Preferential Voting in Denmark: How, Why, and to What Effect? 1

Professor Jørgen Elklit, August 2, 2011 Department of Political Science Aarhus University, Denmark elklit@ps.au.dk Preferential Voting in Denmark: How, Why, and to What Effect? 1 Prepared for delivery at: 1. ECPR General Conference, Reykjavik, Iceland, August 24-27, 2011 (Panel: Electoral Systems and the Personal Vote) 2. APSA Annual Meeting, Seattle, Wash., USA, September 1-4, 2011 (Division 34, Panel: The Determinants of Candidate Choice in Preferential Voting Systems) Abstract: The paper provides an overview of (1) the Danish system for casting a preferential or personal vote in a proportional list system and (2) how seats are subsequently allocated. Denmark differs from most (all?) preferential list PR systems by letting the parties themselves (actually the parties multi-member constituency branches) decide which one of four possible preferential list options they want to employ in an upcoming election. The two dominant options are explained in some detail, to allow a full understanding of how they function. The paper then goes on to discuss how this system can be classified if one looks at the classification schemes proposed by in previous works in this field. The way forward is to understand that Denmark is not one, but at least two different but simultaneously used systems, which belong in different classification categories. The Danish PLPR system is then briefly compared to the Swedish and the Finnish systems before a few empirical results from Danish impact studies are presented. Finally, a model attempts to illustrate how the various explanatory factors interact to produce the parties votes in the Danish multi-member constituencies. 1 Constructive comments and advice from Søren Risbjerg Thomsen and Birgitta Wistrand are appreciated. 0

Preferential or personal voting in proportional list systems can be studied at various levels: Why is there one form in Denmark, another in Sweden, and a third variant in Finland and what are the effects of these differences on how party groups function in parliament (if any)? One can also analyze preferential voting at the party level are some forms more conducive to an increase in intra-party rivalry than other forms or why do some parties prefer one form and not another (if they as in Denmark have an option to choose)? Personal voting can also be studied as a phenomenon at the level of the individual voter. Why do some voters cast a personal vote and others do not, even when it is very clear that the personal votes might decide which candidates actually make it into parliament? This paper aims to document how the Danish sub-system of preferential voting functions. Subsequently, it will situate Denmark in its proper place among those electoral systems allowing at least some degree of preferential voting. The first step in this endeavor is to provide a description of the system, which will hopefully also allow other than Danish electoral experts to understand it. The next step will be to discuss the proper categorization of the system, while the final step is a brief comparison of the Danish system with the preferential voting systems in Sweden and Finland as that might provide useful additional insights. Preferential Voting in Denmark Denmark has three main electoral regions. Today they only comprise a total of ten multimember constituencies, which are still the central elements in the seat allocation system. The ten multimember constituencies each consists of a number of so-called nomination districts. The number of such districts range from two on the island of Bornholm in the Baltic Sea, which is a multimember constituency of its own, to 13 in the South Jutland multimember constituency. Every fifth year, 135 so-called constituency seats are allocated proportionally to the ten multimember constituencies based on the sum of (1) registered population, (2) number of voters in the most recent parliamentary election, and (3) the area in km 2 x 20 (in order to improve the level of representation of sparsely populated parts of the country). The previous allocation of seats to electoral units meant that the South Jutland multimember constituency in 2007 had 18 constituency seats ( in addition to this, South Jutland also got six compensatory seats, but this number followed from the actual allocation of the 16 compensatory seats allocated to the electoral region to which South Jutland belongs, see below) to allocate among parties. Most of the parties presented 13 candidates, one in each of the 13 nomination districts, and the pre-election appointment/nomination of these 13 candidates took place within each party s nomination district branch. Each party s final list of candidates for the multimember constituency was thus 1

primarily the product of 13 unconnected decision-making processes among party members and local party leadership. 2 As the nomination of candidates takes place at the nomination district level, and no formally structured party list is produced at the constituency level, there must also be a system in place to ensure that a party s set of candidates is presented to the voters in a manner that is fair to all the party s candidates, across nomination districts. The current system which has developed gradually since this system was introduced in 1920 (Elklit, 1993; 2005/2008) is the following: When a party is entitled to run in the elections, it presents its list of candidate names to the constituency s electoral authorities (indicating the nomination district where each candidate is the party s nominated candidate) together with the party s own decision on which system to apply for the eventual post-polling selection of candidates to fill seats. So how candidates are eventually to be elected which is also reflected in the way their names are printed on the ballot paper is decided by each individual party (either as part of the national party rules and regulations or as an ad hoc decision in the party s multimember constituency organization). This complicates the presentation of the system as well as the computations, but that s the way it is. It also means that the name order on the ballot papers within a multimember constituency differ between nomination districts. There are two main forms of list organization and presentation, each with two variants. Voters are in all cases entitled to cast one vote, either for the preferred party as such (a so-called party vote) or for one of the candidates running for that party (a so-called personal or preferential vote): 1. Standing by district (in Danish: Kredsvis opstilling) is the traditional form, with one of the party s candidates nominated in each nomination district. The name of the nomination district s own candidate is printed first in the party s section of the ballot paper (in boldface) followed by his/her fellow candidates in the nomination district in question). He/she receives all the votes cast for the party in the nomination district plus all the preferential votes cast for him/her in all nomination districts in the multimember constituency as all the constituency s voters can cast such a vote for him/her. Candidates from the other nomination districts appear in alphabetical order after the nomination district s own candidate. However, 2007 only saw one single occurrence of this form of presenting candidates out of 90 possible occurrences (nine parties x ten multimember constituencies). It shows that this form is no longer so fashionable with the parties as it was when preferential voting in Denmark was first studied (Pedersen, 1966). 2 One of the consequences of the local government restructuring reform in 2005 was that the magnitude and number of some of the electoral administrative units changed, effective from January 1, 2007. Therefore, some of the numbers given below differ from numbers found in expositions about the pre-2007 situations. 2

A party running in this standing by district -way can further indicate that it wants to present its candidates in a particular order decided within the party. This form of list organization is termed party list, and the ranking of candidates is normally decided by a ballot taken among the party s members in the multimember constituency (of course prior to the registration dead line). This system reduces the voters influence on the selection of candidates considerably, as is also the intention (see Elklit, 2010 for details). This option is not easy to identify on the ballot paper, but if the first candidate s name is in boldface and the following candidates are not in alphabetical order, we have the party list form of list organization (as an example see party F on the copy of a ballot paper on page 7). As it is difficult for the voters to change this list order, it comes close to being a closed list, but without formally being so. It has happened, however, that voters have changed the list order. There were only 12 occurrences of the party list organization out of the abovementioned possible 90 in 2007, all among the two leftist parties, so it is probably more relevant to say out of 20 occurrences. In seven of these 12 cases seats were won without changing the list order. 3 Table 1 illustrates this way of doing things: Mr Frank Aaen, elected for Enhedslisten (the Unity List) on a compensatory seat (T), received a total of 3,176 votes, which is the sum of all his preferential votes across the eight nomination districts (348 + 267 + + 254). However, in his own nomination district (# 3), the figure printed in boldface (1,145) is the sum of his 559 preferential votes (see the bottom row, disregarding two independent candidates with very few votes) and the 586 non-preferentialvotes cast for the party as such ( partistemmer ). Since Mr Aaen s vote total was less than a (Droop or Hagenbach-Bischoff) quota calculated for this particular form of list organization (3,917, see Fordelingstal in the lower left corner), the party list order is followed and the party s one seat in this multimember constituency goes to no. 1 on the list, while the other candidates become substitutes in the order of the list (the right-most column). Candidates with more votes than the quota are elected irrespective of their position on the party list, but this only happens rarely. 3 In the three cases, where the Socialist People s Party (SF) won seats using party list, there was some disagreement between the list order and the total number of votes for individual candidates, but not to such a degree that the list order did not prevail. But only in one case (Sydjyllands Multimember Constituency) did an elected candidate obtain fewer personal votes than a non-elected candidate. 3

Table 1. Copy of a result table for the Unity List from a multimember constituency in Metropolitan Copenhagen (Suburbs of Copenhagen) (disregard the two last rows) 2. Standing in parallel (in Danish: Sideordnet opstilling) is what Katz (1986: 89, following Pedersen, 1966) calls simultaneous list ordering. Over recent decades, this form of list organization has become increasingly popular, and in 2007 it accounted for 77 of the 90 possible cases). The basic idea is that all the party s candidates in a multimember constituency formally stand in all the nomination districts, but the candidate from the party branch in the nomination districts is nominated and therefore has his/her name printed first on the party s section of the ballot paper in the nomination district in question, while the other names come in alphabetical order. They all have their names printed in boldface, which signifies that they all stand in the nomination district in question (see parties A, B, C, D on the ballot paper copy on page 7). All candidates get all their individual preferential votes from the entire constituency, but the key point is how the party votes in the nomination districts are distributed among them. As they all stand in all nomination districts, party votes in each and every nomination district is allocated to each and every candidate in proportion to his/her preferential votes cast in that particular nomination district. So each candidate s total vote under this form of list organization is the sum of (up to) 13 piles of preferential votes one from each nomination district and (up to) 13 piles of party votes, namely his/her shares of the party votes in each and every nomination district in the constituency. 4

Under this form of list organization, it is also possible not to have nominated a top candidate in the individual nomination districts, but this option is only rarely used and did not occur in 2007. 4 Table 2 is from the same multimember constituency as the one above and shows the votes for the Social Democrats, who here won four constituency seats (marked with a K). The Social Democrats have decided nationally to use the most usual form of list organization, i.e. the standing in parallel, in all the country s multimember constituencies. This constituency has eight nomination districts, and the party has eight candidates, one nominated in each district. In the upper half of the table, the numbers in italics indicate in which district the candidates are nominated. The table s lower half gives the preferential votes ( personlige stemmer ) for all candidates across all nomination districts, while the row above that ( partistemmer ) gives the total number of votes cast for the party in each nomination district. Table 2. Copy of result table for the Social Democrats in a multimember constituency in Metropolitan Copenhagen (Suburbs of Copenhagen) 4 Under this system, a candidate can also be nominated in more than one nomination districts. This happens rarely, but it is a possible solution to two different problems: Too few candidates or a strong wish to secure the election of a particular candidate. 5

The Social Democrats here had four relatively high profile candidates and four relatively low profile candidates. Ms Mette Frederiksen is a highly respected young politician, who was the party s nominated candidate in nomination district # 8 (Ballerup), as can be seen by her vote total (13,632) in italics. In the lower half of the table, one can see that she obtained a total of 27,077 personal votes across the eight nomination districts (1,772 + 1,994 + 9,165), i.e. most in her own nomination district, where she really took the lion s share of the 10,040 personal votes. Her share of the personal votes in the other seven nomination districts was also very high, so her share of the party votes across nomination districts became 20,757 (e.g. 4,467 in her own nomination district, calculated as 9,165/10,040 x 4,893). Consequently, Ms Frederiksen filled the first of the party s four constituency seats in this constituency as her vote total was higher than any of her party comrades vote totals. The other seven Social Democrats became either MPs or substitutes in declining order of their total vote. It has not been possible to obtain a copy of a ballot paper from the 2007 election. Therefore, one finds below is a ballot paper from the 2001 elections (the upper half of it) which shows how the two currently important forms of list organization is presented to the voters in a nomination district. In this case, it is the 3 rd nomination district in the (then) Southern Jutland multimember constituency (see upper right corner). The text above has already clarified what one (and the voters) should look for: What names are printed in boldface and whether the names after the top name are printed in alphabetical order or not? So in this multimember constituency A, B, C, and D used the standing in parallel way of presenting candidates in 2001, while F had decided to use the standing by district way, and more specifically the party list way of organizing the list. 6

How to Classify the Danish Preferential Voting System? The Danish preferential voting system has apparently been difficult for comparativists to deal with and it was not included in two of the more ambitious studies (Carey & Shugart, 1995; Shugart 2005/2008). At the same time, it is regrettable that the first attempts to include Denmark in comparative studies of preferential voting (Marsh, 1985; Katz, 1986) are now somewhat outdated because of (1) changes in the rules for allocating seats if party lists are used and (2) shifting 7

relative occurrences of the various forms of list organization. The same goes for Pedersen (1966), even though this analysis of the situation in the early 1960s is still of value. Katz (1986), however, is still a useful contribution to the study of preferential voting, i.a. because of his observation that *t+he most important characteristic distinguishing among preference voting systems concerns the preference 5 votes not explicitly cast for any particular candidate (p. 93). The two mains forms of list organization in Denmark differs in exactly this respect, so Katz hits the nail on its head with this observation, as did Marsh when he discarded Bogdanor s older attempt of providing a general typology for preferential voting systems. According to Marsh there are only two types of preferential voting systems. In the first, the voters alone decide which candidates shall fill the seats won by a party; in the second, the decision results from a combination of party ordering and voter choice normally with the weight on the latter factor (Marsh, 1985: 366). The main forms of list organization used simultaneously in Denmark perfectly describe the two types suggested by Marsh, even though the Danish party list form of list organization puts considerably more weight on the party ordering factor than on the voters choice. Later, Carey and Shugart (1995) proposed a method to study the relative value of personal reputation and party reputation to candidates for election under different electoral systems and with different provisions for preferential voting. Three trichotomized ordinal variables in combination with district magnitude were used to develop an ordinal scoring system of electoral systems according to the incentive to cultivate and exploit a personal reputation (p. 424ff). The system is interesting and the authors invite colleagues to test and nuance it in future research at the same time, they are well aware of the difficulties in measuring the dependent variable, i.e. the value of personal reputation to candidates for elective office. Denmark is not one of the countries used to illustrate the combinations of variables. This is understandable as Denmark has the above-described system that allows parties four different options at the multimember constituency level (even though some of the parties internally have decided not to give their lower levels the freedom to choose). However, the problem is diminished considerably as all four options would lead to Denmark being classified in Carey and Shugart s category i (2,0,2; for details refer to Carey and Shugart, 1995). But, even within this category, there is some variation as the two main forms of list organization in Denmark do not provide similar incentives to cultivate and employ a personal reputation over and 5 The word preference should probably be deleted here, as the text makes considerably more sense without it. 8

above the party s reputation. The party list form of list organization provides much less incentive for developing and cultivating a personal reputation than the standing in parallel form. The reason is, of course, that it is difficult for individual candidates to reach the vote quota, which is the only way one can get elected irrespective of one s place in the rank order established by the party. The example above illustrates this well as Mr Aaen is a well known and respected long time representative of his party. So, even for a candidate of Mr Aaen s stature and reputation, the incentives to further cultivate one s personal reputation vis-à-vis the party s are almost nonexisting. More recently, Shugart has provided a useful overview of the field of comparative electoral systems research (Shugart, 2005/2008). The section on intraparty effects is particularly helpful as it brings more clarity to the topic than one has seen previously, not least by separating out preferential list proportional representation systems (PLPR) as a particularly important category. Shugart then proposes a typology for systems falling in this category and ends up with two main categories: open lists and flexible lists. The distinction between these two kinds of PLPR corresponds well with not only the views presented by Marsh and Katz earlier, but they also cover the two main Danish types perfectly, as the standing in parallel -form should be categorized as an open list, while the party list -form is a flexible list. 6 Using this terminology, we can now state that the Danish system is a PLPR system, where party organizations at the multimember district level have the option of choosing 7 between either presenting an open list (the standing in parallel form, as described in some detail above) or presenting a flexible list (the party list form, as also described in some detail above). The former is probably the most difficult to understand, while the latter is more by the book. This categorization covers the two main forms, while the two forms not used currently are disregarded. 8 Shugart has a fine formulation, which deserves to be cited in full: The distinction between open and flexible lists is subtle but crucial. In effect, in the flexible list the voter who opts not to cast a preference vote is delegating to the party the task of deciding the order in which the candidates will be elected, while in the open list, the voter is delegating this decision to the other voters (those who cast preference votes. (Shugart 2005/2008: 43: italics in original) 6 Shugart (2005/2008: 42) says that the term flexible list is borrowed from Marsh (1985). However, Marsh does not really use it, at least only with reference to Bogdanor, whose typology he says obscures important differences. 7 Provided that the national party organization has not decided what form of list organization all the party s multimember constituency level organizations must use. 8 The variant of standing in parallel without nomination would also be an open list, while the traditional form (Pedersen, 1966; Katz, 1986) is less easy to categorize as either an open or a flexible list, but it should probably be placed as an open list. 9

Shugart then goes on to present some issues in empirical research on the intraparty dimension. In the first part, he deals with socio-economic effects and in the second with the impact of the personal vote, primarily building on Carey and Shugart (1995). There is not much empirical research on PLPR systems and their impact on (1) voting behavior (What explains differences in the level of the use of preferential votes?) and (2) candidate behavior (What level of intraparty conflicts emerge under various PLPR systems; What role does personal reputation actually play for the eventual outcome?). The scarcity of empirical studies and in particular of comparative studies calls for further research, even though the field is quite complex because of the many complex interrelations between variables, as demonstrated by the discussions in the literature referred to above. To take just one example: Tables 1 and 2 gave the details of intraparty seat allocation for two Danish parties, one (the Unity List) using using the party list form of list organization (or a flexible list, following Shugart s terminology), the other (the Social Democrats) using the standing in parallel form, i.e. an open list in Shugart s terminology. Since voters voting for the Social Democrats would have more real influence on the selection of candidates among the party s eight candidates than the voters for the Unity List could expect as the election of Mr Aaen was expected by all, party members as well as external commentators it is surprising that the two parties had the same percentage of voters who decided to cast a personal vote: 55 per cent among Social Democratic voters and 54 per cent among Unity List voters. The Social Democrats also had high profile and well respected candidates running in this constituency (in particular Ms Frederiksen and Mr Lykketoft) so why this similarity in the level of personal voting in this multimember district (socio-economic and structural differences are thereby controlled for) at the same election? If we for a moment return to table 2 showing the Social Democrats votes and candidate selection, we can focus on the candidate nominated in nomination district # 1, Ms Christel Schaldemose. Her personal vote total was 729, of which 487 were from her own nomination district. But 1,772 voters in that district voted personally for Ms Frederiksen and 617 for Mr Lykketoft, so she was only third in her own nomination district. 48 per cent of the Social Democratic voters in nomination district # 1 decided to vote for the party as such, while the corresponding figure in nomination district # 8 was only 33 per cent. In the remaining districts the percentage oscillates between 42 and 51. It is indeed surprising at least to this observer that so many voters do not use the personal vote option, either to express a personal sympathy for one of the candidates or to try to act tactically, i.e. forecast who would most certainly be elected (Frederiksen and Lykketoft) and then vote for one s most preferred candidate among the other six candidates. Voting for the two best known and most respected candidates would only add to their surplus of votes, and therefore be less effective than if being used on one of the other candidates. But it is evident that Ms Frederiksen s 10

very impressive number of votes is explained not only by her ability to attract personal votes from voters in other nomination districts than her own, but also by her (and her campaign workers ) ability to convince Social Democratic voters in her nomination district that they should cast a personal vote for her instead of voting for the party (or for one of the other Social Democratic candidates). The more general picture is that 51 per cent of all Danish voters in 2007 did cast a personal vote. The variation across parties and multimember constituencies is, however, considerable: The Conservative Party obtained relatively most personal votes (59 per cent), and the Social Liberals relatively fewest (41 per cent). Relatively most personal votes were cast in the small Bornholm multimember constituency (63 per cent), relatively fewest in the one from which the two examples above were taken, i.e. the suburban area around Copenhagen (44 per cent). If we take a closer look at parties and multimember constituencies together, it becomes evident that most of the cases with high percentages of personal votes (65+ per cent) are due to a national party leader running in that multimember constituency (Folketingsvalget den 13. november 2007, p. 49). 9 Marsh (1985: 376) has argued that the best explanation of variation in the use of the personal vote option is the likely effectiveness of such votes. This is not borne out in the Danish case, and certainly not in 2007: For the Socialist People s Party the average for the three multimember constituencies with a party list was 46.1 per cent personal votes. And in the seven constituencies with the standing in parallel system, the average was 39.2 per cent, i.e. exactly the opposite of what one should expect according to Marsh. And the Unity List used the party list system in all ten constituencies, but the national average of personal votes was still 50.0 per cent, virtually the same as the overall national average for all parties (50.8 per cent). So the picture is very scattered, and it might be wisest not to attempt a conclusion. If one should, nevertheless, dare do that, it appears that the main motive behind personal voting is candidate reputation (in the broad sense used by Carey and Shugart), which in many ways goes against the thesis of the rational voter (or at least broadens the concept of utility): If one votes for a wellknown candidate with a strong reputation, one is rewarded by seeing this person elected, but he or she would probably have been elected in any case. So the reward is primarily emotional, while one could have benefitted more from casting a personal vote, if the vote had been cast for a candidate with less certainty of being elected, as that would have increased the voter s influence on the overall composition of the party group in parliament. A further and rational motive could also be to vote for a candidate (who would certainly be elected in any case) in order to improve that particular candidate s standing/reputation in the party caucus in parliament. 9 High percentages are also found in Bornholm and Western Jutland multimember constituencies with no party leaders running, but because of a certain local/regional tradition for casting personal votes. 11

Marsh has more recently studied whether Irish voters in 2002 were more candidate- or more party-centered in their voting (Marsh, 2007). The various measures all point towards a strong degree of candidate-centered voting, but whether this is primarily due to the STV electoral system as such or is a more general feature of preferential voting systems is not discussed. Shugart may actually have a very valid point when he suggests that there is a fundamental difference between what he terms preferential-list systems (such as the Danish) and the various non-list systems (such as FPTP, STV, SNTV etc.) with regard to the interaction between inter- and intraparty dimensions, i.e. how the systems impact on the distribution on votes among parties and among candidates within the parties (Shugart, 2005/2008, pp. 36-41; see also Carey and Shugart, 1995: 425-431). Carey and Shugart (1995) did not include Denmark among the electoral systems they scored according to the incentive to cultivate a personal reputation, maybe because Danish parties have the option to choose among various ways of presenting their lists and having seats allocated by different rules. However, the more and more popular standing in parallel form fits Carey and Shugart s category i well, while the other main form, the party list falls in the e -category (1-0-2). 10 Finland is also in the i -category, while Sweden which has changed its preferential voting system after (but not because of) the publication of Carey and Shugart s paper should be in the e -category. Carey and Shugart expect the incentive system and district magnitude to interact in such a way that the higher the district magnitude and the later in the alphabet the category is indicated, the higher is the value of personal reputation (as measured by the proportion of personal votes). When comparing Denmark and Sweden, the table gives some support to Carey and Shugart: In the e-category, Denmark has much more personal voting, (maybe) partly reflecting the higher district magnitude, while the difference within the Danish cases is small, but in the expected direction with a (slightly) higher level of personal voting in category-i cases than in category-e cases. So the expectation holds. Level of personal voting (in per cent) as an indicator of the incentive to cultivate a personal reputation, by incentive system and country Category (according to Carey and Shugart, 1995) Sweden Finland Denmark 2007 i (2-0-2) - 100 About 51 (77 cases) e (1-0-2) 20-25 P - About 48 (12 cases) Average district magnitude (all seats) 12.0 13.3 17.5 Average district magnitude excl. small islands forming their own district 12.4 14.2 19.2 10 For details regarding categories and scoring, please refer to Carey and Shugart (1995). 12

The Finnish system is so different from the Danish that the conclusion of this comparison is that the specificities of the various systems may explain much more than, e.g., district magnitude, which of course is a serious complication for comparative research (Raunio, 2005/2008, in particular pp. 478-482; see also Karvonen, 2011). It probably remains an open question whether the impact of personal votes is greater in Denmark than in Finland or if it is the other way round. The conclusion here appears to be that it is difficult to compare even within categories, because district magnitude is less important than, e.g., specificities of the preferential voting system as such. To some degree, this conclusion parallels Swindle s. In a comparison of Japan and Ireland also inspired by Carey and Shugart s suggestion for rank ordering of electoral formulas he came to realize that the role of party leadership should not be underestimated as an explanatory factor reflecting also the different ways STV and SNTV electoral formulas work (Swindle, 2002). Carey and Shugart s categories should maybe be further developed in order to pay more attention to the specificities of electoral systems as that appear to be a soft spot in the categorization system? Or would that be the end to the comparative study of preferential voting systems? Empirical Results from Danish Impact Studies The consequences of intraparty preference voting are more difficult to assess than are the consequences of most other aspects of electoral law (Katz, 1986: 93). Only very few studies of the effects of the Danish system for preferential voting have been conducted, one reason being that the many variants across the country do not combine well with random sampling of voters for survey studies, as very few voters will have been in sufficiently similar contexts to allow comparisons across parties and counties. Therefore, one has traditionally had to rely on official statistics (Pedersen, 1966; Thomsen and Elklit, 2007; Elklit, 2008), as it has only recently been attempted to study the phenomenon using multi-level analysis (Lysgaard, 2010; Thomsen and Sloth, in print 2012). The main conclusion of these studies is that in many cases it is very likely that the candidate s personal reputation has an independent impact on the total vote obtained by the party, i.e. personal votes and party votes combined (Thomsen & Elklit, 2007: 331). The method used in the latter study was to compare the development across multimember constituencies in personal votes and party votes. The picture, which emerged, was fairly easy to interpret when popular politicians either switched from one multimember constituency to another or withdrew from politics. One drawback to this dynamic approach is, however, that candidates who win a large but unchanged vote share in their constituencies in subsequent elections are easily disregarded. The main result namely that candidate reputation can independently of other factors 13

increase a party s total vote remains unchallenged. A recent study based on local elections has confirmed the effect of personal votes on the vote total for parties (Thomsen & Sloth, in print 2012). Lysgaard (2010) has provided an interesting multilevel analysis that aims to explain the voters individual choice of party by comparing the effect of the candidates personal reputation vis-à-vis other factors, using the parties share of personal votes as a proxy for the candidates personal reputation. Measured in this way, the candidate s personal reputation apparently has a highly significant, but relatively low positive effect on the voters choice of party. It complicates things that party leaders only run in one multimember constituency each, i.e., it is difficult to separate the party leader effect from the candidates personal reputation. And also that voters strongly attracted by the party leader (as opposed to local candidates) tend to vote for the party in those nine (previously 16) multimember constituencies where the party leader did not run for election. A further complication not mentioned by Lysgaard is that party leaders tend to run in the same usually large multimember constituency where the party leader effect is expectedly most useful for the total national party vote. Conclusion It has been a daunting challenge to try to present the Danish preferential voting system in such a way that it fits into existing proposals for classification and interpretation. This endeavor was not easy as Danish parties have the option to choose between four different ways of presenting their candidates to the voters, with different consequences for ballot design and post-election systems for allocating seats in parliament to individual candidates. It appears that personal reputation plays a considerable role when voters decide how to vote, irrespective of what candidate presentation system the party in question has chosen for the multimember constituency in question. This clearly overrides the importance of whether we look at standing in parallel or standing by district (and combine that with a ranked party list). Party leaders, however, get many more personal votes than most other candidates, which means that being a national party leader must be included among the elements combining the candidate s personal reputation, which is then measured indirectly through the person s ability to attract personal votes. Comparative studies are difficult to conduct, i.a., because of the differences in the way the electoral systems and the preferential voting part of it function. Country studies are also not easy to conduct due to the difficulties in separating candidate-centered voting from partycentered voting. To use one of the above examples: Does a personal vote for Ms Mette Frederiksen indicate that the voter voted for her 14

(1.) because of her personal reputation as young, dedicated, and concerned about social problems (but not thinking of her as a Social Democrat), i.e. primarily candidate-centered voting? (2.) because she was the local Social Democratic party branch s nominated candidate, i.e. personal voting, but nevertheless strongly party-centered? (3.) because, in the voter s opinion, she was the best Social Democratic candidate in the constituency, i.e. the voter conducts a kind of ranking of all eight candidates, i.e. a more conscious combination of party- and candidate-centered voting? (4.) because she might, at some point in time, challenge the national party leader. To some degree, the voter thus voted against the national chairperson, Ms Thorning-Schmidt, i.e. another form of party- and candidate-centered voting? There might be other reasons as well, but the four above suffice to make the point which is that one cannot do much without survey research, which is almost impossible in Denmark because of the variations across parties and multimember constituencies and the set of candidates available to choose from in a particular constituency. But at least one can hope that further studies using various multi-level methods might provide some of the answers we are still looking for! 15

Attempt to illustrate the interconnectedness of systemic factors, candidate incentives and behavior, and the interplay between personal votes, party votes, and total votes 16

References Carey, John M. and Matthew Soberg Shugart (1995). Incentives to Cultivate a Personal Vote: a Rank Ordering 1920 of Electoral Formulas, Electoral Studies, 14(4): 417-439. Elklit, Jørgen (1993). Simpler than Its Reputation: The Electoral System in Denmark since 1920, Electoral Studies, 12(1): 41-57. Elklit, Jørgen (2005/2008). Denmark: Simplicity Embedded in Complexity (Or is it the Other Way Round?), pp. 453-471 in Michael Gallagher and Paul Mitchell (eds), The Politics of Electoral Systems, Oxford: Oxford University Press. Elklit, Jørgen (2008). Valgkredsreformen 2006 og folketingsvalget 2007: En første evaluering, pp. 233-249 in Karina Kosiara-Pedersen & Peter Kurrild-Klitgaard (eds), Partier og partisystemer i forandring. Festskrift til Lars Bille, Odense: Syddansk Universitetsforlag. Elklit, Jørgen (2010). Valgsystemerne, pp. 15-79 in Anne Binderkrantz et al. (eds), Det demokratiske system. Valg, Folketinget, administrationen, organisationerne, 2nd ed., Copenhagen: Hans Reitzel. Folketingsvalget den 13. november 2007. Danmark, Færøerne, Grønland, Copenhagen: Indenrigsog Socialministeriet. Published October 2009. Katz, Richard S. (1986). Intraparty Preference Voting, pp. 85-103 in Bernard Grofman and Arend Lijphart (eds), Electoral Laws and Their Political Consequences, New York: Agathon Press. Karvonen, Lauri (2011). Preferential Voting in Finland. How much do candidates matter, and to whom and why? Paper prepared for the Annual Meeting of APSA, Seattle, Washington, September 1-4, 2011. Lysgaard, Jesper (2010). Hvad betyder kandidaternes personlige egenskaber for vælgernes individuelle partivalg? En multilevelanalyse af folketingsvalget i 2007. Term paper, Department of Political Science, Aarhus University. Marsh, Michael (1985). The Voters Decide?: Preferential Voting in European List Systems, European Journal of Political Research, 13(4): 365-378. Marsh, Michael (2007). Candidates or Parties? Objects of Electoral Choice in Ireland, Party Politics, 13(4): 500-527. Pedersen, Mogens N. (1966). Preferential Voting in Denmark: The Voters Influence on the Election of Folketing Candidates, Scandinavian Political Studies, 1: 167-187. 17

Raunio, Tapio (2005/2008). Finland: One Hundred Years of Quietude, pp. 473-489 in Michael Gallagher and Paul Mitchell (eds), The Politics of Electoral Systems, Oxford: Oxford University Press. Shugart, Matthew Søberg (2005/2008). Comparative Electoral Systems Research: The Maturation of a Field and New Challenges Ahead, pp. 25-55 in Michael Gallagher and Paul Mitchell (eds), The Politics of Electoral Systems, Oxford: Oxford University Press. Swindle, Stephen M. (2002). The Supply and Demand of the Personal Vote. Theoretical Considerations and Empirical Implications of Collective Electoral Incentives, Party Politics, 8(3): 279-300. Thomsen, Søren Risbjerg & Jørgen Elklit (2007). Hvad betyder de personlige stemmer for partiernes tilslutning?, pp. 307-334 in Jørgen Goul Andersen et al. (eds), Det nye politiske landskab. Folketingsvalget 2005 i perspektiv, Aarhus: Academica. Thomsen, Søren Risbjerg & Christian Eg Sloth (in print, 2012). Valgsystem, personlige stemmer og partivalg, in Jørgen Elklit & Ulrik Kjær (eds), KV09. Kommunalvalg i Strukturreformens skygge (draft title), Odense: Syddansk Universitetsforlag. 18