Hi Gresham Charter Commissioners, 
I want to thank Chair Andaya for writing and sharing the link to the recent May 18th, 2022 FairVote presentation and for asking me to respond. Diane Silver from FairVote is someone we've come across in the past and I've written rebuttals to her presentations before. Each presentation and each RCV presenter is a bit different, and Diane is a passionate advocate for RCV and we have a ton of respect for that, but her presentations are typical of FairVote in that they tend to paint with a very broad brush and get important details wrong. In the field, In the field, FairVote is notorious for a decades long history of glossing over key facts, or worse, deliberately misrepresenting information to their benefit. I recommend this resource which was submitted as official testimony to the Oregon Legislature because it has each common false claim stated succinctly and then it has the citations listed. False claims about voting methods are pervasive in the electoral reform space, but most can be traced back to FairVote and their staffers originally. Fair Vote is the most funded and largest voting reform advocacy group, fundraising and spending many millions of dollars annually between their affiliate groups so most take their word at face value.

For the part of the presentation up to the 1:10 minute mark in the video we agree with everything Ms Silver has to say and we agree that vote-splitting is a fundamental root problem with our current system. After that there are a few points that deserve a closer look. 

At 1:11:43 Ms Silver says, "it is still one person one vote". We actually disagree on this point. Yes, RCV never gives any voter more than one vote, but it does count some votes less than it counts others. Even for voters whose 1st choice is eliminated, and who had a second choice marked, some of these voters will not have their next choice counted. Others will. This is fundamentally unequal. This is the reason RCV (and its related multi-winner methods don't pass the Equal Vote Criterion. In the 1964 Wesberry v. Sanders decision, the U.S. Supreme Court declared that equality of voting—one person, one vote—means that "the weight and worth of the citizens' votes as nearly as is practicable must be the same." RCV does not pass this test. 

In the exact same way that one voter in a gerrymandered district might have less vote power than others, RCV gives more power to some voters while leaving others at a mathematical disadvantage with a less powerful vote. 

RCV has recently been banned in a number of states including Florida and Tennessee for related reasons. It was banned in Texas decades prior. We strongly condemn the banning of alternative voting methods and believe voters should have the right to adopt and try the systems they want. Still, this is a real concern for voters and for the future of the movement at large. 

1:13:00 Ms Silver talks about transparency of results and election integrity. The example she gives is a simple one that doesn't show exhausted ballots. What isn't mentioned here is that RCV requires centralized tabulation and requires that all ballots be in hand and that each round be fully tabulated before tabulation on the next round can start. The impact of this is that real time preliminary results aren't available until well after election day, and that preliminary and final results only include top choice rankings for each round. They can't include the full tally of rankings given since most rankings voters put down will never be counted. Most importantly, the requirement for centralized tabulation means that any issues or errors counting any subset of ballots will impact the tabulation of all other ballots. This is problematic for counting at any scale, but increasingly so the larger the scale of the election. The requirement for centralized tabulation for RCV also makes it incompatible with our current risk limiting audit protocols and our current election security protocols.

The reality is that election tabulation mistakes do happen, and the current ballot printing error in Clackamas County is a timely reminder. The thing to note here is that RCV has the potential to dramatically magnify the impact of any errors made at worst, or at best to make tabulation much slower than expected if an issue arises. In RCV elections it's normal to not have results for days, if not weeks, even when things run smoothly. 

NYC recently used RCV for their 2020 democratic primary and had a massive issue come up. Part way through the counting process of the mayoral election a candidate discovered that their internal campaign numbers didn't match the official results. An investigation was launched and it was discovered that 135,000 extra "test" ballots had been accidentally added to the official count. While this issue wasn't because of RCV specifically, it is highly likely that the much more complicated system made the error more likely to happen in the first place and much harder to catch than it would have been. And, because the error happened in an RCV election it meant that the full election count had to be redone from scratch. There was no way to just remove the test ballots and keep going from there. This article on the matter from NY polysci professor Lindsey Cormack is worth a read, as are her others. She is also a good presenter and would likely be happy to come and answer questions or present. 


1:13:17 Ms Silver makes the point here that RCV leads to better representation for women and people of color. We agree with her that preferencial voting and more expressive voting methods in general will improve equitable representation, and since RCV does mitigate vote-splitting in less competitive elections it does help remove some of the barriers to entry as well. That said, there is also a case to be made that the FairVote paper she is likely referencing here confuses correlation and causation. RCV has been overwhelmingly been adopted in urban areas where women and people of color are winning in higher numbers regardless of voting method, and RCV has mostly been adopted in highly progressive cities that are setting that curve anyways. 

What she doesn't mention is that there is strong statistical evidence that RCV is significantly more likely to exhaust, spoil, or waste the ballots of voters of color, voters who don't speak English as a first language, and voters from less wealthy neighborhoods. 


1:13:40 Ms Silver and her slideshow state that with RCV there is "No more vote-splitting." This is just objectively false and FairVote absolutely knows it. In 2017 their CEO agreed to stop making this and a shortlist of other false claims and signed a joint agreement about this with the leaders of a number of other voting reform orgs, including our own, the Center For Election Science, and leaders from Represenent.Us. In the agreement he pledged that FairVote would use the word "mitigates" when needed and avoid these kinds of oversold false claims. The fact that they continue to make this claim knowing that it is false is deeply problematic considering their influence, the amount of money they spend lobbying for voting reform, and the fact that they are misrepresenting their reforms to their donors and to elected officials like yourselves.

RCV can have serious vote-splitting in any round of tabulation. This is because only one ranking is counted in a given round. RCV literally works just like the current choose-one system, repeated over and over with one candidate eliminated each time. All the pathologies of the current system can still happen in any round of RCV, and changing the order of elimination of non-winning candidates can change the winner. 

While FairVote argues that  pathologies like vote-splitting (and non-monotonicity where your vote backfires and has the opposite of the intended effect) are rare, the findings from statistical analysis results are deeply concerning, and in real life, the full election data from overseas adoptions is not available for analysis. Domestically, until recently, RCV had mostly been adopted in urban jurisdictions that are one party dominated, though that is starting to change. When we look at real elections with RCV, we often are unable to get a sample size large enough to come to firm conclusions, but statistical analysis is very clear and the findings are unanimous. If we look at the close races and competitive elections where alternative voting would actually make a difference, these issues become much more common. Most analysis shows that RCV gets roughly the same outcomes as a primary and a top two general election. The science has been clear for years that in competitive races with three or more candidates RCV is much more likely to have vote-splitting bad enough to cause the system to elect the wrong candidate than almost any other alternative voting method.

Figure_4.a._201_voters,_100000_iterations close.png

(This is a graph from the Merrill Social Utility Efficiency simulations, the original studies on voting method accuracy, but with the models rerun to include STAR. In this image RCV is listed as "Hare," the system's original name. "Runoff" is a primary with a top two runoff. Note that the RCV purple line and the Top Two Runoff brown line both drop off sharply when more than two candidates are in the race, and that all other methods tested do significantly better. For a comprehensive overview of this and other models, including ones that factor in strategic voting this article is a great resource.)

The same bullet point states that with RCV there is "No more strategic voting." This statement is a red flag about the credibility of any source because no voting method can entirely remove strategic incentives across the board. This is widely known and commonly taught in any 101 course on the subject. That said, RCV is not even among the top methods by this metric. Voters in RCV should absolutely not rank a candidate first who they think may be strong enough to hang on through the rounds of tabulation but be eliminated in the end. In close races voters who do so in RCV risk having the vote never transferring, which can have the effect of preventing their preferred viable candidate from winning and help to elect their worst case scenario. In RCV there is a counterintuitive phenomenon called non-monotonicity where voting for a candidate can actually backfire and get you a worse outcome than if you had not voted at all. Studies suggest that this may happen to over 15% of voters in competitive RCV elections. No other voting methods under consideration exhibit this pathology and many consider it to be a deal breaker. 

We've shared this information with Ms Silver when we were both presenting in the same zoom calls before, so this is not news to her. That said, Ms Silver is relatively new to FairVote. We made a point of reaching out to follow up and discuss these points further, but she has not taken us up on the offer to date. 


At 1:15:15 Ms Silver talks about STAR Voting and RCV and says "For me, strategic voting is a very big concern... Voters should not have to play mind games to cast their vote." Then she goes on to say that this is why she likes RCV over STAR. The opposite should be the case since STAR beats RCV on this metric according to all available evidence. The fact that she is apparently unaware that in RCV, just like in the current system, it is often not safe to vote for your favorite in 1st place is very concerning.

RCV seems like it would do better strategic incentives than it actually does. The disconnect between appearances and reality is a big problem. Regardless of the voting system, it's absolutely critical that voters are able to navigate the system correctly so as to cast their vote effectively. In the current system, despite all its flaws, at least it's transparent enough that voters mostly navigate the strategy correctly. If they didn't we'd be in even worse shape. 

In RCV most voters think it's safe to vote honestly, and many do, but then are punished for doing so. In RCV ranking your favorite can and does backfire if doing so prevents your vote from being able to transfer until your other choices are eliminated. This can happen to up to a third of the electorate and when it does voting honestly can actually help elect your last choice. It's happened in real world elections in Burlington, VT in 2009 and again recently in Alaska in 2022. While the outcome itself of those elections were celebrated as a win by some, in both cases the majority preferred another candidate over the winner and voters learned the hard way that they should have been strategic and voted for the lesser-evil. 

In STAR Voting it's safe and smart to vote honestly and voters who do so will have a strong effective vote. Having the incentives for voter behavior aligned with the behavior we need to get the best outcomes is absolutely critical. 

To be clear, for strategic voting and lesser-evil voting the current system is absolutely the worst by a wide margin. RCV is better at mitigating those super harmful incentives where there are only two viable candidates, but it still is not safe to rank your favorite in 1st place in more competitive races. In STAR Voting, voters absolutely should always give their favorites 5 stars and show their honest preference order. In both RCV and STAR voters have good incentives to show their preference order between other candidates beyond their favorites. 

Strategic voting is complex and multifaceted, but these two images should help to sum up the comparisons and findings: 


Strategic Voting - STAR and RCV.jpg

Strategic Voting.jpg


At 1:16:00 Ms Silver really talks through the reasons why she wants a voting method where the best strategy is to vote your conscience, but the voting method she describes is not the one she thinks it is. Condorcet voting methods and STAR Voting in particular do much better at this than RCV.  


We're only a few minutes into the presentation and this is already quite long, so I'm going to address a few more common concerns FairVote tends to bring up directly rather than keep going through point by point. 


FAQ: Addressing more concerns from FairVote: 

Would voters bullet vote in STAR Voting? 

One of the concerns that FairVote often raises about STAR Voting is that voters would "bullet vote." Bullet voting is a voter behavior that can be done in either RCV and STAR Voting where a voter ranks or rates only one candidate, basically voting just like they would have in the old system if they were voting honestly. In most cases this is not actually a "strategic" or dishonest vote, and if the voter actually had a strong polarized opinion this may have been their most honest vote possible. It also is common among voters who are rushed or not well informed.

In other cases, this is a reflection of voter confidence that their favorite will win. A study on bullet voting rates in RCV elections by FairVote showed that on average 32% of RCV voters bullet voted, and that voters whose favorites were likely to win by a huge margin did so the most. This makes sense, because in those scenarios additional rankings wouldn't have made a difference anyways. Our initial findings for bullet voting in real STAR elections are comparable, if not better, and the behaviors seen to follow the same general trends. 

The fact is that in order for RCV or STAR to do their very best, we don't actually need perfect voting behavior. In STAR Voting it's best for voters to show their full preference order to ensure that no matter who makes the runoff, that their vote can go to the finalist they prefer, but the fact is that the system will still do great even if voters are a lot less diligent. In real elections voters usually do have a good idea of which candidates are competitive, and all a voter has to do to get an optimal result is to show a preference between the front-runners. 


What if voters in STAR Voting exaggerate their scores? 

In STAR Voting, one concern people raise is that voters might strategically exaggerate their scores. For example, I might give my favorite 5 stars, my last choice 0, and then be strategic about how to score my second choice. Let's say that my favorite isn't really viable, and that it's likely to come down to my last choice vs a candidate who is the lesser of two evils. Should I give my lesser-evil 1 star (which they deserve) or 4 stars to help them beat my last choice? While this is a tactical calculation I'm making, as long as I've shown my preference order it's not actually a dishonest vote, unlike the kinds of strategic voting that voters have to do in other systems. Here's why: In STAR Voting your scores are relative between the candidates. This is where STAR Voting differs from a 5 star product rating. In STAR, per the instructions, you give your favorite 5 stars even if they aren't perfect, because you are showing your relative opinions on the candidates. So, if you really strongly want to defeat your last choice, it's totally fine to give your second choice 4 stars. Or if your priority is making sure your favorite beats your 2nd choice because your 2nd choice is almost as bad as your last choice it's fine to give them 1 star. Ultimately, it's up to you as the voter. 


What if voters don't use the full 0-5 scale in STAR Voting? 

Another concern people sometimes raise is "What if voters don't use the full 0-5 scale in STAR Voting?" The STAR Voting instructions are clear that voters should give their favorite 5 stars, give their last choice 0, and show their preference order, but it's reasonable to expect that some voters might miss the memo, especially if they are using the system for the first time. This is a primary reason STAR Voting has a top-two runoff built in. Of course ideally voters should follow the instructions, but even if they don't all they have to do to ensure that their vote is just as powerful as anyone else's is show a preference between the finalists. In the STAR Voting runoff your ballot counts as one full vote, so regardless what the scores you gave were your vote is counted as one full vote for the finalist you prefer. A STAR Voting runoff works just like a top two general election, except that voters don't have to vote again. 


Score voting is vulnerable to strategic voting. Isn't STAR voting a type of Score voting? 
Score voting, where the highest scoring candidate wins, is one of the oldest methods out there, and it tests incredibly well with optimal voter behavior. Score Voting actually beats Arrow's famous Impossibility Theorem which won the Nobel prize for proving that no *ranked* method could pass all of the desirable criteria. Score voting actually does pass all of the criteria Kenneth Arrow listed, but Arrow's point held. No voting method can be perfect on every metric. Where Score voting isn't the best is if voters don't vote optimally. For example in Score if voters don't use the full scale their votes are less powerful than voters who did. Score voting also incentivizes voters to exaggerate the difference between the frontrunners to ensure they have as much impact as possible. It's worth noting that these problems, while not ideal, have long been overblown. Score voting is one of the most accurate methods out there, and even when the situation isn't ideal it still outperforms most methods at their best. Still, these concerns are valid, and it turns out they're easy to address. 
STAR Voting corrects for both of these issues by adding a binary runoff. In STAR, voters don't have to exaggerate because in the runoff you vote is binary anyways. Even if you don't give anyone a top score, and even if you don't have as strong a preference between the frontrunners, the STAR system ensures that your vote is fully powerful automatically.
Ranked Choice Voting passes the Later-No-Harm Criteria and STAR fails it. Does that mean RCV is better at empowering voters to vote their conscience? 
Thank you for your time and consideration, if you have any other questions, concerns, or if you'd like additional citations please reach out,