Preferential voting and the party–electorate relationship in Slovakia

In this article we use concentration of electoral support for individual candidates and divergence between party and electorate preferences for candidates to test hypotheses about the party–electorate relationship. We test these using data from preferential voting in Slovak general elections between 1998 and 2010. Our results suggest that low concentration is associated with parties based on ideology and high concentration with parties based on leadership. Age of the party fails to predict the concentration or the divergence. For coalitions, type of coalition matters in regard to divergence. Furthermore, we document that divergence increases with the size of the electoral support, though this seems to be true only for larger parties.


Introduction
The concept of 'preferential voting' encompasses a variety of electoral rules aimed at combining party-based electoral systems with the ability of voters to have their say in the election of individual members of a legislature (Millard and Popescu, 2004b). Originally introduced in Western Europe to foster links between voters and representatives (Bergman et al., 2003: 213;Marsh, 1985: 365;Pedersen, 1966: 172), the mechanism was adopted by five postcommunist countries -Poland, Estonia, Latvia, the Czech Republic and Slovakia -as part of their democratization reforms with the same rationale (Birch et al., 2002).
Two fundamental questions about preferential voting are: (1) where do votes go, i.e. how big is the concentration of preference votes within the party list, and (2) how big is the mismatch between the party's and the electorate's preferences, i.e. how big is the divergence between the party-ordered and the voters-ordered ballot. The literature suggests several determinants of the concentration and the divergence of preferential voting. The most prominent factors include leadership (Lijphart, 1974;Marsh, 1985;Millard and Popescu, 2004a), unity (Katz, 1980;Millard and Popescu, 2004a;Wildgen, 1985) and age of the party (Millard and Popescu, 2004a).
The main aim of this article is to problematize these hypotheses. We use a theoretical framework based on the idea that concentration and divergence cannot be studied separately, but are interrelated. The proposed framework is examined using Slovak preferential voting data.
There are three reasons why we are interested in the case of Slovakia in particular. First, as Millard and Popescu (2004a) document, Slovak voters proved the most willing of the five post-communist countries to alter their parties' candidate order. We would therefore like to explore whether the high mismatch between party and voters' preferences is universal across all Slovak parties, and whether it changes over time.
Second, a characteristic feature of the Slovak party system is that it combines several relatively stable parties with frequent emergence of new and sometimes short-lived parties (Harris, 2010: 188). This enables us to test the relationship between the preferential voting and the age of the party. Furthermore, the emergence of new parties is common in East European countries (Sikk, 2011); Slovakia can thus serve as a representative case.
Finally, our framework implicitly requires data that capture both party and voter preferences. The Slovak electoral system meets these criteria.
The rest of the article is organized as follows. First, we problematize the hypotheses about the preferential voting derived from the existing literature and propose an alternative approach to the study of preferential voting. We then discuss the methodology, and, in the next section, present our findings and test the framework we adopt based on Slovak preferential voting data. The final section summarizes findings and puts them into comparative perspective.

Preferential voting and the party-electorate relationship
The literature offers several hypotheses regarding preferential voting. Leadership, unity/factionalism and the age of the party are the most commonly discussed factors that influence the concentration of electoral support for specific candidates and divergence between preferences of the electorate and the party.

The role of leadership
Leaders' influence as a factor affecting individual's voting decisions has been contested. Some argue that leadership perceptions play an important role in voter preference and choice of a candidate (Maurer et al., 1993;Pillai and Williams, 1998;Shamir, 1994), others point to the leadership effects as weak when other factors, such as sociodemographic variables, values and economic perceptions, are taken into account (King, 2002). This contradiction can be resolved by pointing out that whereas candidates (and issues) are considered short-term factors in determining voters' choice, party is a more powerful predictor in the long-run (Niemi and Weisberg, 1993: 137).
Preferential voting provides further insight into this debate, as it opens up space for personalization of electoral competition. Voters' support tends to be concentrated on incumbent candidates and party leaders (Marsh, 1985: 374). This may be understood in terms of limits on citizens' information absorption (Kinder and Sears, 1985). As Richardson (1988) pointed out using Japan as a case, most voters have information about perhaps one or two candidates. In the case of incumbents, visibility may not be the only factor. For example, by focusing attention on the prime minister as the individual who is accountable for the government's collective performance, the voter finds it easier to deliver reward (or punishment), when compared to an abstract collectivity (McAllister, 2007: 7).
Higher concentration in the case of leader parties may also be related to name-order effects, as leaders usually top the ballot. The position bias, i.e. the notion that election results are influenced by the order in which candidates' names appear on the ballot, was documented for Ireland (Robson and Walsh, 1974), the United Kingdom (Upton and Brook, 1974) and for the United States (Koppell and Steen, 2004;Taebel, 1975).
In general, the more powerful and important the role of the leadership in a party, the weaker its ideology and administration (Katz and Mair, 1995;Panebianco, 1988;Webb, 2004). Thus, we expect to observe higher concentration of preference votes in the case of leader-based parties, and lower concentration in ideological parties, such as ethnic or religious parties, where the collective party identity and common goals are stronger than the personas that attempt to deliver these goals.
Hypothesis 1: Leader-based parties are associated with high concentration of preference votes; ideology-based parties are associated with low concentration of preference votes.
Effects of leadership on the divergence of preference votes are unclear. On the one hand, followers of a charismatic or transformational leader are willing to 'transcend self-interests for the sake of the collective' (Howell and Shamir, 2005, 99). Transformational leadership should therefore be associated with lower divergence, as transformational leaders succeed in bringing the preferences of their voters closer to their own.
However, leadership is determinant on the context. The organization and managerial research suggests there may be multiple collective identities within organizations, representing different and perhaps competing values, sources of identification and preferences (Shamir et al., 1998). Furthermore, a range of more or less fluid identities tends to be associated with large companies (Alvesson, 2000;Ashforth and Mael, 1989). Based on this, it is possible to hypothesize that the size of the political party will matter with regard to divergence. The larger the party, the more likely it is that its electorate will consist of groups of voters with competing preferences.
Hypothesis 2: Transformational leadership is associated with lower divergence. However, its effects might be mitigated by the size of the party: the larger the party, the larger the divergence.

The role of factionalism
Preferential voting can be understood as 'an election within each party' (Katz and Bardi, 1980: 102). It enables voters to take part in intra-party conflicts and support candidates of their preferred faction (Marsh, 1985: 367). Voters are thus more likely to disrupt party list orders in case of parties with competing factions or electoral coalitions (see Katz and Bardi, 1980;Marsh, 1985: 371;Millard and Popescu, 2004a;Wildgen, 1985). The higher mismatch between the preferences of voters on the one hand and those of electoral coalitions or parties with competing factions on the other has been documented for Italy (Katz and Bardi, 1980;Wildgen, 1985) and post-communist countries (Millard and Popescu, 2004a). However, interactions between factions, host parties and voters are dynamic (Belloni and Beller, 1978). As Boucek (2009) shows, factionalism may take different forms in different parties at different times. Boucek defines these forms as threefold: cooperative, competitive and degenerative factionalism. Cooperative factionalism provides a structure of cooperation between separate intra-party groups. It is a tool of consensus building. It aggregates separate groups and blurs cleavages. In this case, it is reasonable to assume lower rather than higher mismatch between a party's and voters' preferences. In contrast, competitive factionalism indicates fragmentation and splits. This type of factionalism diffuses conflict internally, widens voter choice and empowers party followers. Competitive factionalism is thus close to the view that factionalism will lead to a higher divergence between the preferences of the party and the voters.
Hypothesis 3: Parties or coalitions with openly competing factions have higher divergence. Millard and Popescu (2004a: 8, 12, 27) hypothesize that voters of new parties tend to be less informed about candidates and therefore their support concentrates on the few candidates with high visibility. In contrast, the relationship between the age of the party and the divergence is less clear. On the one hand, the divergence should be lower in the case of older parties, which over time learn about their voters and are able to match their preferences better (Millard and Popescu, 2004a: 8). Voters of new parties, on the other hand, lack information about the candidates; moreover, these parties are often leader-based. Therefore, they should be associated with lower divergence as voters defer to the party choice (ibid., pp. 8, 12).

Role of the party age
However, empirical studies provide mixed evidence in support of this hypothesis. The results of Millard and Popescu (ibid.) suggest that voters of new parties do not overall vote more often for first-placed candidates (the case  of Poland, Estonia and Latvia). These findings are not surprising, especially in the light of the recent research on new parties (see Lucardie, 2000 andSikk, 2011). Although it is true that new parties in Eastern Europe are often founded 'from the top down' by regional or national government leaders who are well known and have the necessary resources (Lucardie, 2000: 179), the new parties may compete along two dimensions (Sikk, 2011). First, parties may have a strong ideological motivation, i.e. they will be committed to a set of cohesive and comprehensive policies which they wish to pursue when in government. 'Prophets' (Lucardie, 2000 andSikk, 2011) are close to the traditional view of the role of new parties as that of representing new issues inadequately represented by established parties (Harmel and Robertson, 1985;Hauss and Rayside, 1978;Hug, 1996Hug, , 2001Kitschelt, 1988Kitschelt, , 1995. In contrast to the 'prophets', some parties arise to cleanse an ideology already represented by existing parties; Lucardie (2000) and Sikk (2011) call these 'purifiers'. In regard to Hypothesis 1, then, we should expect a low concentration of preferential votes among new parties. However, parties' ideological motivation to enter the electoral competition may also be low. But whereas 'prolocutors' usually address a particular issue that is disregarded by established parties, parties based on the 'project of newness' seek to change just the manners of doing politics rather than its contents (Sikk, 2011: 3). And because the emphasis on newness itself has little potential for mobilizing voters, often the newness comes in a bundle with the charismatic leader or considerable financial resources (ibid., p. 15). According to Hypothesis 1, therefore, we may expect a higher concentration of preferential voting in parties based on the project of newness.
Hypothesis 4: New parties tend to have higher concentration. Predictions regarding divergence are mixed.
Based on the two dimensions of concentration and divergence, we expect parties to fall within four categories: Category 1. Non-responsive party (low concentration, high divergence). The combination of low concentration and high divergence is typical of the depersonalized party in which the absence of the charismatic leader is substituted by party identity. This identity is strong enough to compensate for the apparent inability or unwillingness of the party leaders to match the preferences of their voters. Competing factions may play the role as well.
Category 2. Leaders and followers (high concentration and high divergence). The party of leaders and followers is one where selection and ordering of candidates differs from what the electorate would prefer, and the electorate itself has a concentrated set of personal preferences. As discussed above, the high divergence may be associated with the size of the party, where a charismatic leader ensures a broad group of followers encompassing competing identities and preferences. Alternatively, it is reasonable to assume that the leader attempts to capitalize on his charismatic appeal and tries to convert his electability into an ordering that does not reflect the voters' preferences.
Category 3. Black box/responsive (low concentration and low divergence). The third type is a party where selection and ordering of candidates is consistent with the views of the party electorate but the party electorate has diffused its preferences among a relatively large number of candidates. There are two competing explanations for this situation -either the party identity (distinct from any identity imprinted by individuals) is so strong that voters give no preference to individual candidates, or, on the contrary, the party works as an alliance of a large number of factions and their representatives, ordered in line with electoral preferences. Such parties could be called 'black-box party', since the insides of the party list are only marginally important to the electorate.
Category 4. Coterie of the attractive (high concentration and low divergence). In this final case, selection and ordering of candidates is consistent with preference of the party electorate and the preferences of the party electorate are relatively concentrated among a few individuals. Thus, we expect to see a strong leadership that responds to the voters' preferences when ordering the ballot.

Methodology
We test our framework and the hypotheses using data from preferential voting in Slovak general elections 1998and 2010. Since 1998, Slovakia has been a single electoral constituency with a weak form of preferential voting. Voters can choose only one party list. However, they are able to modify the party-structured list by using 4 (or fewer) 'preference votes'. In 1998 and 2002, a candidate with more than 10 percent of preferential votes was given priority and assigned the first place on the list. If more candidates achieved this threshold, they were ranked according to the number of their preference votes. In the 2006 election, the threshold was reduced to 3 percent. Such a setting allows us to gather data for both party ballot ranking and preferential voting. Because Slovakia is a single constituency, the data also allow conclusions to be drawn for the country as a whole.
The data samples we use cover all parties elected to the 150-member National Council of the Slovak Republic since 1998, when Slovakia became a single constituency.
To measure the concentration and the divergence, we rely on two quantitative indicators: index of concentration and index of divergence.
Index of concentration. We use an index of concentration as a quantitative measure for the concentration of preference votes within a party list in a given election. We modelled the index on the Herfindahl-Hirschman index, which is defined as the share of preference votes received by each candidate in the sum of preference votes received by all candidates on a particular party list. In other words, if all voters used just one preference vote and gave it all to the same candidate, HHI would be 1. If voters split the preference votes evenly between all candidates, HHI would be close to 0. This index then allows us to measure with a single metric how concentrated the popularity is of the candidates among the party electorate. This measure is an improvement over previous studies, which measured the support for first-placed (or highly placed) candidates only (see, for example, Millard and Popescu (2004a) who focus on the percentage of votes received by the top three candidates at most).
Index of divergence. This index creates a metric for measuring the mismatch between the party's and voters' preferences. To measure the divergence between electoral preferences of the party and those of its electorate, two lists are compared. One is the ordering of the list as submitted by the party for the election. The other is the ordering which would result purely from preference votes. We use the standard deviation -sum of the squared differences between the two orderings -and call it the 'index of divergence'. The importance of divergence differs between various parts of the list, with declining importance towards the end of the list. Therefore, in our analysis we use the divergence between the first 30 places on the list and their ranking by preference votes. The number was chosen because parties represented in the parliament generally reach fewer than or only slightly more than 30 seats. Such a setting enables us to measure how much the voters disrupted the party order in places deemed electable by the party leaders (and the voters themselves).
Both indices are normalized to a 0-100 scale for easier comparison. We calculate average values for parties whose concentration and divergence levels were stable across time. Separate values are listed for parties whose concentration and divergence levels changed over time. These parties enable us to examine the specific conditions under which the party-electorate relationship shifts.
When analysing the results, one should bear in mind that type of electoral system affects the use and impact of preferential voting (Millard and Popescu, 2004a, b;Ortega, 2006). Although Visegrad Four countries (except for Hungary) introduced preferential voting to create links between the voters and the parties, the preference systems themselves were different. In the Polish open-list system, voters obligatorily chose a single candidate. Although they had less choice, their choices made greater difference than in the optional preference systems of the Czech Republic and Slovakia. There, voters could choose up to four candidates, and a candidate's preference votes were taken into account only if at least 10 percent of party voters in a constituency expressed a preference for the candidate. The introduction of the single constituency in Slovakia in 1998 made it even more difficult for locally respected candidates to gain the requisite share of preference votes (Millard and Popescu, 2004a: 9). We discuss our findings from the comparative perspective in the concluding section. Particular attention is paid to the Czech Republic with a weak system of preferential voting similar to that of Slovakia.

Findings: Slovak case study
In this section, we test the hypotheses presented earlier using methodology from the previous section on a case study of four Slovak parliamentary elections since 1998. Table 3 documents that the use of preferential votes is high and growing despite the fact that it has minimal effect on the actual composition of the parliament. Since 1998, we have seen a gradual increase in the number of voters using the option to cast preference votes, i.e. increasing from 54.1 percent of voters in 1998 to 77.5 percent in 2006. The year 2010 witnessed a slight decline to 73.6 percent.
Preferential voting, in Slovakia in particular, is popular because it gives the aura of choice and offers the safety valve of a potential voters' revolt on behalf of particular candidates or groups of candidates (Millard and Popescu, 2004b: 8, 19). At the same time, we would expect the utilization of votes to increase along with increasing 'effectiveness' (Pederson, 1966) of preferential voting, i.e. the impact of the preferential voting on list order. The number of MPs elected through the mechanism of preferential voting began to increase only in 2006, when the threshold the candidate had to achieve for his preference votes to count fell from 10 percent to 3 percent of all the votes cast for the party (Table 4).
In the 2010 election there were 12 MPs who would not have been elected had they not received the minimum number of preference votes. This represented 8 percent of all MPs elected. However, under the old rules the number of MPs would have equalled a single MP. This contradicts findings summarized by Marsh (1985) that utilization of preference votes is directly related to relevance, i.e. that the preferential voting has an effective impact on list order. At the same time, number of preference votes cast by each voter utilizing the mechanism has decreased slightly, indicating that voters are, on average, becoming more strategic in how they use the votes. Second, our findings suggest that the party-electoral relationship indicated by patterns of preferential voting varies widely across the party system, but for most parties the findings are relatively stable over time. The majority of parties fall within the lower cells of the proposed typology, which is characterized by low divergence (see Figure 1). In other words, the mismatch between the party and the electorate preferences is relatively low. However, the casual relationship between the party's and voters' preferences is not clear. It is either that the party responds to its voters or voters accept the preferences of the party.
Looking at the measures of concentration and divergence separately, we find the following. First, our findings support Hypothesis 1 that low concentration is associated with parties based on ideology and high concentration of support valid for parties based on strong leaders 'owning the parties', as evidenced by differentials in the rate of change of party leaders between the two groups.
There were six parties that experienced a low level of concentration (and two borderline cases) and two of these were 'coalition' parties created for the 'critical' elections in September 1998 (Harris, 2010: 189). A broad electoral coalition of five parties was formed in 1997 to challenge Vladimír Mečiar's authoritarian rule and ensure integration into the Western clubs of EU and NATO. When Mečiar responded by controversial electoral law restricting the candidature of such coalitions, the coalition was transformed into a party under the name SDK (Slovak Democratic Coalition). In line with our hypothesis, the coalition leader Mikuláš Dzurinda 'was initially seen as weak' (Haughton, 2005: 134) and the coalition rested more on its pro-democratic and pro-integration programme than on the personal appeal of the leader. Following the same electoral law, the former electoral coalition of three small parties representing Hungarian minority merged into one party (SMK). SMK gradually evolved into a regular ethnic-based party; SDK, in contrast, split into two parties (KDH and SDKÚ ) by the next election. In 2010, SMK was replaced in parliament by Most-Híd, formed by several SMK MPs after an internal struggle. Given these two parties achieved similar levels of concentration and divergence, and the fact they compete along similar agenda, the average was calculated for both these parties.
Other parties with low concentration included four strongly ideological parties. The party with the lowest concentration was KSS, a Communist successor party in terms of ideology and personnel, which promoted state ownership and rejected capitalism and European integration (Rybář  and Deegan- Krause, 2008: 201 f.). Second, the Christian Democratic Movement (KDH) was a centre-right party liberal on economic policy and highly conservative on social and moral issues (Fitzmaurice, 2004: 165;Kopeček, 2007: 325), with a limited emphasis on nationalism. Over the course of its existence, the party was successfully depersonalized (ibid.) and its voters became deeply and often lastingly identified with the party. This is documented by two smooth leader replacements in 2002 and 2009 and stable electoral results. Finally, two new parties with low concentration entered the Parliament in 2010 -the civil liberal SaS (see below) and the ethnic-based Most. The low concentration is thus associated with parties which are not leader-based. All these parties changed their leaders at least once in their history. (Exceptions are SDK, which split into SDKÚ and KDH, and the new party SaS (Freedom and Solidarity), which contested only the latest 2010 election.) Looking at the common denominator of these parties it is possible to argue that the low concentration is associated with parties based on ideology (prodemocratic, Christian, ethnic/minority, neoliberal) rather than on leading figures.
High concentration, on the other hand, is found in the case of six parties which share strong personalization. These parties have not changed their leaders, or, if they had, it would have led to electoral disaster.
HZDS was a dominant party on the Slovak political scene between 1992 and 2002 (see Haughton, 2001 andBaer, 2001), with its leader Vladimír Mečiar becoming Prime Minister three times. The party was personalitybased (Day, 2002: 413), combining the personal appeal of its leader with a broad appeal of managed economic reform and a concern for those who lost out from marketization. This agenda was moreover articulated in a national accent and targeted also at Christian voters (Haughton, 2005: 133).
The nationalist SNS has projected itself as the party of ethnic Slovaks (Haughton, 2001: 747). It is highly conservative on social issues (Kopeček 2007: 441) but economically leans more towards the left. Similar to HZDS, it revolves around the polarizing leader Ján Slota.
Finally, in the group of parties with high concentration, Smer has been the strongest party in Slovakia since 2006, as it gradually took over the position of HZDS. It was created by Robert Fico in 1999 on the basis of transcending the existing cleavages. As we discuss below, this concept of 'newness' (Sikk, 2011) was central also in the case of other new parties, namely SOP and ANO.
An apparently unlikely member of the group is SDKÚ , Slovak Democratic and Christian Union. The party was a successor to the SDK coalition and was created before the 2002 election as a result of the growing frustration of Prime Minister Dzurinda and a number of senior colleagues over the persisting ties of those elected on SDK's list to their mother parties (Haughton, 2005: 135). Approximately half of the SDK parliamentary club and all SDK ministers went over to SDKÚ . Although Dzurinda was initially a weak leader, he gradually asserted himself and became more confident in his second term (Haughton 2005: 134). But although the attachment to the personality of Dzurinda was important, the potential of the party was based mostly on the legacy of SDK (Kopeček, 2004) and the bold programme of economic reforms. Furthermore, in terms of personnel, SDKÚ was not a 'one-man party' (ibid.); in fact, the concentration just for the top name on the list would put this party in the low concentration camp. In this sense, it is different from the other parties with high concentration and unique in combining high concentration of support for the top team with low concentration of support for the individual leader.
Although it has been suggested that new parties are usually leader-based and therefore with higher concentration, Slovak data do not fully support this hypothesis. A number of new parties have been formed in Slovakia (SOP, ANO, Smer, SaS) which have not formed a single cluster. The first three entered the Slovak political scene sharing high concentration and low divergence. These three can be characterized as parties based on the project of newness (Sikk, 2011: 15). They promoted the cause of change but were vague on specific policy commitments.
SOP (Party of Civic Understanding), which entered Parliament in 1998, was a catch-all populist party formed by the popular Mayor of Košice, Rudolf Schuster (Fitzmaurice, 1999). After Schuster was elected president in 1999, the party became virtually leaderless (Fitzmaurice, 2004). It did not cross the 5 percent threshold in the 2002 election, and in 2003 it merged with Smer.
The 2002 election saw two new parties based on the project of newness: ANO, based around the media mogul Pavol Rusko, and Smer, based around the charismatic Robert Fico. Both parties attempted to offer an alternative to the existing parties. Survey evidence shows that whereas ANO provided a popular alternative for rightwing voters dissatisfied with Dzurinda, left-wing voters opted for Smer (Krause, 2003: 70). But whereas ANO disappeared, mostly a result of financial scandals involving its leader, Smer turned out to be the most successful representative of the party based on the project of newness (Sikk, 2011: 15). The party originated as a populist centrist party (Harris, 2010: 77) targeting voters weary of polarization between the authoritarian and prodemocratic forces (Rybář and Deegan-Krause, 2008: 502). After the disappointing performance in the 2002 election and the failure to enter the government, Smer made substantial changes in the social and economic programme of the party (Marušiak, 2006). This move away from the empty concept of the 'newness' towards a concrete set of policies was probably the main reason behind Smer's successful transformation into a regular party, in contrast to the short-lived SOP and ANO. SOP, ANO and Smer share several common features. Their original ideological motivation was rather low and parties served mainly as locomotives to posts in the executive. All parties competed on the basis of 'newness', an alternative to existing parties. The lack of ideology in combination with popular leaders is in line with hypothesis 1: all examined parties achieved high levels of concentration, and all scored low on divergence as a result of a number of candidates less known to the public. Indeed, the introduction of 'new faces' was emphasized by Smer and its leader Fico in the 2002 election campaign.
However, as discussed earlier, the new parties cannot be associated solely with leader-based parties. SaS is a typical example of a party that offers an ideology not represented by any existing party. SaS attempted to meet the demands of the classical liberal party or civil libertarian party. It advocated freer markets, fiscal conservatism, liberalization of drug laws and same-sex marriage. Its programme, '120 ideas' offered a clear set of policies, and therefore attracted both the right-wing voters interested in increased economic freedom and left-wing voters interested in social-liberal issues (Hospodárske noviny, 3 June 2010). The ideological motivation of the party was also reflected in the concentration, which was remarkably low compared to other new parties, as well as most of the existing parties. Intensive Internet campaigning was another factor that contributed to the wider spread of preference votes across candidates, who used personal blogs and Facebook to increase their visibility.
These results suggest that there are more important characteristics of the party that determine the concentration of support (leader-based versus ideology-based) and divergence (united versus fragmented parties).
Our findings suggest that the type of factionalism matters with regard to divergence: it is the competitive factionalism that is associated with higher divergence. As already discussed, SaS ran on a liberal platform in the 2010 election, but de facto was a coalition, since the last four places on its ballot (147-150) were offered to the Christian conservative group 'Ordinary People', whose political profile differed significantly from the classical liberal agenda of SaS. Despite their position on the ballot, Ordinary People were elected to the Parliament based on the high number of preferential votes. The divergence can thus be attributed to the so-called 'competitive factionalism' (Boucek, 2009), which later translated to conflicts between the factions and eventually an exclusion of the faction leader from the SaS parliamentary club.
However, coalition is not always a clear-cut determinant of divergence. Our results suggest that what matters more is the type of coalition rather than the coalition per se. Other electoral coalitions -SMK and SDK in 1998 and Most in 2010 (which similarly to SaS offered four positions on its ballot to members of the other party) -experienced much lower levels of divergence. As already discussed, both SDK and SMK emerged in response to changes in the electoral law. Each of the coalitions was held together by the shared goals. The parties forming SDK aimed to challenge Mečiar, therefore the main concern was to show unity and its capacity to become the largest party (Fitzmaurice, 1999: 294). SDK thus served as a tool for aggregation of parties from different poles of the political spectre without threatening their identities. SMK was motivated by the protection of ethnic and minority rights. In contrast to SDK, which later split into two parties, the effect of 'cooperative factionalism' on SMK was more profound. The coalition of three Hungarian parties motivated by the protection of ethnic and minority rights evolved into an integrated party.
A border-line case is KDH competing on a Christian platform whose divergence increased over time. The increase in 2006 can be attributed to competing factions, which led to fragmentation of the party and the departure of four key members of KDH two years later. This departure was probably reflected in the further increase of divergence: as the departed members were key figures in the party, the support for these members spread over other candidates.
Finally, we document that divergence increases with the increasing size of the party. High concentration and high divergence were documented for only two parties -Vladimír Mečiar's Movement for Democratic Slovakia (HZDS), which dominated the Slovak political scene for a decade between 1992 and 2002, and Robert Fico's Smer, which won the 2006 and 2010 elections. Both parties are based around strong, charismatic leaders. The high divergence associated with charismatic leaders does not support hypothesis 2, namely that charismatic leadership decreases divergence. Furthermore, divergence levels of both parties shifted significantly over time. This enables us to examine the conditions under which parties move from one box of the two-by-two to another.
First, it should be noted that both HZDS and Smer moved to the upper-right box characterized by high concentration and high divergence at the point when they both reached more than 25 percent of the vote and thus became large parties by Slovak standards (generally, the vote ranges between 5 and 15 percent support per party). Figure 2 documents the positive relationship between the electoral result (percentage of votes) and the level of divergence. As HZDS decreased in size, so did its divergence. Similarly, as Smer began to dominate the Slovak political scene, its divergence rocketed upwards. As HZDS and Smer are similar in appeal (populist and nationalistic), we wanted to control for the party-type effects by including a party from the other part of the political spectrum. Therefore, we include the coalition SDK, the only party whose electoral support could match that of HZDS and Smer. Its position in the graph fits nicely into the exponential curve. These findings support the view that the large parties encompass groups of voters with competing preferences. However, the fact that the SDK was a coalition may play some role in determining its divergence.
Furthermore, the ballots of HZDS and Smer seem to be a combination of very popular figures at the top of the list and then an ordering which places a number of candidates with very low electoral appeal after the electoral favourites at the top. Once the party leaders expect their electoral results to be high, they are more willing to order their ballot in a way that does not reflect the preferences of their voters. Party leadership can convert its electability into an ordering that reflects its other preferences without incurring a significant electoral penalty. This fact does not turn voters off preference voting, nor does it stimulate it -in both instances, the utilization of electoral votes was around the average.

Conclusions
The main aim of this article was to examine the party-electorate relationship using preferential voting data. To this end, we used a concentration of voters' support and divergence between party and electorate preferences as two inter-related dimensions. Using this framework to examine preferential voting in Slovakia enabled us to test the existing hypotheses, which suggest certain levels of concentration and divergence based on the unity, age or leadership of the party; furthermore, it led us to examine the circumstances under which parties' relationship with the electorate changes. We come up with seven distinct findings.
First, the option to indicate preference for individual candidates in Slovakia is popular and increasingly so despite the fact that it has limited effect on the actual composition of parliament. This contradicts findings from the literature which linked utilization to relevance, i.e. the effective impact of preferential voting on list order (Pederson, 1966). This is in contrast to the situation in the Czech Republic, where the increase in the number of preference votes from 2 to 4 and the reduction in the threshold from 7 to 5 percent, adopted in October 2006, had a profound impact on the number of candidates through preferential voting. A total of 13 election leaders of five parties were not elected as a result of preferential voting in the 2010 election despite topping their party list in a given region (Týden 31 May 2011). This can be attributed to the fact that a higher number of electoral districts (14 in the Czech Republic compared to 1 in the Slovak Republic) effectively allows not just to promote lower-ranked candidates, but to promote sufficient number of them to de-select a leader of the list in a given district.
Second, the patterns of party-electorate relationship in Slovakia indicated by concentration and divergence of support differ widely across parties, giving rise to the following set of conclusions about individual parties. First, we observe a higher concentration in the case of leader-based parties and lower concentration in the case of parties with collective identity, where the presence of a charismatic leader is usually substituted by a clear ideological agenda (pro-democratic, Christian, ethnic/minority, neoliberal, communist). These findings were shown to hold also in the case of the Czech Republic. Similar to the Slovak Communist Party (KSS), leaders of the Communist Party of Bohemia and Moravia (KSČ M) gained the same or lower percentage of preference votes than other candidates on the party ballot in the 2005 Czech election (Voda and Pink, 2009: 174). The Czech Green Party was another example of an ideologically based party with low concentration (Voda and Pink, 2009: 177). The highest number of preferential votes went to the leaders of the Civic Democratic Party (ODS) and Czech Social Democratic Party (Č SSD) (Cevro Revue, 2006). These two parties were characterized by a single leader who came to personify the party but remained responsive to other party officials (Deegan- Krause, 2006: 90). In contrast, Polish voters seemed less enthusiastic about their party's list-leader: by 2001 almost 40 percent of first-placed candidates received less than 25 percent of the vote, and less than 10 percent of firstplaced candidates received over half the vote (Millard and Popescu, 2004a: 17 f.). However, the results might have been influenced by the length of the list (ibid.).
Third, age of the party does not determine the concentration. New parties whose main appeal rested on popular leaders and celebrities experienced high concentration. On the contrary, new parties that competed along ideological lines experienced low concentration. Thus, leadership was a more powerful predictor than age of the party. The above-mentioned case of the new Czech Green Party supports this view: absence of a strong leader and introduction of a number of independent candidates from the arts sphere and civic organizations resulted in the spread of preferential votes over the ballot (Cevro Revue, 2006;Voda and Pink, 2009: 178). Similarly, older and stable parties did not have lower divergence between voters' and party's preferences, as suggested by the literature.
Fourth, in line with the literature, we observe higher divergence in the case of parties with competing factions. However, factionalism per se is not predictive of divergence; electoral coalitions achieve both low and high divergence. What matters, therefore, is type of factionalism. Divergence tends to be highest in the case of parties that are de facto coalitions -with factions with clearly identifiable and competing programmes. This effect was strong also in the 2002 Czech election: out of eleven deputies elected by preference votes, ten came from one coalition list that brought together four parties (Millard and Popescu, 2004a: 22 f.).
Fifth, for the majority of parties, the relationship between them and their electorate is relatively stable. Most fall within the bottom categories of the proposed typology, which is characterized by low divergence. Major shifts in the relationship are documented for parties whose size and electoral strength change markedly over time. In Slovakia, the threshold is approximately 25 percent of all votes cast. This is true especially in the case of two parties with charismatic leaders that used to dominate the Slovak political scene. We dub these parties 'coalition of leaders and followers' and argue that charismatic leaders convert their electability into an ordering that does not reflect their voters' preferences without incurring a significant electoral penalty.
Sixth, there was a single instance of a party with high divergence which was neither a coalition nor a large party with a charismatic leader -the Christian Democratic Movement (KDH). In this case, we explain the increasing divergence by fragmentation of the party and by departure of key members of the party. We believe that an increase in divergence was the result of the spread of their preferential votes among other candidates in the next election.
Seventh, only a minority of parties in Slovakia achieved high levels of divergence relative to other parties. Thus, most of the parties seem to be on a par with each other in matching their voters' preferences. High divergence was thus achieved only with three types of parties: (1) coalitions with competing factions, (2) parties of leaders and followers, large in size and organized around charismatic leaders, and (3) a single exception -a party whose key members left and the voters most probably spread their support among the remaining candidates, hence the high divergence.