Electoral Systems Matter

An Experimental Examination of Different Systems         Richard J. Timpone


        Concern about the influence of electoral systems has grown in the United States in recent years, because of a number of factors, including the decision by the Supreme Court in Shaw v. Reno, the rulings by federal courts dealing with cumulative voting in Maryland and the failed nomination of Lani Guinier to the Department of Justice.
        This increased interest is not purely academic because decisions regarding the implementation of electoral systems have substantial consequences for political representation. It is no surprise that the rules of a game can affect its outcome. In the realm of elections, however, the distinct outcomes resulting from different 'rules' can have major implications for governmental policy.
        While it is clear that the electoral structure can constrain and influence election outcomes, the exact influence is often difficult to gauge. This is due to the simple fact that no two elections are exactly alike. Each election has its own dynamics resulting from the interaction among the candidates running, the campaign issues, and the cultural context.
        Even two elections in the same community at two points in time will have different dynamics. The difficulties are even greater when comparing elections across areas that have different political cultures, as is most apparent in cross national comparisons. It can therefore be quite difficult to identify with any precision the impact of electoral rules on outcomes and representation.

In the realm of elections, the distinct outcomes resulting from different "rules" can have major implications for governmental policy.

An Experimental Approach
        In order to clarify our understanding of the true impact of electoral institutions on behavior, it is useful to supplement comparisons of actual electoral results with more controlled studies. In the fall of 1994, an experiment was conducted that attempted to answer some of the questions about the relationship between individual political behavior and the electoral systems of single member districts, at-large systems and cumulative voting.
        In essence three mock elections were conducted
where all aspects, such as the information available and the candidates involved, were controlled and equivalent across the experimental conditions. The only variation was the electoral system in which the participants (353 undergraduates of State Univ. of New York-Stony Brook) evaluated the information.
        The participants in this experiment were presented with electoral guides that they were told to evaluate to determine if they were useful and whether their counties should provide them for voters. These flyers were constructed for the experiment with three sections: the first urged people to vote, the second described the electoral system in the county and the third section gave brief biographies of the six candidates.
        The first and third sections of all flyers were identical, and only the description of the electoral system was systematically manipulated. After answering a preliminary set of questions, the participants in the study were told to imagine they were voters in Sullivan County (the county presented in the flyer) in 1993, and asked how they would have cast their ballots. After casting their hypothetical ballots, participants were asked a battery of questions including those tapping their satisfaction with the systems and the difficulty they had with them.
        The hypothetical legislature in the study consisted of nine members. In the single member district case, participants could choose among the six candidates for the representative of their district. The participants in the two multi-member conditions -- at-large and cumulative voting -- were informed that the legislature used a staggered system with three year terms and three seats up for election each year. Thus, these participants could cast three ballots and were told there would be three winners. The candidates were an anti-corruption reform Democrat (Anderson), an anti-crime Republican (Dallin), an experienced Democrat (Froehlich), an African-American Independent (Morris), a business leader Republican with prior political experience (Nikkels) and a female Independent (Williams).


Election Outcome by System

Candidate (party id.)

Single Member District

At-Large Election

Cumulative Voting

Percent of People who Plumped

Anderson (Dem)

11.9

15.0

16.0 (3)

30.8

Dallin (Rep)

22.0 (2)

22.0 (1)

28.7 (1)

48.4

Froelich (Dem)

28.0 (1)

17.4 (3)

12.3

12.5

Morris (Ind)

7.6

8.6

6.2

40.0

Nikkels (Rep)

20.3 (3)

19.6 (2)

23.5 (2)

48.7

Williams (Ind)

10.2

17.4 (3)

13.3

25.6

n-voters

118

116

110

-

n-ballots

118

327

324

-


Impact of Electoral System on Election Outcomes

        It is appropriate that the bulk of the work on elections has focused on their impact on electoral outcomes because elections are the central mechanism for translating citizen preferences into representation in modern democracies.
        Because all other factors have been controlled,
this experiment permits a clean test of how different electoral systems affect election outcomes. Treating each of the experimental conditions as a separate 'election' allows for comparisons in the context of the current study. Table 1 provides the breakdown for the candidates in each of the electoral conditions.
        The first finding from Table 1 worth noting is simply that the rank ordering of the candidates and thus the election outcomes did differ across the electoral systems. As noted earlier, the rules of the game affects who wins, and this is clearly demonstrated in the results.
        These results show that the differences in how the systems translate preferences into representation may have severe consequences for the electoral outcomes. This can be seen where the individual winner in the single member district system (Froelich) wins none of the three seats available in the cumulative voting condition.
        In addition, the study demonstrates how some candidates with smaller cores of supporters may benefit from multi-member systems and thus broaden the base of citizenry represented in legislative bodies. I will briefly summarize the results here as they pertain to two of the hypothetical candidates, Froelich and Williams, because they present clear demonstrations about the nature of representation as mediated by the rules of the game.

Questions About a Plurality Winner

        The most dramatic differences between the different conditions can be seen for Froelich, the experienced Democrat, who won the plurality of the vote in the single-member district condition and tied for third in the at-large condition, yet came in fifth place in the cumulative voting condition.
        These distinctions highlight the concerns many have about plurality, single member district systems: a candidate can win who has a plurality of support, even if (1) this support is weaker than that held for other candidates, and (2) the candidate is considered a weak alternative among non-supporters.
        The fact that support for Froelich is weaker than the support for other candidates can be seen by the participants' unwillingness to cast multiple ballots for him in cumulative voting, and that a smaller proportion of Froelich supporters plump their ballots (i.e. cast more than one ballot for a candidate) than the supporters of any of the other candidates.
        The claim that fewer non-Froelich voters consider him a clear alternative to their first choice can be seen in the comparison of the single member district results to the at-large condition. When voters are given two more ballots to 'spread around', Froelich is the only candidate to do substantially worse in comparison to the other candidates. That is, relatively fewer non-principal Froelich supporters considered him an adequate second or third choice.
        Taken together, these two comparisons provide clear evidence of the potential dilemma for democratic theory in translating voter preferences into representation under plurality rules when there are more than two candidates competing for a single office.

Comparisons for a Women Candidate

        The other case highlights the influence of electoral systems on the selection of Williams, the female Independent candidate. Scholars have often argued that women candidates do better in multi-member district elections than in single member districts, although the evidence is somewhat mixed.
        The experimental outcome lends some support for the argument that women candidates may benefit when voters have an opportunity to help elect several candidates. While Froelich was shown to do worse relative to the other candidates when voters were given more ballots, Williams was the main beneficiary of that change. In spite of not having a base of partisan support, participants were willing to give her one of their votes when they had them to spread around.
        However, much of Williams' gain in support was lost when participants were allowed to plump their ballots for their preferred candidates. Thus, while women may gain some benefit in at-large systems, it may not necessarily hold across all types of multi-member district elections. This raises important questions for those who claim that cumulative voting is a good alternative system to bolster the electoral empowerment of both minority and female candidates.

Individual Reactions to the Electoral Systems

        Since the critical aspect of electoral systems deal with the translation of preferences into representation, they play a central role in democracy. These systems are not only important for deciding the composition of elected bodies, but for maintaining legitimacy of the outcomes. Therefore, how alternative systems are perceived will affect how seriously they are considered, as was seen in the outcry over Lani Guinier's "undemocratic" beliefs about the inadequacy of "winner-take-all" electoral systems.

Perceptions of Electoral Choice and System

Satisfaction with Expression Single Member District At-Large Election Cumulative Voting
Candidate Preferences 0.648 0.697 0.705
Candidate Strength 0.615 0.617 0.752
Perceptions of System
Difficult to Understand 0.312 0.245 0.230
Fairness of Selection 0.668 0.703 0.727

        
Note: All scales range from 0 to 1 with higher values reflecting greater satisfaction in the top two rows and perceptions of greater difficulty and fairness in the bottom two. For example, participants found single-member districts harder to understand than cumulative voting.

        Some recent research has examined individuals' satisfaction and difficulty with new election systems like cumulative voting. The findings from these studies suggest that arguments about cumulative voting's complexity for voters are overstated, but analysis of voter satisfaction has often been tainted by the real world dynamics created by cumulative voting systems being adopted after a judicial mandate to change a local system. The experimental context of this study allows for a more objective examination of perceptions of the fairness of different electoral systems as well as their difficulty.
        After selecting candidates in the mock election, participants were asked about their satisfaction with their ability to express preferences and the difficulty they had understanding the electoral systems. Table 2 provides the average responses for individuals in each electoral system regarding these perceptions.
        The first two questions in the table give the level of satisfaction individuals felt regarding their abilities to express which candidates they preferred in the election and to express how strongly they preferred specific candidates. These scales range from 0 for extremely dissatisfied to 1 for extremely satisfied, with a 0.667 being moderately satisfied.
        The experimental results demonstrate that individuals can distinguish among electoral systems in how their preferences are tallied. Individuals in the two multi-member electoral systems, at-large and cumulative voting, were significantly more satisfied with their ability to express the candidates they preferred than individuals voting in the single member district election.

Not only are systems such as cumulative voting not considered too difficult to understand, they are also considered as fair as those more commonly used such as single member plurality elections.

        Also, the participants in the cumulative voting condition were significantly more satisfied than those voting in the other elections regarding their ability to express the strength of their preferences. Thus the concerns addressed by democratic theory dealing with the translation of preferences into representation are not lost on the electorate.
        Another concern with unfamiliar electoral systems is that they may be difficult for citizens to understand. Participants were asked if they agreed or for strong agreement.
        It is clear from Table 2 that participants did not find any of the three systems too difficult. In fact, those in the at-large and cumulative voting conditions disagreed significantly more than those voting in the single member district election. Not only are systems such as cumulative voting not considered too difficult to understand, they are also considered as fair as those more commonly used such as single member plurality elections. Participants in all three systems are seen to moderately agree that the system they used gave everyone a fair chance to select officials of their choice.
        This study demonstrates that electoral systems have a clear influence on the outcomes and perceptions of elections. The manner that individuals can cast ballots and the way that they are tallied can clearly lead to different results. These variations occur, in part, because people use distinct decision rules in different institutional settings; in this case, voters expressed different preferences under different electoral rules.
        Given this finding, the goals of elections must be seriously considered when evaluating the merits of specific electoral systems. These goals will generally need to address how preferences are to be tallied for translation into representation. These should be the principal concerns because a number of studies, including this one, have shown that citizens can readily learn the mechanics of different systems. This experiment augments the examinations of actual electoral outcomes by providing truly comparable results across different systems. Combining these different approaches will help us gain a fuller understanding of the nature of elections and, more broadly, representation.
        Richard Timpone is an assistant professor of political science at the State University of New York at Stony Brook.

Table of Contents
Chapter Four