party system. This association, thus, remains ambiguous.
There is, however, an association, which has been rather explicitly stated and accepted, but which has not been empirically tested at all – one with the level of democracy. This is something that was clearly present in the works of Downs (1957) and Sartori (2005), but which has not made its way into the analysis of polarisation yet. In brief, the argument of Downs was that ideological polarisation will make government policy across governments unstable, which can result in political chaos and possibly even revolution (Downs 1957, p. 120). The argument of Sartori was similar – anti-system parties at the extremes result in a politics of outbidding, which undermines the political life of a country. If the extremes become too strong and overtake the stabilising centre, the political system can fall apart. Therefore, one could expect an association between increased polarisation and lower levels of democracy. With democracy measures abound (e.g. Coppedge et al. 2016b), this is rather straightforward to test empirically.