A new version of this paper from Barbara Klein and the League of Women Voters was released in Feb 2021. A commentary on the new piece will be available here as soon as it's available. Unfortunately the updated version didn't incorporate most of the new information, citations, or corrections we had submitted and most of our original comments still apply.

Below you will find the original comments on the 2019 piece. 

Dear Norman,

As promised, here is a list of notes for the problematic LOWV paper circulating. Feel free to connect me with Barbara Klein directly. These points were all addressed in my meetings with Keli Osborne and Lane LOWV in 2018 and 19 and with PDX LOWV in 2020, but I have not had the opportunity to meet with Barbara yet. It is our hope that there is a more current paper that could be shared with these points corrected, but if not I'd be happy to consult on that. All the points below can be independently verified, and I've tried to include key citations for each point despite the short notice. 


Points flagged for update in the "STAR and RCV condensed paper, Barbara Klein 2019" paper attached:

• Intro: Debate among reformers props up the status quo. 
  Electoral science is deep field with a long history. Reforms proposed need to be supported by the science, which unfortunately has not always been the case, especially in recent decades, as the field has evolved. Not all reforms are created equally, and with modern statistics and data analysis which can very accurately measure and predict the frequency of problematic outcomes. In particular, RCV has a long history of being sold with pervasive but factually false claims: 1. RCV eliminates vote-splitting and/or spoilers, 2. No wasted votes. 3. If your favorite can't win your next choice will be counted. 4. Elects majority winners. 5. RCV is less vulnerable to strategic voting than STAR or Approval. 6. RCV is more simple.
  
  Having outreach efforts and voter education derailed by false claims and misinformation is damaging to the movement and oversold claims were directly linked to a number of RCV repeals after voters realized they had been misled. In every case these repeals soured voters against reform, and have left us back at square one with the empirically worst system- plurality voting. Many are slow to retract claims or correct the record because claims have been perpetuated by seemingly reputable sources, especially FairVote, and reformers often don't know who to believe until they've invested significant time researching the science themselves. Many organizations in the PNW and around the country have taken this feedback to heart and are doing better, including many local RCV activists, but FairVote's national organization has refused to take responsibility for correct common false talking points within their local chapters or by their partners, even when they admit that these claims are false or misleading. 
  
  There are legitimate pros and cons for comparing voting methods and raising the bar on voter education and factual integrity would allow us to have more positive debates and better solidarity between reformers from different methods. This is where we need the leadership of groups like the League to correct the record and raise the bar. 

• Row 1: Expressiveness 
  Looks good. Note that cognitive load theory supports 5 star rating over ranking for large fields.
  https://www.surveymonkey.com/curiosity/rating-vs-ranking/

• Row 2: Recommends STAR for non-competitive elections only
  STAR has since been used in the statewide Ind Party primary 2020, the 2020 presidential delegate elections for Oregon from the DPO, and the MultCo Dems elected 95 winners using STAR last Saturday for officer positions, CD 1, 3, and 5 delegates, and the SCC. (The winner set was more diverse than the previous delegation.) It has also been used widely by many private orgs. 
  starvoting.us/case_studies is a good source but note that it's not fully updated due to the number of orgs now using STAR, particularly in the last year and even week. :) 

• Row 3: on Experience
  See above

• Row 4: on Strategy
  Claims against STAR and for RCV are contradicted by the studies on this. Failing to eliminate vote-splitting means that voters should still vote lesser-evil if their favorite is strong but not likely to win. RCV is a series of Plurality elections, so fundamental biases and flaws in Plurality are maintained, though sometimes mitigated. In Burlington 2009 almost a 1/3 of voters should have voted lesser-evil, failing to do so caused their votes to backfire, electing their last choice. The system was then repealed. 

  https://electionscience.org/voting-methods/runoff-election-the-limits-of-ranked-choice-voting/
  http://electionscience.github.io/vse-sim/VSE/ 
  https://en.wikipedia.org/wiki/2009_Burlington_mayoral_election (see citations) 
  Even FairVote has conceded that their white paper on this subject was misleading after it was discredited by leaders in the field as self-contradictory and misleading. They agreed to take it down but then did not follow through. (See attached.)
  
  

• Row 5: Later No Harm
  No voting method which passes this criteria can eliminate vote splitting and spoilers, the number one cause of strategic voting and implicit bias in the vote.  This is one of two mutually exclusive criteria, the other being Favorite Betrayal which ensures that it's safe to vote for your favorite. STAR does a good job at both of these criteria, but in general it's key to look at frequency of issues not just criteria pass/fail. No voting method can pass all criteria and there are many that are important and mutually exclusive.
  https://www.starvoting.us/farewell_to_pass_fail
  http://starvoting.us/criteria (see citations.)
  
  

• Row 6: Condorcet
  STAR elects condorcet winners on par with actual Condorcet methods. (RCV came in 48th on this metric in one study. These findings have been broadly corroborated.) 
  https://www.equal.vote/science

  https://www.starvoting.us/accuracy (30+ see citations)
   

• Row 7: Inconsistent Personal Translation of Preference into Score
  Every ballot is one equally weighted vote in STAR. IRV does not have an equally weighted vote because some (relevant) rankings may be are ignored, and because some ballots are exhausted. 
  https://www.sciencedirect.com/science/article/abs/pii/S0261379414001395
  https://www.starvoting.us/equal_vote
  
  

• Row 8: Nullify votes 
  This section misunderstands that in STAR a vote of no preference between two candidates still is counted and does have an impact comparing these candidates with others. Compare this with an Exhausted Ballot in RCV which is literally full ballots not included in the deciding round, even when those voters preferences were relevant. STAR has no exhausted ballots. All ballot data is counted. Every ballots is counted in both rounds.
  PEER REVIEW: "The rate of ballot exhaustion [in IRV] was high in each election, ranging [from] 9.6%-27.1%."  
  https://www.sciencedirect.com/science/article/abs/pii/S0261379414001395
  
  

• Row 9: Hand count audit
  Doable for STAR, also STAR is compatible with Risk-Limiting-Audits (RLAs) (RCV is not, particularly at the local level and for larger elections. Ballot data must all be centralized for tabulation and auditing.)

• Row 10: Reduces Negative Campaigning? True for all alternative voting where voters can support more than one candidate, STAR, RCV, Approval, etc. 
  

• Row 11: Monotonicity 
  STAR is monotonic. Previous claims otherwise by FairVote were retracted years ago. (See criteria checker chart.) RCV is likely to have this type of vote-splitting issue throw off results in 15% of elections or more in competitive elections.
  
  This peer reviewed paper "Frequency of monotonicity failure under Instant Runoff Voting: Estimates based on a spatial model of elections"begins by stating:
  "It has long been recognized that Instant Runoff Voting (IRV) suffers from a defect known as nonmonotonicity, wherein increasing support for a candidate among a subset of voters may adversely affect that candidate’s election outcome. The expected frequency of this type of behavior, however, remains an open and important question, and limited access to detailed election data makes it difficult to resolve empirically... We conclude that monotonicity failures in three-candidate IRV elections may be much more prevalent than widely presumed (results suggest a lower bound estimate of 15% for competitive elections)."Dr. Joseph T Ornstein of the University of Michigan and Dr. Robert Z. Norman of Dartmouth College, 2013.