The prototype ballots included one with ranking, one for cumulative voting, and two versions for approval voting. Participants sometimes struggled to understand both how to express their choices through marking and how the ballot would be counted, even with brief instructions and a voter guide. This suggests the need for a strong voter education program when introducing a new voting method that covers both the mechanics of voting and the mental model for the method.

Relying on existing mental models caused confusion

We saw three issues where voter mental models conflicted with the voting method as represented on the ballot. Although participants followed instructions and marked the ballots, these issues are important points of confusion.

• Participants expected to mark only the same number of candidates as the number elected. 
• Some felt that voting for 1 is a more efficient and legitimate system. They felt that fewer options causes voters to more carefully consider their votes and makes voting easier.
• Some were concerned that voting for a party would be confusing or conflict with voting directly for candidates.

Participants expected the number of marks to match number to be elected:

Based on their prior experience with elections, when the numbers did not match, participants needed clarification, were confused, and expressed negative sentiments. They finally voted without understanding the different numbers.

Cumulative voting is an exception, because the number of selections always matched the number elected.

A few participants used the words “vote” and “elect” interchangeably, especially when discussing the ranking ballot.

We asked which ballots are more “American”

We asked participants which ballot felt most American. Familiar and simple ballots felt more American. Most said ballots feel American when they are similar to ballots currently used.

• The approval ballot was the top response. Even though the approval ballot was most disliked, 5 participants said the approval ballot was the most American.
• Ranking may remind people of surveys commonly used. One participant said ballots feel American if they are similar to surveys.
• Voting for a party on the German ballot raised concerns, although some liked the idea of voting for a party to support efficient government and a party even if not all candidates

Each of the ballot designs had pros and cons for participants

• Ranking. Most participants preferred the ranking ballot. They described it as orderly, transparent, and clear. Ranking made them think about their votes and allowed them to express more nuanced intent. The only errors were on the ranking ballot with  3 overvotes and 1 overrank. 
• Cumulative voting. The cumulative ballot resulted in careful choices. No one gave more than one mark to any candidate. Participants said they would give multiple marks to one candidate if they had strong feelings about that candidate and more information about platforms.
• Approval. The approval ballot was the least favorite.  They felt the approval ballot did not convey nuanced voter intent. Some preferred it for the simplicity. 

Understanding how the ballots are counted was a struggle

After reading a brief description of the counting methods, only half of the participants correctly matched the counting method to the ballot design that used it. 

Participants used ranking to try to understand counting. 

• Two correctly associated ranking with counting in rounds.
• Two incorrect matches associated the top rank with the highest number of votes.

Percentages in counting methods caused issues. Reading instructions for proportional RCV with a 25% winning threshold, participants associated  it with both a cumulative and approval ballot. only include the most important points here, and close by directing readers to the full report above.

This research was conducted by Misty Crooks and Emma Werowinski.

We tested 5 ballots using 4 voting methods: 

• Ranking
• Cumulative
• Approval voting (2 versions)
• German multi-member proportional system. 

The prototype ballots included different levels of instructions and information about the counting methods. 

The research included 11 participants in Bridgeport, CT and Allentown, PA. Connecticut voters have experience with fusion voting. Pennsylvania voters have no experience with any of the voting methods used in this testing.