Analysis of Instant Runoff Voting. - Alex Hobel; Research intern for Berkeley City Council Member Gordon Wozniak.

Transcription

1 Analysis of Instant Runoff Voting - Alex Hobel; Research intern for Berkeley City Council Member Gordon Wozniak.

2 Typical Solution to Determine a Winner. General election followed by a runoff election between the top two candidates. This ensures a majority. Problems: Lower voter turnout for runoff Expenses involved with having a second election.

3 Instant Runoff Voting (IRV) Instant Runoff Voting is a system of voting which combines the runoff with the general election to determine a majority winner. It allows every voter to rank their choices for all or some number of candidates for the position. After the first count, the candidate with the smallest total is eliminated and these votes are redistributed. This process is repeated until a majority winner is determined.

4 Vocabulary of IRV : each round in the instant runoff is called a pass. Generally one candidate is eliminated per pass. Exhausted Ballots: Ballots whose choices do not include any of the candidates left. These ballots are not included in the vote total for that pass.

5 Different Methods of IRV Full-Preference Voting: Used by the Australians, voters rank all candidates on the ballot. Supplementary Voting: Used in London, voters can vote for a 1 st and 2 nd choice; involves only two rounds. Partial Preference Voting: Used in San Francisco under the name of Ranked Choice Voting (RCV), allows voters to rank up to three candidates.

6 Used In SF Board of Supervisors Elections for Districts 1,5,7, and 11 in Nov, 2004

7 Results (First Round) District 1 District 5 District 7 District 11 Jake McGoldrick 11,815 Ross Mirkarimi 9,947 Sean R. Elsbernd 10,505 Gerardo Sandoval 7,477 Lillian Sing 8,989 Robert Haaland 5,124 Christine Linnenbach 6,784 Myrna Viray Lim 4,280 Matt Tuchow 2,864 Lisa Feldstein 3,257 Isaac Wang 2,728 Jose Medina 2,869 David Heller 2,012 Nick Waugh 3,025 Gregory Corrales 2,560 Anita Grier 2,806 Rose Tsai 1,595 Andrew Sullivan 2,477 Milton "Rennie" O'Brien 2,372 Rolando A. Bonila 2,293 Leanna Dawydiak 1,380 Bill Barnes 1,664 Vernon C. Grigg III 2,091 Rebecca ReynoldsSi 1,816 Jeffrey S. Freebairn 132 Jim Siegel 1,540 Shawn Reifsteck 1,108 Tom Yuen 1,328 Dan Kalb 1,398 Michael Patrick Mallen 975 Fil M. Silvero 307 Total 28,787 Susan C. King 977 Pat Lakey 763 Michael E. O'Connor 868 Svetlana Kaff 546 Total 23,176 Brett Wheeler 832 Art Belenson 510 Joseph Blue 802 Sheela Kini 349 Tys Sniffen 686 David Parker 348 Phoenix Streets 657 Julian Davis 418 Total 31,639 Emmett Gilman 393 Francis Somsel 368 Rob Anderson 336 Vivian Wilder 130 Patrick M. Ciocca 91 Phillip House 62 H. Brown 57 Total 35,109

8 Did IRV change any of the results for the SF Supervisor races in 2004? Is it likely that the voters 2 nd and 3 rd choices will change the standings from the 1 st round?

9 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Did IRV Change the Overall Results? Votes For Top Two candidates in First and Last Round (All Districts) # vote 1st round # vote last round Jake McGoldrick Lillian Sing Ross Mirkarimi Robert Haaland Sean Elsbernd Christine Linnenbach Gerardo Sandoval Myrna Viray Lim

10 Did IRV Change any Candidates Rank after a There were a total of 40 passes made in all of the BOS elections, with 314 opportunities for one candidate to pass another. Only 4 times did redistribution of votes push a candidate ahead of another. None of these candidates made it into the final pass of their ballots.

11 Vote Redistribution D7 Progression of Votes 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2, Sean R. Elsbernd Christine Linnenbach Isaac Wang Gregory Corrales Milton "Rennie" O'Brien Vernon C. Grigg III Shawn Reifsteck Michael Patrick Mallen Pat Lakey Svetlana Kaff Art Belenson Sheela Kini

12 How to define Majority? To get a majority, does a candidate need to get 50%+1 of the total ballots cast, or just 50%+1 of the ballots remaining by the final pass? San Francisco chose to determine the majority winner only counting ballots remaining on the final pass. This may leave the winner of the election with only a plurality of the total votes.

13 Was a true majority achieved? Trying to Reach a Majority 100.0% 75.0% Percent of Total Votes Gained from Redistribution Percent of Vote Won in First Round 50.0% 25.0% 0.0% District 2 District 3 District 9 District 1 District 5 District 7 District 11

14 1 st Round Distribution of Votes District 1 1st Round Distribution of Votes Total Votes: 28,787 Total Votes: 35, 109 District 5 1st Round Distribution of Votes Others (5) 28% Jake McGoldrick 41% Others (19) 48% Ross Mirkarimi 28% Lillian Sing 31% Lisa Feldstein 9% Robert Haaland 15% Total Votes: 31,639 District 7 1st Round Distribution of Votes District 11 1st Round Distribution of Votes Total Votes: 23,176 Others (11) 46% Sean Elsbernd 33% Others (6) 50% Gerardo Sandoval 32% Christine Linnenbach 21% Myrna Viray Lim 18%

15 Following District 7 s Distribution of Votes District 7 1st Round Distribution of Votes District 7 Last Round Distribution of Votes District 7 Last Round Distribution of Votes, Not Including Undistributed Ballots Others (11) 46% Sean Elsbernd 33% Undistributed Ballots 23% Sean Elsbernd 44% Christine Linnenbach 43% Sean Elsbernd 57% Christine Linnenbach 21% Christine Linnenbach 33% Total Votes: 31,639 Total Votes: 31,639 Total Votes: 24,325

16 What happened to reduce the number of Total Votes? As more and more candidates become eliminated, voters second and third choices become reassigned to candidates already eliminated from the race. This leads to ballots becoming exhausted, significantly reducing the number of Total Votes.

17 Exhausted Ballots These undistributed ballots are called exhausted ballots, and in some cases outnumber the ballots that are redistributed. District 7 Redistribution of Votes 8,000 7,000 6,000 Available Redistribution Actual Redistribution Exhausted Ballots 5,000 Votes 4,000 3,000 2,000 1,

18 Exhausted Ballots Votes Redistributed to Top 2 candidates (District 7) 100% 80% 60% 40% Exhausted Ballots Other Christine Linnenbach Sean Elsbernd 20% 0%

19 Exhausted Ballots by Round Undistributed Ballots (district 1) Undistributed Ballots (district 5) 3 10% 2 7% All other es 37% 19 (Last ) 32% 4 (Last ) 83% 17 12% 18 19% Undistributed Ballots (district 7) Undistributed Ballots (District 11) All other es 15% All other es 12% 9 8% 10 14% 11 (Last ) 63% 4 13% 5 26% 6 (Last ) 49%

20 Redistribution of Votes Redistribution of Votes (District 1) Redistribution of Votes (District 5) , , , ,783 8, , Votes to be distributed Votes Redistributed to 1st or 2nd Place Votes Not Counted Votes to be distributed Votes Redistributed to 1st or 2nd Place Votes Not Counted Redistribution of Votes (SF District 7) Redistribution of Votes (District 11) , , ,550 Votes ,036 7, , Votes to be distributed Votes Redistributed to 1st or 2nd Place Votes Not Counted 0 Votes to be distributed Votes Redistributed to 1st or 2nd Place Votes Not Counted

21 Redistribution of Votes Redistribution of Votes (District 5) Redistribution of Votes by Percent , % % 100% % ,783 8,998 60% 46% 54% % % % Votes to be distributed Votes Redistributed to 1st or 2nd Place Votes Not Counted Votes to be distributed Percent Redistributed to 1st or 2nd Place Percent of Votes Not Counted In District 5, 54% of votes available for redistribution were Exhausted, and therefore not counted in the final pass. Does this limit the effectiveness of IRV?

22 Exhausted Ballots and the Margin of Victory Margin of Victory vs. Total Exhausted Ballots 10,000 9,000 Margin of Victory Total Exhausted Ballots 8,000 7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 District 1 District 5 District 7 District 11

23 Exhausted Ballots The exhausted ballots created in the election have the same effect as low voter turnout in a runoff election. Fewer people are able to voice their opinion on the top two candidates. It is in the last few rounds of redistribution that the majority of votes are exhausted

24 As Rounds Progress, Fewer Votes are Counted. Vote Conservation: Are all votes counted? (District 7) 35,000 Number of Votes Counted 30,000 25,000 20,000 15,000 10,000 5,

25 Vote Conservation: Percent of Votes Being Counted. 120% Vote Conservation: Percent of Total Votes Being Counted (District 1) Percent of Total Votes 120% Vote Conservation: Percent of Total Votes Being Counted. (District 5) Percent of Total Votes 100% 100% 80% 80% 60% 60% 40% 40% 20% 20% 0% 0% Vote Conservation: Percent of Total Votes Being Counted. (District 7) Vote Conservation: Percent of Total Votes being Counted. (District 11) 120% Percent of Total Votes 120% Percent of Total Votes 100% 100% 80% 80% 60% 60% 40% 40% 20% 20% 0% %

26 Are all Voting Systems Fair? Are all voters treated the same Variation in Hardware: Punch Card Ballots Optical Scan ballots Electronic Touch Screen Technology Does IRV allow all voters to be treated equally?

27 Voter Equality in SF elections English Speakers had prior knowledge of IRV at 70%, Chinese speakers knew at 69%, Spanish only knew at 56%, while other languages knew at 63% An Assessment of Ranked-Choice Voting in the San Francisco 2004 Election by SF State s PRI, page 41 Ranking three candidates was least common among African-Americans, Latinos, voters with less education, and people whose first language is not English Page 13

28 Voter Equality in SF Elections The study found a strong correlation between voters who did not have prior knowledge of IRV and voters who ranked only one candidate. If a voter ranks only one candidate, unless he or she votes for one of the top two, then the voter will not have his or her ballot counted in the final round.

29 3 different types of voters. 1 st - Voter AA casts 1 st choice for one of the two finalists 2 nd - Voter BB casts 1 st choice for someone who placed 3 rd or worse. Casts 2 nd or 3 rd choice for one of the two finalists. 3 rd - Voter CC casts 1 st choice for someone who placed third or worse. Casts 2 nd and 3 rd choices for candidates who don t make it to final round.

30 Whose Votes Count? 1st Round 2nd RoundFinal Voter AA Voter BB Voter CC Vote Counted 0-Vote not Counted

31 What Would Happen in Full- Preference Voting? We distributed the exhausted ballots to the candidates in proportion with the already distributed ballots, so each winning candidate would get a majority. We were trying to determine if fullpreference would make a difference in the margin of victory.

32 Percent Between First and Second Place (Comparing real results with results when we redistributed exhausted ballots proportionally) Partial Vs. Full Preference Voting (District 1) Parital Vs. Full Preference Voting (District 5) 14.0% 35.0% 12.0% 30.0% 10.0% 25.0% 8.0% 20.0% 6.0% 15.0% 4.0% 2.0% 0.0% Delta with partial preference (Cumulative) Delta with full preference (Cumulative) Partial Vs. Full Preference Voting(District 7) 10.0% 5.0% 0.0% 2 Delta with partial preference(cumulative) Delta with full preference (Cumulative) Partial Vs. Full Preference Voting (District 11) % 25.0% 20.0% 20.0% 15.0% 15.0% 10.0% 10.0% 5.0% Delta with Partial Preference(Cumulative) Delta with full preference (Cumulative) 5.0% Delta with partial preference (cumulative) Delta with full preference (Cumulative) 0.0% %

33 How can we make all votes count? In a Full Preference system, there would be no exhausted ballots But can there be a system that limits exhausted ballots without forcing voters to rank all the candidates? We ve come up with an alternate form of IRV that prevents exhausted ballots.

34 Alternative IRV System Voters can rank as many choices as they want. Once a candidate is eliminated, all possible votes are to be distributed to candidates that remain. If there is no way to pass on the votes, either because votes are distributed to candidates already eliminated or no other choices were made, then those ballots remain with the candidate. (This total cannot be added to in subsequent rounds) This leads to no exhausted ballots but does not guarantee a majority.

35 Alternative IRV Alternative IRV for District 11 12,000 10,000 8,000 Gerardo Sandoval Myrna Viray Lim Jose Medina Anita Grier Rolando A. Bonila Rebecca ReynoldsSilverberg Tom Yuen Fil M. Silvero 6,000 4,000 2,

36 District 11: Final Vote Distribution. San Francisco's RCV final round 12,000 10,679 10,000 8,000 7,628 6,000 4,869 4,000 2, ,000 12,000 12,019 Simulated Full-Preferece IRV Gerardo Sandoval 10,000 8,000 8,694 Myrna Viray Lim 6,000 4,000 2,000 Exhausted Ballots 0 6 Final Round of District 11 using Alternative IRV 12,000 10,679 10,000 8,000 7,628 6,000 4,000 2,392 2,000 1,

37 Summary Instant Runoff Voting did not change the ranking of the top two candidates in any election. Exhausted Ballots created by people ranking already eliminated candidates prevent the winner of the election from gaining a true majority. Voters whose ballots become exhausted before the last round do not get a say in the outcome of the election. There are other types of IRV that could be more effective than Partial-Preference Voting.

38 Looking at the 2000 Florida Presidential Election If IRV were used in the 2000 Florida Presidential Election, the Results after the first round would be the same as the present results. Florida 2000 Presidential Election 1st Round Democrats 48.84% Other 0.68% Green Party 1.63% Republicans 48.85%

39 Looking at the 2000 Florida Presidential Election In the final round, only the two main parties would remain, giving way to one of the two possibilities shown below. Although effective in this situation, IRV could prevent third parties from getting the credit they would have gotten from a regular election. Rep. 49% Dem. 51% Rep. 51% Dem. 49%

40 Sources Neely, Francis, Lisel Blash and Corey Cook. An Assessment of Ranked-Choice Voting in the San Francisco 2004 Election. San Francisco State s Public Research Institute (March 2005).