Casting multiple votes is easy for voters, but ranking causes issues

Publikováno dne:

For democracy to function well, it is not enough for people to simply show up to vote – they must also cast a valid ballot. The complexity of some voting systems is often mentioned as a barrier to electoral reform. But is this concern justified? Our experiment explored which methods voters struggle with and which they can use effortlessly.

In a representative survey, the detailed results of which can be found here, we had over 2,000 people vote using six different voting methods that could potentially serve as an alternative to the two-round system. We approached 1,253 people on the street, who voted using paper ballots, while another 912 voters were given the option to vote online. Unlike standard public opinion polls, our survey allowed respondents to make mistakes – thus replicating the conditions of a real voting booth. This provided us with deeper insights into how different voting methods might work in practice.¹

‍

The graph above shows how many voters cast an entirely invalid ballot using each method. Approval voting performed best, mainly because it is almost impossible to make a mistake: the only possible error is not filling anything in. Casting a valid ballot, however, does not necessarily mean that voters voted for the candidate they intended. It is possible for a mistake to be made without making the ballot completely invalid. The D21 method with plus votes also performed very well, with even more people casting valid votes compared to the single-vote system.²

The biggest challenge for voters was instant runoff voting. About 9 percent of voters were unable to cast a valid ballot under this method that requires them to rank all candidates. In the last presidential election, this would have meant nearly half a million votes wouldn't be counted towards the result! In reality, only about 0.8 percent of ballots – approximately 44,000 – were invalid.

When discussing the need for electoral reform, instant runoff is often considered one of the top options. But we sometimes overlook how complicated this system can be for voters. It is easy to make an error when ranking nine candidates, such as skipping a number or placing two candidates in the same position. Moreover, the requirement to rank even the candidates a voter may have no opinion on can discourage some people from voting at all. It should be noted that if the voting system were truly changed, people would become more familiar with it than during a poll, which would reduce error rates. However, the number of invalid votes in Australia shows that the excessive complexity of the system is an issue that won’t just disappear with practical implementation (although in Australia, mandatory voting also contributes to this problem). Introducing a version where ranking all preferences is optional could help reduce the number of invalid votes, but it would also make the shortcomings of this method even worse, which means it would end up yielding results similar to the first-past-the-post system (read more in our comparison of voting methods).

‍

At the same time, our survey shows that multiple-vote methods do not have the same problem. The success of voters in casting a valid ballot did not depend on how many votes they were allowed to use. One vote, three votes, or an unlimited number of votes all resulted in low error rates. The D21 method with multiple plus votes was handled by voters without any major issues. Even in the version with a minus vote, the number of invalid ballots remained relatively low, although the proportion of invalid minus votes was higher. Approximately 7.5 percent of people marked a minus vote without being entitled to do so, having cast only one plus vote. Minus votes were also incorrectly cast by 11 percent of respondents who assigned them to more than one candidate. While their plus votes still went to the candidates they wanted, the final result did not fully align with their intention.

And what about range voting? In this method, unmarked candidates are automatically given the lowest rating. Moreover, if a mistake is made in rating one candidate, the ratings for the others are still counted. Therefore, we cannot discuss invalid votes when it comes to range voting. However, 2.3 percent of voters did make an error, with 1.6 percent assigning multiple ratings to a single candidate. Another 0.7 percent made this mistake for multiple candidates.

Voters handled multiple-vote methods the best, and range voting also did not pose much of a problem for them. However, instant runoff performed poorly, with nearly one-tenth of voters unable to cast a valid ballot. Understanding how and why voters make errors is an important potential issue to consider when discussing changing the electoral system.

¹ The survey was conducted by SC&C from January 3 to 13, 2023. The results provided in the linked survey have been re-weighted to align with the results of the presidential election, during which there was a significant shift in voter preferences. For the purposes of this article, the data have not been re-weighted, and therefore the election results differ. Josef Středula, who withdrew shortly before the first round, is included in the data. Results of single-vote (FPTP) elections in the dataset used for this article: PP: 26.7%; AB: 26.2%; DN: 22.4%; JB: 8.0%; PF: 5.8%; MH: 4.0%; JS: 2.5%; KD: 2.4%; TZ: 1.9%.

² The difference between methods with lower proportions of invalid votes was not statistically significant in this sample, meaning we cannot generalize the results to the entire Czech Republic.

‍