Legislator activities in parliament and electoral performance in a mixed-member electoral system. The case of Hungary,

Transcription

1 Legislator activities in parliament and electoral performance in a mixed-member electoral system. The case of Hungary, Zsófia Papp Research Fellow Centre for Social Sciences Hungarian Academy of Sciences papp.zsofia@tk.mta.hu First draft Presented at the IPSA World Congress in Poznan, 25 July 26 Abstract Students of legislative behaviour appear to agree with Mayhew s starting point, and consider MPs as single minded seekers of re-election. Consequently, most of the things MPs do in the present are attributed to their individual desire to future re-election. However, within the premises of legislative studies, little attention has been paid to whether MP behaviour in parliament indeed affects electoral performance. The main question of the paper is whether there is a connection between activities carried out by legislators and their performances at the next elections. I wish to investigate this matter in a mixed-member electoral system permitting multiple candidacy which produces at least two types of MPs with different sets of incentives to act. To establish the effect of parliamentary activities on electoral outcome, I use parliamentary activity panel data from three electoral terms. The results of the multilevel models suggest that there is no connection between the intensity of work MPs carry out in parliament and how they perform in single member district on the one hand, and where they are placed on closed party lists on the other. The findings partially question the existence of activity-based personal accountability in Hungary. Keywords Legislators; Parliament; Hungary; Accountability; Elections

2 Introduction Students of legislators behaviour might differ regarding the activities they study, the factors with which they explain the variance in these activities and the methods they apply to uncover causal relationships. However, the majority of authors seems to agree with Mayhew s oft-cited starting point, and consider Members of Parliament (MPs) as single-minded seekers of reelection (Mayhew, 974). Consequently, most of the things representatives do in the present are attributed to the MPs individual desire for future re-election. At the same time, Hazan and Rahat point out that under centralised selection rules with a high number of safe seats, the aim of re-selection is superior to the purpose of being re-elected (Atmor, Hazan, and Rahat, 2; Hazan and Rahat, 2). In the former case, where re-election incentives dominate, candidates who are able to attract extra votes will be selected, whereas in the latter, where re-selection motivation prevail, candidate loyalty is considered as priority. Nevertheless, even in safe districts, to maintain the electoral base, it might appears to be reasonable to follow strategies that highlight the individual legislator without endangering party unity (see also Russo, 2). Within the premises of legislative studies, less attention has been paid to whether MP behaviour in parliament and in the constituency indeed affects electoral performance. Extensive research has been done on the determinants of legislative behaviour, all based on the presumption that the MPs perceptions on the utility of their work is more important than the actual electoral benefit. Legislators work more and better because they think it will increase their re-election chances regardless of whether this is indeed the case. While this argument holds in several cases, it does not bring scholars closer to understanding the implications of member behaviour and whether it is realistic to assume that such behaviour yields extra votes or better positions on party lists. Lessening the gap in the literature, my main question is if there is a connection between activities carried out by legislators in parliament and their performances at the next elections in the case of Hungary. I wish to investigate this matter in a mixed-member electoral system permitting multiple candidacy for several reasons. Firstly, the case of mixed electoral system is particularly interesting for students of the electoral formula, as it enables to test the effects of two different set of rules within the same electoral and cultural context. Secondly, being a candidate on multiple electoral tiers at the same time creates a larger pool of viable strategies compared to system with one legislative tier on the one hand, and mixed-member system without multiple candidacy on the other. Thirdly, besides being able to investigate the effects of electoral activities on candidate votes share, by examining party list positions on closed lists, we can test party priorities in candidate selection at the same time. The main hypothesis of the paper is that both, selectors and voters evaluate the goodness of the legislators through their parliamentary activities. Voters seek to elect good representatives who are willing to do hard work in parliament, and they may use prior performance to evaluate future potential. Additionally, it can be argued that parties will also consider this aspect when organizing the party lists. Putting hardworking MPs on the top of the lists makes sure that they have reliable legislators to do the leg-work in parliament. The paper is structured as follows. First, I will briefly outline the theoretical framework of the paper, which is followed by a short introduction of the Hungarian case. In the data and methods section, the dependent and independent variables of the analysis are introduced. In the analysis section I apply various types of multilevel regression to establish the effect of legislator activities on electoral performance. Finally, I will draw conclusions in the last section. 2

3 Theoretical framework The missing link in the literature bears theoretical relevance especially for research establishing the connection between electoral rules and legislators personal vote-seeking. Namely, the linkage between MP behaviour and voter decisions, which authors refer to as the accountability linkage, is used to describe the mechanism that creates the effect of the electoral formula on MP activities in the first place. Figure clarifies this issue by displaying the relationship between the three factors: electoral rules, MP s activities and their electoral performance. As to the causal effects, (a) stands for the effect of electoral formula on member behaviour, while (b) covers the relationship between behaviour and electoral outcome. Establishing connection (a) in legislative studies often implies that connection (b) is already manifested. Figure. The relationship between electoral rules, member behaviour and electoral performance The underlying argument is that, since one constituency has one single representative in SMD systems, voters will be able to determine who to reward or punish for the positive or negative outcomes (Lancaster, 986; Lancaster and Patterson, 99; Norris, 24), thus making direct sanctioning possible (Mitchell, 2). In Multi-Member Districts (MMDs) with a geographical overlap among legislators (Heitshusen, Young, and Wood, 25) the accountability link becomes confused, and the incentive to free-ride increases. The reason for this is that the voters have difficulties identifying the representative responsible for the benefits, leaving little room for recognition and reward (Cain, Ferejohn, and Fiorina, 987; Lancaster, 986; Scholl, 986). In MMD-systems, members are more accountable to the party leadership than to the electorate (Norris, 24), which makes the contest for re-election a fight for higher positions on the party list (Curtice and Shively, 29). Single member district MPs on the other hand are more dependent on local support, since their electoral fortunes depend on the votes cast for them, and these votes cannot be enhanced or diminished by the electoral performance of fellow party members (Curtice and Shively, 29; Mitchell, 2). One MP is responsible for the whole constituency, and cannot delegate problem-solving to other representatives (Gallagher and Holliday, 23; Norris, 24). As voters in SMDs can easily identify who to approach with their problems (Scholl, 986), and as members are usually better known by the electorate (Cooper and Richardson, 26), single member districts tend to generate larger demand for constituency service, than MMDs (Scholl, 986). And since SMD MPs are more dependent on local support, they are more vulnerable to the pressure from the electorate as well (Cooper and Richardson, 26). In short, voters have expectations that are called forth when casting the vote; and whether they can easily decide whether legislators had met these expectations or not depends on electoral rules. The first models of representation described it as some kind of agreement between citizens and the representatives. The main goal of elections was to choose the best individuals to defend the interests or values of the community or group (Colomer, 2, p. 2). However, the emergence of political parties shifted this relationship toward an interaction between citizens 3

4 and parties instead of citizens and the individual legislators representing their districts. This was further enhanced by the extensive usage of multi-member Proportional Representation (PR) electoral systems, where voters were offered to choose between parties instead of individual candidates. This made party representation more stressful against personal representation. The shift from single member majority systems toward PR also made it difficult to apply theories of representation that were based on the logic of one member representing one constituency, further diminishing the linkage between representatives and the represented. The main principle of party representation is the representation of policy preferences: party representation is necessary in order to select the most relevant issues in the public agenda and to design public policy (Colomer, 2, p. 7). Personal representation on the other hand aims to ensure the good quality of representation, by appointing individuals who are reliable and responsive to the voters demands. In modern societies, the two aspects of representation appear in a special mixture in which the composition is heavily influenced by the given setup of the electoral system. This fits the conclusion of Eulau and Karps (Eulau and Karps, 977) as well as Mansbridge (Mansbridge, 23, 29) on the presence of the different types of responsiveness and models within the same system. This indicates that even in systems, where electoral rules or special circumstances support party representation, some forms of personal representation should be detectable. This involves that voters will be inclined to increase the quality of representation by electing representatives of the good kind. Practicing accountability on the basis of former promises for a good representative, they might do this by judging incumbent MPs for their actions during the previous terms. There are several ways in which Members of Parliament can mark out of the crowd of fellow legislators to advertise their goodness. They engage in constituency service in the districts: they hold office hours, carry out ombudsman-like activities and deal with problems arising in the area as well as respond to individual petitions (Cain et al., 987; Fenno, 978; Norris, 997; Norton and Wood, 99). However, MPs do not necessarily need to be present in their constituencies to gain visibility. As they spend most of their time in parliament, they must find different ways to differentiate themselves from other members. On the one hand, the contents of parliamentary activities reflect the focus of one s legislative work. The local focus of bill initiation (Crisp and Ingall, 22; Marangoni and Tronconi, 2), membership in certain committees (Manow, 23; Stratmann and Baur, 22), speeches (Hill and Hurley, 22), parliamentary questioning (Martin, 2) and the willingness to desert the party line at roll-call (Carey, 27; Tavits, 29) can reflect general concerns about issues that are potentially important for citizens of a well-defined geographical location. On the other hand, the intensity with which MPs engage in these activities is a simple proxy for their efforts to establish the reputation of a hard-working representative. Voters and parties gain important information on legislators simply by looking at their track records. Bill sponsorship, (Bowler, 2; Bräuninger, Brunner, and Däubler, 22), the frequency of floor speeches (Anderson, Box- Steffensmeier, and Sinclair-Chapman, 23) and tabling questions (Lazardeux, 25; Rasch, 29) or simple attendance indicate that they work just as hard as other members without having to assess the local nature of these activities. On the basis of the above, the question of whether MP behaviour is relevant in explaining electoral performance is important to make our assumptions stronger when tying MP strategies to electoral systems. More importantly, answering this question supplies us with ideas on how well personal accountability works in a given country. A steady relationship between member behaviour and electoral performance would suggest that a certain type of legislative accountability mechanism is in place. If MPs are judged individually on the basis of their performances as legislators, they are indeed encouraged to perform better to gain extra votes at 4

5 the next elections. Contrarily, no correlation between MP behaviour and their electoral fate would indicate that Members of Parliament are not held personally accountable for their actions while in office. Thus, the question regarding the link between behaviour and performance might not only justify our basic assumptions in legislative studies, but contribute to the discussion on the personal accountability of individual MPs and the broader topic of representation. The Hungarian case During the period under investigation (998-2), Hungary has a three-tier electoral system, which is commonly classified as mixed-member majoritarian with partial compensation (Shugart and Wattenberg, 2). A total of 386 legislators are elected from 76 single member districts (SMD tier), 2 regions (regional tier) and the national level (national tier). From the regional tier, a maximum of 52 representatives gain their mandates under the rules of proportional representation 2. A minimum of 58 seats and unallocated mandates from the regional tier are distributed on the national tier 3. Both list tiers apply closed party lists. Parties may nominate candidates on multiple electoral tiers at the same time. This practice is quite widespread in Hungary: between 99 and 2, almost 4 % of the candidates were nominated on more than one tier. Furthermore, candidates may be nominated by more than just one party. Joint candidacy is possible on all electoral tiers, but it is most common in the case of SMD candidates and is usually restricted to a rather small amount of candidates. There are some exceptions to this between 99 and 2, particularly on the right, namely the electoral coalitions of Fidesz and MDF in 22 and Fidesz-KDNP in 2 which concerned all candidates on all electoral tiers. Additionally, from 994 on, legislators are allowed to hold multiple elected offices such us various positions on the local level. In the Hungarian parliament, House rules entitle individual legislators to sponsor bills and submit questions. There are four types of questions in the Országgyűlés: interpellations, written and oral questions as well as direct inquiries. These types differ in terms of the form of the required response as well as sanctions in case the questioning MP does not accept the response to the question. Interpellations, oral and direct questions must be answered at the plenary session, while written questions require written response. With the exception of direct questions, all questions have to be submitted in writing prior to the session. Interpellations are said to be the most powerful among the question types, because if the questioning MP does not accept the answer of the government, a plenary vote is called to decide on the response. Hungarian legislators may also make brief speeches on the plenary during debates. However, PPG leaders and their deputies enjoy a wider range of opportunities to take the floor in the parliament than backbench MPs. Last but not least, representatives may sit in committees as permanent members. However, as the partisan setup of each committee is fixed, committee assignments are decided by the PPG leadership. Data and variables The analysis builds upon a dataset that collects publicly available electoral 4, sociodemographic 5 and parliamentary activity 6 data. Information on the parliamentary activities of 9 counties and Budapest. 2 Seats on the regional tier are distributed using the Hagenbach-Bischoff formula. 3 Seats on the national tier are distributed using the d Hondt formula and the dataset of the Centre for Elite Research at the Corvinus University of Budapest

6 MPs is unavailable for the first two electoral terms (99-994, ). The raw dataset is available in wide form 7, but for the sake of the analysis it is transformed into long form 8. As we follow a relatively large number of individuals (large N) throughout a relatively short period of time (small T), the data can be handled as panel. Since there are also MPs who did not participate in every single election under investigation, the panel is fairly unbalanced. Throughout this paper t will denote the year of the election under investigation. MPs are elected at election t-, serve between t- and t, and evaluated by the voters and parties at t. Overall, the dataset in long form has 232 observations. This figure consists of 3 individual MPs serving between 99 and 2. For the above reasons, the analysis is restricted to the period between 998 and 2, which leaves us with 2 observations for 755 individuals MPs only served once during the three electoral terms under investigation, 26 served twice and 2 three times. Table shows the distribution of observations across the different values of t. Table. The distribution of MPs across the different electoral terms t- t Elected MPs (t-) Candidates (t) SMD candidates Regional list candidates (t) National list candidates (t) (t) Number of observations in long form Dependent variables The analysis section investigates the effect of legislator activities on three types of dependent variables. First, candidate vote share in the SMDs at election t is explained. In this case, the selectors have already decided that the given MP will be running as an SMD candidate at the next election. Thus, the first dependent variable shows how voters react to the legislative record of the different MPs. In this case MPs who were nominated in SMDs at election t are taken into account. Nevertheless, legislative activities may not only influence voters in their electoral decisions, but also parties and selectors at the candidate nomination stage. As being an SMD MP in Hungary is considered to be a somewhat more prestigious position compared to holding a list mandate, the second DV will differentiate between SMD and list candidates. Nominating legislators in SMDs may be part of a reward mechanism for hardworking representatives regardless of which type of mandate they hold between elections. Thirdly, it can be assumed that it is in the parties best interests to select MPs who work hard in parliament, as long as they are loyal to the party. Also, if they want them to be elected, parties should put these legislators into positions that results in election with a rather high probability. In a mixed-member electoral system, especially with a large number of safe seats and closed party lists, the electoral performance of the candidates heavily depends on which district the MP was (re-)nominated 7 Records in the datasets represent individual candidates and MPs. Each candidate appears in the dataset only once (Weiss, 26, p. 24). 8 Records in the datasets represent individual candidates and MPs at the given election. Every candidate and MP has as many records as the number of elections he or she has participated in (Weiss, 26, p. 24). All descriptive statistics in this paper are calculated on the dataset in the long form. 9 Note, that MPs elected in 2 are not part of the analysis, as the election of 24 is not investigated due to the changing electoral rules. SMD MPs have more money at their disposal. Additionally, having to demonstrate that they also have a considerable local support, these representatives often have more leverage within their parties. 6

7 in, and where she is placed on the party list. In this paper, I will test the effect of legislator activities on the second set of variables, namely the MPs relative positions on the regional and national party lists at election t. Figure 2 visualizes the distribution of the continuous dependent variables. With respect to candidacy in SMDs, 39.8 % of the MPs were nominated in the constituencies at the next election (t)..5. Density SMD vote share at t Relative list positions are obtained by dividing absolute list positions with the number of candidates on the list. 7

8 Density Relative position on the regional list at election t Relative position on the national list at election t Figure 2. The distribution of candidate vote share, relative list position on the regional and the national lists Legislator activities I use several measures to measure legislator activities in parliament. MPs may engage in a wide range of activities. They attend plenary sessions, take the floor, sponsor bills, submit and table various types of questions or take part in the work of different committees. With the exception of attendance, information on all these activities is available on the official website of the Hungarian parliament. Table 2 lists variables that will be taken into account in the analysis as main independent variables. Again, the main question of this paper is if parliamentary activities affect the share of votes and candidacy in SMDs and list positions on the regional and national tiers. The majority of the activity variables are heavily skewed to the right. Therefore, with the exception of committee membership, the log transformed versions will be used in the analysis (for the distribution of the main independent variables see Appendix ). To keep observations with no activity, I added a very small number to zero (.) before logaritmizing the variables. The bivariate relationships between legislator activities and the dependent variables are shown in Appendix 2.a-d. Table 2. The list of legislator activities in parliament Variables Description Transformation Bills The number of bills sponsored by the MP per electoral term a Natural log Questions The number of questions submitted by the MP per electoral term b Natural log Speeches The number of speeches delivered by the MP per electoral term Natural log Floor time Time spent delivering speeches by the MP per electoral term (in minutes) Natural log Committees The number of committees in which the MP serves as a regular None member per electoral term a Co-sponsored bills are included b All question types are taken into account (interpellations, oral, written, direct). 8

9 Results To test the hypothesis of the paper, namely whether the intensity of parliamentary activities influences re-selection and re-election chances, I follow a multivariate approach. In all cases, I use multilevel regression to estimate the effect of the activity variables. However, as different types of variables are to be explained, three different approaches are implemented. First, in the case of candidates vote share, a simple multilevel linear model will be built. Second, SMD candidacy being a binary variable, multilevel binary logit regression is the proper choice. Third, as the distribution of relative list positions is skewed to the right both in the cases of regional and national lists, I run gamma multilevel regressions. For this list of control variable see Appendix 3. Table 3 presents the results of the models. Table 3. Multilevel models explaining candidate vote share, SMD candidacy, and relative positions on regional and national lists Variables Model a DV: Candidate vote share at t Coefficients Model 2 b DV: SMD candidacy at t Odds ratios Model 3 c DV: Relative regional list position at t Coefficient Model 4 c DV: Relative national list position Coefficient Activity variables Bills (logged) -.2 (.3).2 (.2). (.) -. (.) Questions (logged) -. (.4).3 (.2). (.). (.) Speeches (logged).32 (.9)***.3 (.6).2 (.2) -.6 (.2)** Committee -.47 (.3).98 (.4).4 (.6)**.9 (.6) Control variables Time (t) -.5 (.2).9 (.3)**.5 (.)***.3 (.2) Party vote share in SMD (t).88 (.3)*** SMD MP (t).58 (.5) 3.29 (.5)*** -.3 (.4).25 (.6) Regional list MP (t).38 (.).89 (.3) -.5 (.6).29 (.6)* SMD candidate (t-) 4.96 (.7)*** Tenure (t).24 (.33).66 (.9)***.3 (.6) -.3 (.8)* Joint candidate (t).7 (.88) Dominant party (t).3 (.85).2 (.33). (.4).67 (.22)*** Government party (t) -.3 (.64)**.6 (.28) -.26 (.)**.22 (.6) District competition (t) -.9 (.36)*** Multiple candidacies (t) (3.4)* Mayor (t).79 (.78)** 2.5 (.74)** -.25 (.3)* -.7 (.2) Minister (t) -.82 (.82) 5.3 (3.2)** -.4 (.32) -.36 (.23) Party leader (t) -.57 (.57).93 (.23) -.38 (.)*** -.59 (.5)*** Parliamentary position (t).44 (.66).49 (.63) -. (.24) -. (.3) Committee chair (t).2 (.45).29 (.33) -.4 (.9).8 (.3) Constant (243.65) 2.33e *** (44.94).85e+85*** N id sd(_cons) 2.4 (.5). (.4).36 (.6).59 (.7) sd(resid) 2.72 (.32) Log(pseudolikelihood) Wald χ *** 7.45*** 46.27*** 66.54*** *** p <., ** p <.5, * p <. a: Multilevel linear model (xtmixed) b: Multilevel binary logistic model (xtmelogit) c: Multilevel generalized linear model (meglm, family(gamma)) 9

10 Starting with candidate vote share (Model ), obviously, party vote share explains a huge proportion of the variance. %points increase in the party s vote share in the SMD increases the candidate s vote share by.88 %points. A perfect correspondence between the candidate s and the party vote share would be indicated by a coefficient around. The fact that the change in the party vote share does not induce the same size of a change in the candidate vote share, leaves space for other variables to step in. Among the key independent variables only the number of speeches appears to be significant. Its positive coefficient indicates that the more frequently an MP has the floor in the parliament, the larger the share of votes she can expect at the next elections. Besides the insignificant effects of the other activity variables, they do not point to the expected direction: the number of bills, questions and committee memberships effect vote share negatively. To disentangle the effect of questioning, I checked if the frequency with which MPs use the different question types influences their vote share at the next elections. The results are shown in Table 4. The coefficients of Model 5 suggest that generally there are no statistically significant differences between the various types of questions. However, we also find an exception: the number of direct questions slightly increases vote share. Table 4. The effect of the different question types on candidate vote share, SMD candidacy, and relative positions on regional and national lists Variables Model 5 a DV: Candidate vote share at t Coefficients Model 6 b DV: SMD candidacy at t Odds ratios Model 7 c DV: Relative regional list position at t Coefficient Model 8 c DV: Relative national list position Coefficient Parliamentary questions Interpellations (logged) -.2 (.3).3 (.2)*. (.). (.) Oral questions (logged) -.3 (.2).2 (.7). (.) -. (.) Written questions (logged). (.3). (.2). (.). (.) Direct questions (logged).6 (.3)**. (.2) -. (.). (.) Control variables included N id sd(_cons) 2.46 (.). (.4).36 (.6).64 (.7) sd(resid) 2.69 (.36) Log(pseudolikelihood) Wald χ ***.34*** 49.2*** 59.2*** *** p <., ** p <.5, * p <. a: Multilevel linear model (xtmixed) b: Multilevel binary logistic model (xtmelogit) c: Multilevel generalized linear model (meglm, family(gamma)) Apart from party vote share the government/opposition divide, district competition and mayoral positions affect candidate vote share. As to the government/opposition divide, voters appear to punish government MPs at the next elections. Karácsony (26) has shown that voters evaluate parties retrospectively on the basis of government performance. The reason for the negative effects is that Hungarian voters have unrealistic expectations of the governments ability to improve the economic environment 2. Therefore, voters may end up punishing governments that execute successful economic policies. Furthermore, as voters pursued trial-and-error tactics, a large share of votes was protest in nature, and at the same time, a proclamation of trust toward the new government (Karácsony, 26; Rose, 992). The effect of district competition is straightforward: the larger the number of effective candidates in the district, the smaller the 2 Duch (2) argues that in new democracies, voters cannot have established ideas on what to expect from governments.

11 share of votes a competing candidate may get. And finally, being a mayor boosts vote share by.79 %points, which suggests that voters take local attachments seriously when casting a personal vote. Turning to the candidate selection strategies of the parties, the results of Model 2 (Table 3) indicate that the observed parliamentary activity measures do not influence re-selection chances on the SMD level. Obviously, selectors prioritize SMD MPs and list MPs with previous SMDlevel electoral experience. Knowing the results of Model, it is also not unexpected that mayors are more likely to be nominated in SMDs than legislators with no local leadership positions. Ministers are also more likely to be given a shot at the SMD competition. Last but not least, somewhat surprisingly, tenure has a negative effect on re-nomination chances. The more experienced the MP is, the lower the probability of being selected in an SMD. This could indicate that parties constantly seek for amenable and potentially more successful candidates who are able to increase the district vote share by obtaining a larger amount of personal vote. These results indicate the parties take certain activities or skills (or both) into account when selecting candidates. But these activities (skills) are rather connected to other arenas of a legislator s work than activities in parliament. Although SMD candidacies may indicate who party leaders want to be re-elected, a better measure for these kinds of preferences in a mixed-member system is party list position. Most prominent partisans are very likely to be placed on the top of the lists either on the regional or the national level. Models 3 and 4 on Table 3 display results related to relative list positions. As the smaller the value of the dependent variable (i.e. the smaller the quotient of list position and the length of the list) the higher the MP is on the list. Therefore, a negative coefficient represents movement upward on the list. The results regarding parliamentary activities are rather ambiguous. The number of speeches legislators delivered during the electoral term seems to matter in the case of the national list composition. More speeches mean better positions on the national list. With respect to the regional lists, interestingly, the number of committee memberships has a negative effect on list placement. Paradoxically, the more an MP works in the committees the more unfavourable her position on the list. Apart from the activity-related measures, party affiliation and party leadership positions shape the ranking of the candidates on the lists. First, MPs of the dominant parties (Fidesz and MSZP) are placed lower on the list on average than the other parties candidates. The reason for this lies with the setup of the parliament during the three electoral terms under investigation. Dominant parties had substantially more MPs in parliament than any other parties. Therefore, while other parties could place their legislators on the top of the party lists, Fidesz and MSZP had to place at least a part of the re-nominated candidate pool down the lists. Thus, the average placement takes a larger value in their case. Second, party leaders are placed significantly higher on both the regional and national lists. The above findings are based on the variables presented by Appendix. Here, all MPs were taken into account, even if there is no registered activity. However, as zero cannot be logaritmized, zeros were added a very small number (.) to make logaritmization possible. Due to this transformation, the distributions of these variables are not continuous: they take a larger negative value in the case of inactive legislators, and are continuously distributed in the case of others. Thus, one unit change in the independent variable cannot be translated to every value of the variable. Therefore, to test the hypotheses (and confirm the above findings), I re-ran the analysis using only data on active members. Results are displayed in Table 5. The number of speeches is no longer significant (unlike in the cases of Models [candidate vote share] and 4 [relative national list position]), while bill-sponsorship matters in terms of regional

12 and national list positions. The more bills the MPs submit, the more advantageous their positions on the lists. Table 5. Multilevel models explaining candidate vote share, SMD candidacy, and relative positions on regional and national lists (Active MPs only) Variables Model 9 a DV: Candidate vote share at t Coefficients Model b DV: SMD candidacy at t Odds ratios Model c DV: Relative regional list position at t Coefficient Model 2 c DV: Relative national list position Coefficient Activity variables Bills (logged).88 (.342).25 (.25) -.77 (.52)*** (.76)*** Questions (logged) -.25 (.245).925 (.8).24 (.39).76 (.64) Speeches (logged).347 (.326).224 (.99). (.6) -.53 (.9) Committee (.336)*.4 (.22).47 (.7)**.77 (.79) Control variables included N id sd(_cons) 3. (.388). (.).335 (.66). (.) sd(resid) 2.33 (.539) Log(pseudolikelihood) Wald χ *** 72.3*** 43.3*** 98.7*** *** p <., ** p <.5, * p <. a: Multilevel linear model (xtmixed) b: Multilevel binary logistic model (xtmelogit) c: Multilevel generalized linear model (meglm, family(gamma)) Last, but not least, floor time was excluded from the models because it s high correlation with the number of speeches. Table 5 shows the results of models that include floor time as an independent variable. The number of minutes for an MP takes the floor is only relevant if nonactive MPs are taken into account. In this case, floor time influences candidate vote share in the SMD and the relative position on the regional party list. In the former case, the total length of the speeches affects constituency performance positively, while in the latter, MPs spending more time with speaking on the floor, get higher on the party list than those with less floor time. 2

13 Table 6. Multilevel models explaining candidate vote share, SMD candidacy, and relative positions on the regional and national party lists Variables Model 3 a DV: Candidate vote share at t (all MPs) Coefficients Model 4 a DV: Candidate vote share at t (active MPs) Coefficients Model 5 b DV: SMD candidacy at t Odds ratios Model 6 b DV: SMD candidacy at t Odds ratios Model 7 c DV: Relative regional list position at t Coefficient Model 8 c DV: Relative regional list position at t Coefficient Model 9 c DV: Relative national list position Coefficient Model 2 c DV: Relative national list position Coefficient Activity variables Bills (logged) -. (.3). (.34).2 (.2).7 (.2). (.) -.6 (.5)*** -. (.) -.27 (.8)*** Questions (logged) -. (.4) -.3 (.23).3 (.2).93 (.). (.).4 (.4). (.).7 (.6) Floor time (logged).29 (.8)***.6 (.3). (.5).27 (.9).2 (.2) -.4 (.6) -.6 (.3)** -.6 (.8) Committee -.47 (.3) -.62 (.34).98 (.4).3 (.2).4 (.6).5 (.7)**.7 (.6).7 (.8) Control variables included N id sd(_cons) 2.44 (.5) 3.3 (.36) 3.52e- (.4) 5.45e-7 (.34).36 (.6).34 (.6).62 (.6) 3.6e-33 (9.85e-3) sd(resid) 2.7 (.32) 2.29 (.5) Log(pseudolikelihood) Wald χ *** 75.22*** 8.4*** 72.94*** 47.6*** 42.22*** 65.75*** 98.74*** *** p <., ** p <.5, * p <. a: Multilevel linear model (xtmixed) b: Multilevel binary logistic model (xtmelogit) c: Multilevel generalized linear model (meglm, family(gamma)) 3

14 Conclusions In this paper I investigated the question if the intensity of legislator activities in parliament affects electoral performance at the next election. Electoral performance was perceived in two different ways. First, in the case of SMD candidates, SMD vote share was taken into account. Second, as in mixed-member systems, there are multiple ways into the parliament, I tested the effect of parliamentary activities on SMD nomination as well as regional and national list positions. In the case of SMD candidacy it was assumed that it serves as a reward mechanism to hard-working MPs. With regards to list positions, as party lists are closed, electoral performance depends only on two things: () the number of seats a party wins with the list, and (2) the rank of the candidate on that particular list. Thus, parties can decide who gets elected just by ranking candidates. The results are mixed at least. The number of speeches was found significant with regards to candidate vote share and national list positions, and points toward the expected direction. Furthermore, under certain circumstances, the number of bills also appears to have some effect on how party leaders construct party lists. Nevertheless, none of these results are consistent across different model specifications. Therefore, it cannot be confirmed that legislative activities have a clear impact on the electoral fate of the legislators. This means that MPs are not held accountable either by the voters or by their parties for how intensively they work in parliament. With other words, the goodness of the legislators are not assessed based on what they do in parliament, if it is assessed at all. It is not the promise of re-election that brings MPs to work hard, but something else. To identify these other motivations, more research is needed, probably qualitative in nature. Despite the insignificance of the main factors in this paper, certain findings of other nature may add to our understanding of how voters chose between candidates and along what lines parties construct party lists. We know now that voters value local attachment at the polling box. Additionally, the results indicate that parties take certain activities or skills into account when finalizing the order of candidates on the lists. However, these activities (or skills) are rather connected to other arenas of a legislator s work than parliamentary activities. Mayors and ministers have better chances to being selected to SMD candidacy, while party leaders are found at better positions on the party lists. And finally, a few words about the causality in the presented models. Although, it was assumed that there is a causal relationship between the independent and the dependent variables, one has to be very careful in interpreting positive effects under such model specifications. As a certain proportion of the MPs work in parliament is controlled exclusively by the parties, both the number of bills for example and the list positions may be influenced by a third factor, which makes parties selecting certain MPs to certain tasks. This third factor is probably the MPs position within the party which cannot be captured either by party leadership positions or any other political position that can objectively be measured. Nevertheless, negative results still suggest that there is no connection between legislative activities and electoral performance. With regards to future research, to disentangle the case of individual accountability, whether it exists and through what mechanism, one has to take other, non-parliamentary activities into account as well. As to activities in parliament, one should look behind the intensity and look at the content of those activities. Policy aspects as well as local orientation could be two directions in which we could investigate further to see if voters and parties react to the context behind sheer numbers. 4

15 Appendix Appendix. The distributions of the main independent variables Density Number of sponsored bills (logged) Number of questions (logged) Number of speeches (logged) Density Floor time (logged) Number of committees Appendix 2.a. The bivariate relationships between candidate vote share at election t and the main independent variables Number of sponsored bills (logged) Number of questions (logged) Number of speeches (logged) Floor time (logged) Number of committees 5

16 Appendix 2.b. The bivariate relationships between SMD candidacy at election t and the main independent variables Number of questions 2 (logged) 3 4 Number 2 of speeches 4 (logged) No Yes No Yes No Yes Floor time (logged) Number of committees No Yes No Yes Appendix 2.c. The bivariate relationships between relative regional list positions at election t and the main independent variables Number of sponsored bills (logged) Number of questions (logged) Number of speeches (logged) Floor time (logged) Number of committees 6

17 Appendix 2.d. The bivariate relationships between relative national list positions at election t and the main independent variables Number of sponsored bills (logged) Number of questions (logged) Number of speeches (logged) Floor time (logged) Number of committees 7

18 Appendix 3. The list of control variables Variable Description Time (t) Election (22, 26, 2) Party vote share in The vote share of the MPs party in the respective SMD at election t. SMD (t) SMD MP (t) The MP was holding an SMD mandate at the time of election t Regional list MP (t) The MP was holding a regional list mandate at the time of election t SMD candidates (t- The MP was an SMD candidate at the previous election (t-) ) Tenure (t) The number of terms served as an MP at the time of election t Joint candidate (t) The MP was nominated by multiple parties jointly in the SMD at election t Dominant party (t) MP of Fidesz or MSZP Government party MP of a government party at the time of election t (t) District The number of effective candidates in the SMD at election t; competition (t) N,where pi is the vote share of each candidate competing in 2 p i the respective SMD Multiple The MP was nominated on multiple tiers of the electoral system at candidacies (t) election t Mayor (t) The MP is a mayor at the time of election t Minister (t) The MP is a minister at the time of election t Party leader (t) The MP is a party leader at the time of election t Parliamentary The MP holds office in parliament as a PPG-leader, speaker, vice positions (t) president or clerk) at the time of election t Committee chair (t) The MP serves as a committee chair at the time of election t 8

19 References Anderson, W. D., Box-Steffensmeier, J. M. and Sinclair-Chapman, V. (23) The Keys to Legislative Success in the U.S. House of Representatives, Legislative Studies Quarterly, 28(3), Atmor, N., Hazan, R. Y. and Rahat, G. (2) Candidate selection, in J. M. Colomer (ed.), Personal representation. The neglected dimension of electoral systems. Colchester: ECPR Press, p Bowler, S. (2) Private Members Bills in the UK Parliament: Is There an Electoral Connection?, The Journal of Legislative Studies, 6(4), Bräuninger, T., Brunner, M. and Däubler, T. (22) Personal vote-seeking in flexible list systems: How electoral incentives shape Belgian MPs bill initiation behaviour, European Journal of Political Research, 5(5), Cain, B., Ferejohn, J. and Fiorina, M. P. (987) The personal vote: constituency service and electoral independence. Harvard University Press. Carey, J. M. (27) Competing Principals, Political Institutions, and Party Unity in Legislative Voting, American Journal of Political Science, 5(), Colomer, J. M. (2) Introduction: personal and party representation, Josep M. Colomer (ed.) Personal representation. The neglected dimension of electoral systems. Colchester: ECPR Press, p. 2. Cooper, C. A. and Richardson, L. E. (26) Institutions and Representational Roles in American State Legislatures, State Politics & Policy Quarterly, 6(2), Crisp, B. and Ingall, R. E. (22) Institutional Engineering and the Nature of Representation: Mapping the Effects of Electoral Reform in Colombia, American Journal of Political Science, 46(4), Curtice, J. and Shively, P. (29) Who represents us best? One member or many?, Hans- Dieter Klingemann (ed.) The Comparative Study of Electoral Systems. Oxford: Oxford University Press. Eulau, H. and Karps, P. D. (977) The Puzzle of Representation: Specifying Components of Responsiveness, Legislative Studies Quarterly, 2(3), Fenno, R. F. (978) Home Style: House Members in Their Districts. Glenview, IL: Longman. Gallagher, M. and Holliday, I. (23) Electoral Systems, Representational Roles and Legislator Behaviour: Evidence from Hong Kong, New Zealand Journal of Asian Studies, 5(), 7 2. Hazan, R. Y. and Rahat, G. (2) Democracy within Parties: Candidate Selection Methods and their Political Consequences. Oxford: Oxford University Press. Heitshusen, V., Young, G. and Wood, D. M. (25) Electoral Context and MP Constituency Focus in Australia, Canada, Ireland, New Zealand, and the United Kingdom, American Journal of Political Science, 49(), Hill, K. Q. and Hurley, P. A. (22) Symbolic Speeches in the U.S. Senate and Their Representational Implications, The Journal of Politics, 64(), Lancaster, T. D. (986) Electoral Structures and Pork Barrel Politics, International Political Science Review, 7(), Lancaster, T. D. and Patterson, W. D. (99) Comparative Pork Barrel Politics, Comparative Political Studies, 22(4), Lazardeux, S. (25) Une Question Ecrite, Pour Quoi Faire? The Causes of the Production of Written Questions in the French Assemblée Nationale, French Politics, 3(3), Manow, P. (23) Mixed Rules, Different Roles? An Analysis of the Typical Pathways into the Bundestag and of MPs Parliamentary Behaviour, The Journal of Legislative Studies, 9(3),

20 Mansbridge, J. (23) Rethinking Representation, American Political Science Review, 97(4), Mansbridge, J. (29) A Selection Model of Political Representation, Journal of Political Philosophy, 7(4), Marangoni, F. and Tronconi, F. (2) When Territory Matters: Parliamentary Profiles and Legislative Behaviour in Italy (987 28), The Journal of Legislative Studies, 7(4), Martin, S. (2) Parliamentary Questions, the Behaviour of Legislators, and the Function of Legislatures: An Introduction, The Journal of Legislative Studies, 7(3), Mayhew, D. R. (974) Congress: The Electoral Connection, Second Edition. Yale University Press. Mitchell, P. (2) Voters and their representatives: Electoral institutions and delegation in parliamentary democracies, European Journal of Political Research, 37(3), Norris, P. (997) The puzzle of constituency service, The Journal of Legislative Studies, 3(2), Norris, P. (24) Electoral Engineering: Voting Rules and Political Behavior. Cambridge: Cambridge University Press. Norton, P. and Wood, D. (99) Constituency Service by Members of Parliament: Does It Contribute to a Personal Vote?, Parliamentary Affairs, 43(2), Rasch, B. E. (29) Opposition Parties, Electoral Incentives and the Control of Government Ministers: Parliamentary Questioning in Norway, in S. Ganghof, C. Hönnige and C. Stecker (eds.), Parlamente, Agendasetzung und Vetospieler. VS Verlag für Sozialwissenschaften, pp Russo, F. (2) The Constituency as a Focus of Representation: Studying the Italian Case through the Analysis of Parliamentary Questions, The Journal of Legislative Studies, 7(3), Scholl, E. L. (986) The Electoral System and Constituency-Oriented Activity in the European Parliament, International Studies Quarterly, 3(3), Shugart, M. S. and Wattenberg, M. P. (2) Introduction: The Electoral Reform of the Twenty-First Century?, in M. S. Shugart and M. P. Wattenberg (eds.), Mixed-Member Electoral Systems. The Best of Both Worlds?. Oxford: Oxford University Press. Stratmann, T. and Baur, M. (22) Plurality Rule, Proportional Representation, and the German Bundestag: How Incentives to Pork-Barrel Differ across Electoral Systems, American Journal of Political Science, 46(3), Tavits, M. (29) The Making of Mavericks Local Loyalties and Party Defection, Comparative Political Studies, 42(6), Weiss, R. E. (26) Modeling Longitudinal Data. New York: Springer Science & Business Media. 2