On Tuesday, November 5, 2024, Portland, Oregon will find out how the first tabulations of City Council, Mayoral, and Auditor votes work under our new ranked choice and multi-winner voting systems. For City Council elections, three winners will be selected from four brand new districts (the Mayor and City Auditor are elected under single winner RCV).

This new multi-seat method has some very interesting math and campaign implications. Votes can be transferred both from eliminated candidates and from winning candidates who exceed the number needed to win.

This summer, the Multnomah County Elections Division ran a practice election to create sample data and work through the counting procedures. This exercise was a great investment because it allowed the team at the Elections Division to show observers and researchers the detailed mechanics of running these unique elections. It also produced sample data for analysis and exploration. This past week, the team here at the Elections & Voting Information Center (EVIC) has been fine-tuning our use of these data to prepare for Election Day!

The sample election involved 25 candidates and required 25 rounds to identify a winner. In the council races, each round is a step in the tabulation where totals are taken from the first pick for each ballot. In round one, if there is no winner, the candidate with fewer votes is eliminated and those ballots have their second choice moved to the next round. If a candidate wins in a round — which means the candidate has enough votes to exceed the 25% plus one threshold to be a winner — those ballots are allocated as a fraction to second (or third) choices. Rounds are repeated until three candidates win.

This process produces a LARGE table of results that can be hard to read initially. The image below shows the official report – with each round’s result and transfer of votes. Can we visualize this in a simpler form?

To create a visualization we use a variation of the Bump Plot. A Bump Plot shows the changing rankings of things over time. These are popular for showing sports standings over time. But in this case, we are not just showing the ranking – we want to show the magnitude of vote totals. The Area Bump Plot (or Sankey Bump Plot) is a solution here.

We chose to first visualize the eventual winners in three shades of green to help the reader follow the progression of these candidates. Here you see two of the eventual winners started out in first and second place in the early rounds. The third winner started out in sixth place and even dropped a few positions over rounds until Round 16 where other candidates being eliminated added to their total.

On election night, many of us will look at these patterns to see how voters connected candidates in their rankings. In this sample data, two clear front runners never left the top two positions. Here we’d expect candidates with widespread recognition we also see how much of the vote transfers are not going to these candidates, suggesting the third candidate race is distinct in issues or personality.

In this hypothetical example, Alisha Ali starts lower in the rankings and even slips over rounds until Round 16 where the shift begins. In this round, voters that had their first pick (or potentially second, third, or even further choices) eliminated have Ali as their next choice at a sizeable margin – about 30%. In each subsequent round this pattern holds, meaning Ali was a consistent back up choice for voters who supported candidates Ouedraogo, Hornby, Dillon, Itoya, and Kato. In a real race, we would ask: what connects these candidates? Did they share platforms, networks, or other traits? We will wait to ask these questions tomorrow!