Abstract
Do different types of preferential-list PR systems create different incentives for how Members of Parliament vote? To examine this, we compare the quasi-list system of Poland, where only preference votes determine which candidates win seats, to the flexible-list system in the Czech Republic, where the 5 percent preference vote threshold required to override the party ranking of candidates gives the party greater power in influencing which candidates become MPs. We analyze roll call votes in the 2007–2011 Sejm and the 2010–2013 Czech Chamber of Deputies and, after controlling for party and MP characteristics, we find that in both countries, MPs with lower preference vote shares are more likely to vote along with their party. But, when we compare the strength of this relationship, we observe substantial differences. The magnitude of this relationship in the Czech Republic is ten times stronger than in Poland, which can be attributed to the more prominent role Czech electoral rules give to the party.
Keywords
Introduction
The type of electoral system used in a country impacts many aspects of political life. While much attention has been paid to the effects on party systems, campaigns, and vote choice, less is known about how different types of electoral systems motivate the legislative behavior of Members of Parliament (MPs). In this article, we examine how electoral system rules influence the voting behavior of MPs in the lower houses of the Czech and Polish Parliaments.
Elections to the Polish Sejm and the Czech Chamber of Deputies use preferential-list proportional representation (PLPR), but with slightly different rules. In both systems, voters must first select the ballot of the party they will vote for. This party ballot lists the candidates in an order determined by the political party. In Poland, voters must select exactly one candidate from their party’s list, whereas in the Czech Republic indicating candidate preferences on the ballot is optional. The important distinction in the rules for our comparative analysis of parliamentary behavior is that in Poland the number of preference votes won by candidates determines who wins seats. The original party ranking plays no direct role. This stands in contrast to the Czech Republic, where the initial party ranking of candidates holds unless a candidate(s) surpasses a 5 percent preference vote threshold of votes for the party in that district. This rule difference means that the party ranking is more important in the Czech Republic in determining which candidates become MPs. Drawing on Shugart’s 1 electoral system typology, the Polish electoral system may be described as “quasi-list,” whereas the Czech system is an example of a “flexible list.”
Since the Czech electoral system gives parties more control over who wins because of the preference vote threshold, does this create stronger incentives for electorally vulnerable MPs to adhere to the party line compared to those in the more candidate-centered Polish system? Or is the opposite the case? Are electorally vulnerable candidates more willing to deviate from the party line to improve their standing in the eyes of their constituents? The two aforementioned questions suggest the existence of different mechanisms working in the opposite directions. The first centers on the candidate nomination process, whereas the second draws on the concepts of the electoral connection 2 and the personal vote. 3 Because of the conflicting incentives that exist in the open-list PR system, the answer to these two questions is not trivial. We solve this puzzle using candidate data from the 2007 Polish and 2010 Czech parliamentary elections and the voting records of the MPs during the subsequent parliamentary term.
First, consistent with the party nomination approach, we hypothesize that MPs who won higher preference vote shares will be less likely to vote with their party in parliament than MPs with lower preference vote shares, while those with lower preference vote shares will show stronger party unity. The logic in this is that MPs with higher preference vote shares will feel less beholden to their party and will exhibit lower party unity, because they have the popular support needed to win seats in future elections. MPs who have lower preference vote shares need to stay in good standing with their party to secure a high ballot rank in the next election as a way to increase their chances of winning reelection. Previous research on Poland and the Czech Republic has shown that voters are more likely to cast preference votes for candidates higher on the ballot. 4 In other words, in Poland, the party candidate ranking has an indirect effect on who wins elections, because it affects how some voters cast their preference votes. In the Czech Republic, the party ranking has a direct effect because it is maintained unless a candidate surpasses the 5 percent preference vote threshold within a district list.
Second, we propose a hypothesis consistent with the electoral connection approach, which suggests the opposite effect of the one proposed by the party nomination hypothesis. According to this explanation, the candidates are primarily driven by their interest in winning the support of their local electorate, leading to more frequent deviation from the party line in parliamentary votes. This mechanism has been identified in the consolidated democracies with a strong personal vote, especially when coupled with a presidential system and weak parties. 5 If this mechanism exists in the cases we examine, we should expect to find particularly low levels of unity among candidates who performed poorly in their districts in the previous election. Conversely, those who performed well are expected to defect less often.
Finally, we consider differences between two countries. If our party nomination hypothesis holds, then we would expect the relationship between preference vote share and party unity in parliamentary votes to be stronger in the Czech Republic than in Poland. This is because the threshold requirement in the Czech Republic means that the initial party ranking of candidates is very important in determining who wins seats, much more so than in Poland. Even in cases where a few Czech candidates surpass the threshold, the order of all the remaining candidates under the 5 percent threshold remains the same. Their rank is lowered, but the names stay in the same order. Thus, if one candidate positioned at the bottom of the ballot manages to pass the 5 percent threshold, that candidate finishes the election in the top position. The candidates who had initial ballot rankings of 1, 2, 3, etc. each move down one position, to 2, 3, 4, etc. Thus, a less popular candidate initially positioned by the party at the top of the ballot still has a good chance of winning a parliamentary seat in the Czech Republic’s multimember districts.
If the electoral connection hypothesis holds, then we would expect a stronger effect in Poland than in the Czech Republic, due to the greater importance of preference votes in determining which candidates win seats to the Polish Sejm (it is the sole determinant) compared to the Czech Chamber of Deputies, with its threshold requirement.
We begin our study by placing these hypotheses in the context of the literature on party unity and the effects of electoral systems on parliamentary behavior. Then we provide a discussion of party unity in both countries, before turning to our multivariate analyses and results. Our empirical analyses, conducted separately for Poland and the Czech Republic, find support for the party nomination mechanism in both countries. Polish and Czech MPs with higher preference vote shares exhibit lower party unity on parliamentary votes. Further, when comparing the effects in both countries, our finding is consistent with what we expected with the presence of party nomination effects. The relationship between preference vote share and party unity is much stronger in the Czech Republic than in Poland.
Party Unity: Theoretical Reflection and Individual-Level Hypotheses
Democratic transitions made it possible to expand the study of legislative behavior to the countries of Central and Eastern Europe. The early research on the Polish and Czech parliaments focused on the description of newly created democratic institutions and patterns of party politics. 6 The emergence of stable electoral rules and their successful application in several elections drew the attention of students of comparative politics. This stability, however, was coupled with diverging patterns of party system development. Measured by the incidence of party switching, the Czech party system consolidated earlier than its Polish counterpart. 7 In Poland, the “fluid” party system 8 was evident with a high degree of electoral volatility between consecutive elections. 9 This picture changed in 2007 when the Polish party system experienced consolidation and started to resemble the Czech system. Because of the improvement of governing party discipline (as measured by lower rates of party switching) and a decrease in electoral volatility, the Polish party system can be described as quasi-institutionalized during the time period of our study. 10
While previous research has focused on institutional and party system characteristics, we delve deeper into an examination of party unity in parliamentary voting and how electoral rules impact unity levels. Party unity is closely related to the phenomena discussed in the aforementioned work, but until now has drawn relatively little attention in the context of Central and Eastern Europe. A notable exception is the work of Tavits 11 explaining party unity by the strength of party organization.
The analysis of voting unity in parliaments has a long tradition in legislative studies, which can be attributed to the importance of unity for democratic representation. 12 Voters can only hold parties accountable for their decisions when legislators of the same party act in unison on a regular basis in parliament. Carey 13 highlights that “whether voters can know what they are getting in elections depends partly on legislative voting unity. If the voting behavior of a party’s legislators is unrelated to the positions in its national platform, then the party’s label has no informational value.” Moreover, party unity is a central requirement for governing parties to achieve their goals and to fulfill their electoral promises. 14 Since most parliaments are characterized by government-opposition voting dynamics, 15 a divided governing party will likely have trouble passing bills in parliament. Hence, although legislators are free to decide how to vote on legislation, high levels of party unity are often observed in parliamentary democracies.
Even in parliaments characterized by high unity levels, at times, legislators defect from the party line. Scholars have discussed the degree to which the electoral system creates an “incentive to cultivate a personal vote” 16 for candidates, as an important factor affecting defection rates. When candidates have more incentive to cultivate a personal vote, it is assumed that they have less incentive to follow their party line, and therefore we would observe a higher likelihood of defection. 17 More specifically, it is argued that electoral systems that allow legislators to receive a personal vote, that is, a vote for a specific candidate instead of an overall support for the closed-party list, yield lower levels of party unity. 18 From this perspective, defecting from the party line in parliament can be interpreted as a strategy of legislators to secure personal votes in the next election. Some voters might dislike the position of a party on a certain issue, yet they can choose to vote for a legislator who voted against the party line on these bills. 19 As a consequence, legislators face conflicting incentives for their parliamentary voting behavior. On the one hand, high levels of unity are important for the party in order to be held accountable by the voters. On the other hand, personal vote earning strategies might create incentives for legislators to strategically deviate from the party position.
While many studies address the effect of personal vote earning strategies on vote unity by comparing unity levels between different parliaments and countries with different electoral systems, there is little research on the question of how unity diverges between candidates within electoral systems with high incentives to cultivate a personal vote. Put differently, we do not know how different legislators respond to the competing principals. Are some legislators more likely to follow the party line than others, and which characteristics predict such a behavior?
Voting unity can be defined as the “observable degree to which members of a group act in unison.” 20 Yet, the observed level of voting unity can have different causes. In this regard, the difference between cohesion and discipline is important. 21 While cohesion refers to fundamental policy preferences shared by legislators of a party, discipline is defined as “either . . . a special type of cohesion achieved by enforcing obedience or . . . a system of sanctions by which such enforced cohesion is attained.” 22 Therefore, high levels of party unity do not necessarily reflect high levels of agreement among the respective legislators. Instead, unity could also be achieved by specific incentives set by the parliamentary group leadership. 23 Discipline is usually enforced by imposing fines or suspending the privileges of deviating MPs (e.g., suspension of rights as parliamentary faction member). One may ask in this context: “Do individual legislators vote with fellow party members in spite of their disagreement about the policy in question, or do they vote with fellow party members because of their agreement about the policy in question?” 24 In short, party unity can be perceived as the end product of multiple individual decisions by legislators. Thus, it is useful to consider what motivates MPs to vote in line with their party. Here we focus on two of these potential factors: the number of preference votes received by a candidate and, in order to compare the results from Poland and the Czech Republic, the degree of party-centeredness of the electoral system.
Since party unity is the product of individual legislators’ preferences, it is important to analyze their motivation in more detail. Following the work of Müller and Strøm, 25 legislators are simultaneously policy-, office-, and vote-seekers. If legislators were solely interested in policy outcomes, the means associated with discipline would be ineffective. The voting behavior of legislators would be determined solely by their policy-preferences. In contrast, both office- and vote-seeking motivations create incentives for MPs to deviate from their preferred policy-position. Office seeking describes legislators’ career ambitions, which are more likely to be fulfilled if they remain loyal to the party. Parties eagerly reward legislators for adhering to the party line in their voting behavior. 26 The rewards may be a cabinet office, an attractive committee position, or a high ballot position, which increases an MP’s chances of reelection. 27
In contrast, vote-seeking may result in deviating voting behavior, since legislators act as agents of their voters as described by the concept of “electoral connection.” 28 Therefore, legislators may feel more indebted to their voters than to the position of the party. On the one hand, they are motivated to toe the party line in order to secure an “attractive” position on the ballot; on the other hand, they may want to vote against the majority of their party when a bill they support is unpopular, in the hope of attracting voter attention. Under PLPR systems, winning a seat depends both on preference votes as well as support of the party. MPs can therefore be perceived as agents acting on behalf of two principles. 29 As a consequence, office- and vote-seeking incentives may influence legislators’ voting behavior in different directions and describe an essential conflict between loyalty to the party and loyalty to the voters.
While all politicians are confronted with the dilemma of how to vote on unpopular bills, an MP’s position within the party likely conditions his or her final decision. Mershon and Shvetsova 30 posit that “a politician’s disloyalty to her electoral party is costly at the next election.” This is true especially for politicians with weak positions in their district, since they are more dependent on party support than politicians who can count on strong support of their constituents. In PLPR systems, losing a popular politician may also be costly for the party. These MPs can help the party attract voters, but, at the same time, these popular MPs may use their prominence to adopt more independent positions on parliamentary votes. This independence, in turn, will be tolerated by the parties who need them as “list-pullers.” 31 As shown by Gherghina, 32 using the example of Romania, parties are willing to accept unreliable politicians (e.g., those who changed party affiliation), because they expect benefits in the form of visibility, popularity, and professionalization.
As a consequence, in PLPR systems, popular MPs who receive a high preference vote share may feel less dependent on their party. In contrast, the party relies heavily on these legislators because they can potentially increase the party’s support. Therefore, “list-pullers” are less likely to be sanctioned by the party leader and consequently more likely to follow their own or their supporters’ preferences. A party may tolerate this behavior as long as it is limited to the few popular politicians, and may reward these MPs with attractive ballot positions. Therefore, we hypothesize:
Hypothesis 1a: The higher an MP’s preference vote share, the lower his or her unity level in the subsequent parliamentary term.
We test this hypothesis in both countries, separately, though, as we will see in hypothesis 2, we expect the magnitude of the relationship to differ.
The literature on the personal vote and the electoral connection suggests the opposite effect. Cain et al. 33 define the personal vote as the “portion of a candidate’s electoral support which originates in his or her personal qualities, qualifications, activities, and record.” This is distinct from the party vote that is generated by the voter’s attitude towards and evaluation of the political faction a candidate belongs to. In contrast to the party vote, individual legislators can directly influence their personal vote, which in turn has implications for party cohesion in the legislature. 34 Under PLPR systems legislators have an incentive to cultivate their personal vote. In the purest form, every additional preference vote increases their chances of finishing closer to the top, surpassing party colleagues competing in the same multimember district. If this logic prevails, we should expect to find politicians who were less successful in attracting preference votes in the previous election to be more prone to defect from the party line during the parliamentary votes, compared to those legislators who won many preference votes. Therefore, fewer personal votes might lead to less disciplined behavior and consequently legislators with more personal votes should show higher levels of vote unity. Therefore, this hypothesis, which we test separately in both countries, is the opposite of the one presented in hypothesis 1a:
Hypothesis 1b: The higher an MP’s preference vote share, the higher his or her unity level in the subsequent parliamentary term.
Party Unity: System-Level Hypotheses and Case Selection
Now we turn to how system-level differences might influence the strength of the relationship between preference vote shares and party unity. Previous comparative studies on unity have focused primarily on the effects of presidential and parliamentary regimes. 35 Legislators in presidential democracies usually follow the party line less strictly than MPs in parliamentary regimes. 36 This is explained by the government’s dependence on the parliamentary majority in parliamentary democracies. 37 When a government has no confidence in its parliamentary support base, its continuity cannot be secured. 38
The work examining the relationship between electoral rules and legislative behavior in countries with mixed-member electoral systems is particularly revealing. These studies find that MPs elected in the majoritarian tier exhibit lower levels of party unity than their colleagues elected in the PR tier, because of the candidate-centered focus of majoritarian elections. 39 In some cases, however, the effects are relatively weak and cannot be observed until the second or third legislative period after the introduction of the mixed-member system. 40 Other researchers, however, find little difference in the legislative behavior of deputies elected in different tiers and speak of possible contamination due to the party’s role in candidate selection and candidates running in both tiers. 41
A different perspective is offered by Sieberer. 42 In the context of the German mixed-member proportional system, he observes how different groups of MPs adopt distinctive strategies in reacting to legislation about which they are critical. Sieberer identifies several mutually exclusive categories of MPs: representatives of government versus opposition, leaders and nonleaders, and district and pure list candidates. MPs who appeared only in the proportional tier (pure list candidates) are more likely to use the “explanation” of the vote without actually voting against their parties than other types of candidates. The same effect is found among the government MPs as opposed to legislators representing the opposition. But regardless of whether in government or opposition, party leaders are less likely to use the explanation option than rank and file MPs. 43 Sieberer’s findings concerning the behavior of legislators elected in the majoritarian tier are less conclusive.
In this study, we are comparing across two countries rather than across two electoral tiers within the same country. As neighboring countries with similar social, economic, and political experiences over the past century, Poland and the Czech Republic provide an ideal setting to test the impact of more nuanced differences in the electoral rules. In terms of electoral system design, they can be considered most similar systems. Both use PLPR for elections to their lower houses of parliament and each party determines the order of its party candidates on the ballots in each electoral district. The power of the party in terms of constructing the candidate ranking means that the most prominent candidates are listed at the top of the party ballot, while inexperienced and lesser-known candidates appear near the bottom.
Although both countries use PLPR, the Polish system is considered “quasi-list” and the Czech system “flexible list.” 44 To understand the implications of the electoral rules for parliamentary behavior we turn to the list of variables enhancing personal vote-seeking identified by Carey and Shugart. 45 On the first factor, control over access and rank of candidates on the ballot, the countries are the same. The party has ultimate control over candidate ranking on the ballot. Poland and the Czech Republic do diverge on the other two characteristics. In Poland, voters cast a single intra-party vote, while in the Czech Republic, they are allowed to cast up to four intra-party votes or a party-level vote. Furthermore, the countries differ with respect to the “degree to which candidates are elected on individual votes independent of co-partisans.” 46
The quasi-list, used in Poland, is an open-list system where the final order of candidates is determined solely by the number of preference votes each candidate wins, and requires that the voters must cast a preference vote. The highest vote-getter wins the top party position in that district, while the candidate with the fewest votes ends up last. Theoretically this method gives every candidate, independent of his or her placement on the party list, identical chances of being elected. In reality, however, Polish candidates placed close to the top of the party ballot attract more votes than those placed further down the list. 47 Thus, candidates placed near the top of the ballot have a higher chance of winning a seat. Among candidates of all parties contesting the Polish parliamentary election 2007, the chances of winning a seat if running from the first (top) position on the ballot amounted to 47 percent. This drops to 32 percent for candidates listed second on the ballot. Candidates in ballot positions between the 25th and 75th percentiles had just a 1.6 percent chance of winning, while those in the bottom quartile had a 0.9 percent chance. Still, the relationship between the likelihood of election and the ballot position is not deterministic in Poland. There were forty-five candidates who won parliamentary seats despite their position in the central section of the ballot (25th to 75th percentile) and fifteen who entered the Sejm running from the lowest quartile. As demonstrated by Gendzwiłł and Raciborski, 48 roughly 20 percent of seats in the Polish Sejm were allocated solely due to preference votes. These candidates would not have entered parliament under the closed-list system.
In flexible-list systems, like the one used in the Czech Republic, the final postelection order of the party candidates “takes into account both the party-provided rank order and preference votes” often through the use of a quota or threshold. 49 Prior to the 2010 election, the Czech Republic revised its preference voting rules by increasing the number of preference votes each voter could cast from 2 to 4 and reducing the threshold from 7 to 5 percent. 50 Czech voters are not obligated to cast their preference votes; it is optional. The 5 percent threshold is determined by the number of preference votes received by the candidate divided by the total number of ballots cast for that party in that district. Candidates who surpass this threshold move to the top position, while the order of the other candidates remains unchanged, though each would have a lower rank. If multiple candidates exceed the 5 percent threshold, the final order of all those who passed the threshold is determined by the number of preference votes. The candidate above the threshold with the highest number of preference votes would be in the first position. If no candidate surpasses the threshold, the party ranking stands.
Drawing on Carey and Shugart, 51 we contend that the differences described above would provide stronger incentives to cultivate the personal vote in Poland than in the Czech Republic. As a result, we expect to observe a difference in the magnitude of the relationship between preference votes and party unity. This difference, however, is conditional on our findings concerning hypotheses 1a and 1b. Under the party nomination mechanism (hypothesis 1a), we expect that in the Czech Republic, MPs with lower preference vote shares will strive to stay in good standing with the party in order to secure a high ballot position in the next election. This strategy would increase their chances of reelection, because the 5 percent threshold requirement gives the party, through its original ranking of candidates, a good bit of control over who wins. In Poland, because preference votes solely determine who wins seats, there is less incentive for this behavior.
Hypothesis 2a: If hypothesis 1a (party nomination mechanism) can be corroborated in both countries, the effect of preference votes on MP party unity will be stronger under a more party-centered electoral system (Czech Republic) than under an electoral system with stronger ballot position effects (Poland).
However, if we find evidence of the electoral connection mechanism (hypothesis 1b), the opposite difference at the country level should be observed. The incentives to cultivate the personal vote should be weaker in the Czech Republic than in Poland, because the Czech system provides a stronger role for the party, while election to the Polish parliament is directly linked to number of preference votes.
Hypothesis 2b: If hypothesis 1b (electoral connection mechanism) can be corroborated in both countries, the effect of preference votes on MP party unity will be weaker under a more party-centered electoral system (Czech Republic) than under a less party-centered electoral system (Poland).
Data and MP Characteristics
Our analysis covers every MP who participated in a parliamentary vote as a member of the party from which he or she was elected in the 2007–2011 Sejm and 2010–2013 Czech Chamber terms. 52 MPs who left the party in the course of the legislative term were taken into consideration only for the period of time they belonged to their original faction. We also do not account for parties created through splits in existing factions, since the members of these new groups ran for office from a different party.
In total, our analysis focusing on Poland includes 507 MPs representing four major parliamentary factions present in the sixth Sejm: PO, PiS, SLD and PSL. This is higher than the number of MPs the Sejm seats (460) because as vacancies opened up in parliament, the parties filled them with the next candidates on the party lists. The left-of-the-center faction in the Polish parliament will be referred to as the SLD, in spite of its name changes (LiD, Lewica, and finally SLD) and internal divisions which occurred during the legislative period, because members of the SLD constituted from the beginning the overwhelming majority of this parliamentary group.
In the Czech Republic, we analyze the parliamentary votes of 215 MPs who participated in at least one vote as members of one of five parliamentary parties (ODS, ČSSD, TOP 09, KSČM, and VV). Like in Poland, 215 is larger than the 200-seat Chamber of Deputies. Again, this is due to filling vacancies.
Information on the unity level of each MP was calculated based on parliamentary votes he or she cast during the 2007–2011 Polish and 2010–2013 Czech parliamentary terms. 53 Altogether, our analysis includes 8,437 votes in the Polish Sejm and 5,898 votes in the Czech Chamber of Deputies. In Poland, 34 votes were excluded from the analysis, since they were recorded only as a “valid vote.” This type of code was used when deputies did not have an option of voting with “yes,” “no,” or “abstain,” but had to support one of several options, such as during the election of the parliament’s president. Since virtually all votes in Poland 54 and the Czech Republic 55 are roll-calls, our data are not affected by selection bias. 56
Party unity is operationalized as the percentage of roll-call votes by a given MP that conforms to the party line, 57 where the party line is inferred from the behavior of the majority of party members. 58 It is the “modal vote choice for all members present and voting.” 59 Similar to the approach taken by Hix, 60 we take into consideration all three possible decisions: yes, no, and abstain. If the majority of a given party voted yes, an MP “is considered to have defected from this group in a particular vote if she registered either a no or an abstain in the vote.” 61
Tables 1 and 2 present the descriptive statistics on MP party unity by party in Poland and the Czech Republic. Looking at the Polish unity rates, we observe that party unity levels among Polish MPs range from 73.75 percent to 100 percent. Overall, on average, MPs are voting with the majority of their party 98.78 percent of the time. As in several preceding legislative periods, voting cohesion in Poland is stable and very high. 62 Such high cohesion levels are not unusual and are comparable, for example, with the situation observed in the British House of Commons prior to the 1970s. 63 This does not mean, however, that there is no dissent. The vote on same-sex civil unions that took place in the Sejm in January 2013 proves the contrary. Out of 200 voting members of the largest faction, the Civic Platform, 46 voted together with the conservative opposition against the project sponsored by their party colleague, Artur Dunin, and 8 abstained. This made it possible to reject the legislation supported by the Prime Minister Donald Tusk that would legalize the civil unions of same-sex couples. 64
Poland: Party Unity Descriptive Statistics (2007–2011)
Czech Republic: Party Unity Descriptive Statistics (2010–2013)
We observe only slightly different levels of party unity across the Polish parties. The Civic Platform (PO), the party of PM Tusk, had the highest unity level, 99.60 percent. The PO is followed closely by the strongest opposition party, Law and Justice (PiS) with 98.65 percent and the Polish People’s Party (PSL), the junior coalition partner, at 97.98 percent. The lowest unity level, 96.41 percent, was recorded by the Democratic Left Alliance (SLD), which witnessed several changes in the composition and name of its parliamentary caucus. That the SLD has the lowest unity level, on the surface, contradicts the findings of Tavits 65 who recognizes the SLD as one of the best organized political groups in Poland and therefore should exhibit high unity. But the deterioration of the SLD’s position and the defection of its prominent members occurred after the period Tavits 66 studies.
The range of unity levels in the Czech Republic (Table 2) is wider than in Poland. The lowest MP unity level is 57.87 percent while the high is 97.07 percent. The overall mean party unity level among the members of the Chamber of Deputies is 87.15 percent. The new party and junior coalition partner, TOP 09, had the highest level of party unity at 91.37 percent. This was closely followed by the Communist Party of Bohemia and Moravia (KSČM), which was recognized by Tavits 67 as the best organized party in the Czech system. The KSČM is unique in post-communist Europe in that, as a largely unreformed communist party, it continues to have a following, with a national vote share above 10 percent in all parliamentary elections in the post-communist era. 68 TOP 09 and the KSČM were followed by the second junior coalition partner, also a new party in 2010, Public Affairs (VV) with 90.13 percent. Despite VV’s high unity level, it was severely affected by scandals and the splintering of the parliamentary party. 69 The unity levels of these three parties differ considerably from the levels of the two major parties, the Civic Democratic Party (ODS) and the Social Democratic Party (ČSSD). The average unity level of the ODS, which led the government under Prime Minister Petr Nečas, was 87.83 percent. The Social Democrats, as the dominant opposition party, were the least disciplined (81.27 percent).
In hypothesis 1a, we expect a negative relationship between preference vote shares and party unity—the higher the preference vote share, the less unity. In looking at individual MP profiles, we find many who exemplify this relationship. While examples do not prove the existence of the relationship, they do provide a way to further appreciate the statistical relationships. In the Polish case, Józef Zych of the PSL and Marek Balicki of the SLD depict the behavior of popular politicians. Zych won 56.36 percent of votes cast for the PSL in his home district and belonged in the group of legislators who least often supported his party with a 95.12 percent unity score. The mean unity for the PSL was 97.98 percent. Balicki won 52.10 percent of his party district list vote, but voted together with his party colleagues in 95.44 percent of votes, below the party mean of 96.41 percent. The backbenchers, Aleksander Sopliński (PSL) and Sylwester Pawłowski (SLD), provide contrasting examples. They obtained 5.59 percent (Sopliński) and 4.79 percent (Pawłowski) of the district list vote, while their respective unity levels of 98.76 and 97.97 percent were higher than their party averages.
In the Czech case, the popular leaders of the two new parties, Karel Schwarzenberg (TOP 09) and Radek John (VV), exhibited particularly low unity levels compared to the rank and file legislators of their respective factions. Schwarzenberg, who won 31.4 percent of preference votes, followed the party line in just 70.41 percent of parliamentary votes. This compares to the TOP 09 party mean of 91.37 percent. Similarly, John attracted a 28.06 percent preference vote share and voted with the majority of his colleagues 82.88 percent of the time, well below the party average of 90.13. Examples of rank and file MPs from the same parties include Jaroslav Eček (TOP 09) and Radim Vysloužil (VV). Eček won less than 1 percent of the votes for his party, and in parliament voted along with the majority of TOP 09 MPs 96.73 percent of the time. Vysloužil garnered a preference vote share of 6.51 percent and had a unity level of 96.19 percent.
From these examples and the data in tables 1 and 2 we observe that the average party unity in Poland is much higher than in the Czech Republic. Since we are comparing MP behavior in two different countries, these different levels of unity could result from differences in cultures and norms within the parties or parliaments in the two countries. So how does this mesh with hypotheses 2a and 2b? To test our expectations, we compare the strength of the relationship betweeen preference vote share and party unity in each country, rather than the levels of unity.
The primary independent variable of theoretical interest measures each MP’s support in her or his electoral district. It is expressed as the percentage of preference votes won by an individual MP computed from a total of all preference votes cast for that party at the district level. The observed values of this variable range from 0.38 to 86.34 percent among Polish legislators (N = 507) and from 0.73 to 31.42 percent among their Czech colleagues (N = 215).
As an alternative operationalization of an MP’s strength at the time of the last election, we include his or her standardized rank on the ballot. The ballot position is the direct expression of a candidate’s standing within the party, since the order of names on the ballot is determined by the party elites. Ballot rank is related to preference vote shares, since voters gravitate to the candidates at the top of the list. 70 However, this relationship is not deterministic. Because of the differences in ballot length, the rank variable is standardized so that first position is defined as 0 and the last position as 1. The intermediate positions reflect a given percentile of the overall ballot length where a candidate’s name appears.
In addition to the preference vote share and ballot rank variables, we account for other characteristics that likely affect how often MPs vote with their party. Given the different unity levels across political parties in both countries, we control for MP party membership. We also control for whether or not the MP ran in the 2007 Polish or 2010 Czech elections as an incumbent. Of the Polish Sejm members included in the analysis, 56.61 percent held a seat in the preceding legislative period. In the Czech Republic, this was true of only 41.86 percent of the MPs, which is due to the entry of two new parties to the Chamber (TOP 09 and VV) and the failure of two other incumbent parties to surpass the 5 percent national threshold for seats (the Christian Democrats and the Greens). In Poland, two parties failed to pass the electoral threshold (Self-Defense and the League of Polish Families), but no new party won seats.
Finally, we include two sociodemographic variables, age and gender. 71 The Polish MPs were between 22 and 78 years old. In the Czech Republic, the age of deputies ranged from 23 to 73 years old. The mean age of MPs in both countries is approximately 47 years. In both parliaments males outnumber females by large margin. Among the 507 Polish MPs, 79.88 percent are males and merely 20.12 percent females. These rates are nearly identical among the Czech MPs in this study: 78.60 percent are men, 21.40 percent are women.
Empirical Results
Tables 3 and 4 present the results of the OLS regression models for party unity in Poland and the Czech Republic, respectively. Model 1 in both tables accounts for the preference vote share and party membership. After establishing the impact of preference vote shares and party membership on MP unity, in model 2, we introduce MP characteristics: incumbency, age, and gender. Then, in model 3, we operationalize the strength of the candidate in the previous election by his or her ballot rank instead of preference vote share. Thus, we can directly measure the standing of candidates within their parties. Model 4 is the extension of model 3 using identical controls as those in model 2. Finally, model 5 simultaneously tests the effects of ballot rank and preference vote share.
OLS Regression: Explaining MP Party Unity in Poland (2007–2011)
Note: White’s heteroskedasticity-robust standard errors in parentheses (*p < 0.05, **p < 0.01, ***p < 0.001). OLS = ordinary least squares; RMSE = root mean square error.
OLS Regression: Explaining MP Party Unity in the Czech Republic (2010–2013)
Note: White’s heteroskedasticity-robust standard errors in parentheses (*p < 0.05, **p < 0.01, ***p < 0.001). OLS = ordinary least squares; RMSE = root mean square error.
Looking first at the Polish results in Table 3, we see that as expected in hypothesis 1a, preference vote share has a statistically significant and negative effect on party unity. The higher the preference vote share the MP received in the election, the less loyal the MP is to the party line in the following legislative term. A one percentage point increase in an MP’s preference vote share results, on average, in a decrease of 0.02 percentage points in the unity level. The magnitude of the effect of preference vote share remains similar across all models. As anticipated based on the average party unity scores by party in Table 1, the party dummy variables are statistically significant. In models 2 and 5, we see that while preference vote share and the parties remain statistically significant, none of the MP characteristics has an impact on unity. Models 3 and 4 show that the standardized ballot rank of a candidate has a statistically significant effect on unity. We note that the positive sign on the ballot rank coefficient indicates that the higher the ballot rank (lower on the list), the greater party unity. The higher the value of the ballot rank, the larger the candidate’s distance from the candidate occupying the first position on the ballot. An increase in the ballot rank from the first (top of the list) to the last position, leads to an increase of unity by between 0.47 (model 5) and 0.95 (model 4) percentage points. In the model including both preference vote share and ballot rank (model 5), preference vote share maintains its statistical power, but ballot rank fails to be statistically significant. Despite the existing correlation between ballot rank and preference vote share, the VIF values on both variables remain lower than 2.
The Czech models presented in Table 4 offer a similar picture as far as the main hypothesis is concerned. The effect of preference vote shares on party unity is consistent with hypothesis 1a. A one percentage point increase in the percentage of preference votes won by an MP decreases the MP’s unity by roughly 0.27 points. Like in Poland, this relationship persists after controlling for the characteristics of MPs (models 2 and 5). The party variables are statistically significant, though VV loses this status in models 2, 4, and 5. We also see in models 2, 4, and 5 that incumbency and age have no effect, but the gender variable is positive and statistically significant. In the Czech Republic, after controlling for party membership and preference vote share, female MPs are more likely to vote with their party than their male colleagues. Compared to Poland, the effect of ballot rank on unity is statistically significant in all models in which it is included. The movement from the first (top) to the last position results in an increase of unity by between 3.16 (model 5) and 3.82 (model 3) percentage points.
The direction and the level of statistical significance of the effects allow us to decisively reject the electoral connection mechanism suggested by hypothesis 1b. There is no evidence, in either country, that MPs who performed worse in the previous elections work to cultivate the personal vote through defection in parliamentary votes.
Now that we have established that the negative relationship between preference vote shares and party unity holds in both countries, we turn to our hypotheses concerning the country differences. We can reject hypothesis 2b, since the electoral connection hypothesis (H1b) did not hold in either country. So we evaluate hypothesis 2a, which hypothesizes that the relationship between preference vote share and party unity will be stronger in the Czech Republic than in Poland. To assess the relative strength of the relationship between preference vote share and party unity, we can compare the coefficients. Looking at model 5 in Tables 3 and 4, we observe that the preference vote share coefficient amounts to −0.27 in the Czech Republic compared to −0.02 in Poland. The magnitude of the coefficient is much stronger in the Czech Republic than in Poland. We can see this, as well, through interpretation of these coefficients. For example, in the Czech Republic, a 10 percentage point increase in preference vote share results in a 2.7 percentage point decrease in party unity, whereas the same increase in preference vote shares in Poland results in just 0.22 of a percentage point decrease in voting with the party. The effect of preference vote share is 10 times stronger in the Czech Republic than in Poland, providing strong confirmation of our hypothesis.
Another interesting difference between Poland and the Czech Republic is the effect of gender on party unity. In both countries, women are underrepresented in parliament and traditional gender role stereotypes prevail. So, the surprise for us is not that gender is statistically significant in the Czech Republic (female MPs are more likely to vote with their party than males by 3.38 percentage points), but that gender is statistically insignificant in Poland. At this point, we can only speculate why we observe this, but in terms of gender representation, an important difference between the two countries occurred during the legislative periods under consideration. Poland adopted legal gender quotas that went into effect for the 2011 election, while the Czech Republic does not have such a law. The law obligates Polish parties to allocate at least 35 percent of positions on the party list to females and at least 35 percent to males. 72 It is possible that the introduction of legal gender quotas could have reduced incentives for women already holding a seat in parliament to exhibit strong loyalty towards the party, since parties needed to actively seek more women as candidates to meet the quota. Perhaps female MPs felt that their positions on the party list were secure, as secure as male MP positions. In order to understand whether the introduction of gender quotas has an impact on the voting behavior or female MPs, we would need to look at Polish parliamentary votes over time, including the first complete legislative period after the introduction of gender quotas in 2011.
Discussion
In both Poland and the Czech Republic, MPs vote in predictable ways. After controlling for party and MP characteristics, we find that MPs with high preference vote shares are less likely to vote along with their party on parliamentary votes, while those with low vote shares are more likely to toe the party line. However, we find important differences in the strength of this relationship between Polish and Czech MPs. In the Czech Republic, where the party’s initial ordering of candidates on the ballots plays a much stronger role in determining which candidates win seats, less “popular” MPs are more likely to vote with their party on parliamentary votes to better ensure a high ballot position in the next election. In Poland, where only the number of preference votes determines which candidates win seats, the relationship between preference vote shares and party unity is weaker. The analysis using initial ballot ranking as the explanatory variable leads to similar conclusions. Individuals occupying less attractive positions on the ballot at the time of election defect from their party in the parliamentary votes less often than candidates placed closer to the top of the list. This effect is, nevertheless, not statistically significant in Poland when we control simultaneously for preference vote share and candidate characteristics.
We began this article by asking whether the electoral system rules provide incentives for certain types of legislative behavior. Previous literature has analyzed this question by either comparing different electoral systems or different tiers in one mix-member system. Here, we have examined two PLPR systems: the more candidate-centered quasi-list system of Poland and the more party-centered flexible-list system of the Czech Republic. Our results indicate that these rule differences within PLPR systems matter for parliamentary behavior. Where the party has more control over who wins parliamentary seats, the relationship between MP preference vote share and party unity is stronger.
The results of our study provide further evidence for the mechanisms relating MP popularity among voters to lower adherence to the party line identified by Crisp et al. 73 in the Slovak case. We show that those who entered parliament enjoying strong popular support (a higher preference vote share) are less faithful to their parties in the subsequent parliamentary term, while Crisp et al. 74 show that these popular “list-pullers,” with low unity levels, are assigned to high ranks on the ballot and attract many votes for the party.
A potential extension of our study would be the inclusion of data on previous legislative periods. This could be especially interesting in terms of post-communist development. There is evidence from the voting behavior literature that it took Central European voters time to “learn.” 75 Likewise, we might expect it to take time for MPs to learn the incentives created by the electoral rules. The gradual consolidation of the Polish party system 76 (at least until the 2015 parliamentary election) and the simultaneously observed crisis of the leading Czech parties 20 years after the democratic transition 77 may be another interesting aspect influencing the relations between parties and individual legislators over time.
An additional question revealed by our analysis concerns the differences in the parliamentary voting behavior of male and female MPs. While the 2011 introduction of legal gender quotas in Poland drew our attention, the situation of female legislators in the Czech Republic should undergo closer scrutiny to understand why female MPs, on average, vote the party line more often than their male counterparts. The 2010 Czech election marked the high point in the number of women elected to the in the Chamber (22 percent) 78 in the post-communist era. 79 We would need to look at voting in earlier parliaments to see if there is a relationship between the percentage of women in parliament and the corresponding unity levels. One might expect that as more women enter parliament, the behavior of male and female MPs would become more similar, after controlling for other relevant characteristics. As far as Poland is concerned, we concur with Górecki and Kukołowicz, 80 who call for additional analyses of the gender quotas after the parliamentary election in October 2015. Once the first parliamentary term after introduction of gender quotas has been completed, we have gained an opportunity to evaluate their effect on the parliamentary voting behavior of female MPs in more detail.