An analysis of FairVote's Look at STAR Voting

Any voting system proposed for adoption in public elections should be subjected to scrutiny, and because STAR Voting is a new system that has never before been used in public elections, it is doubly important that it be given deep peer review. This is why we are so thankful that FairVote, the nation’s leading voting system reform advocacy organization, took the time to dig into STAR in their article "Explaining FairVote’s position on STAR Voting.” FairVote took great care to probe the system for potential weaknesses, particularly focusing on the effects of strategic voting.

FairVote’s article gives us a boost in confidence in STAR’s viability, precisely because the article’s proposed manipulation strategies are ineffectual in STAR Voting. We’ll break down the article, section by section, to show why this is so.

Section 1: (Wait, what’s this article about, anyway?)

STAR Voting, the titular subject of FairVote’s piece, is actually nowhere to be found above the fold. The first section, instead, is about the recent adoption successes for Instant Runoff Voting (which FairVote also calls RCV). We at the Equal Vote Coalition also celebrate the successes of voting reform advocates across the country, and are the first to agree that nearly any voting system alternative would be an improvement to our seriously broken “Choose Only One” plurality method.

To be clear, our objections to Instant Runoff Voting have nothing to do with the passion of its advocates, nor the goals we all share in common. Many Equal Voters were, in fact, longtime advocates of IRV. After deep consideration, however, we simply think that STAR Voting is vastly superior: STAR is monotonic, IRV is not. STAR is simpler, more accurate, more expressive, and is precinct summable to boot.

FairVote concludes this section with the thesis for the article: “Our position is grounded in a belief that, unlike RCV, [STAR Voting] would be subject to tactical voting and, unlike RCV, has serious viability challenges due to allowing a candidate to lose despite being the first choice of more than half the voters.”

This conclusion is incorrect on both points. RCV (aka Instant Runoff Voting) is subject to considerable tactical voting, as evidenced by FairVote’s own article, and the proposed viability challenge posed by the Majority Criterion ignores the fact that STAR, like Instant Runoff Voting, passes the Relaxed Majority Criterion, which has implications in the discussion below.

Section 2: How STAR Voting Works

One great feature of STAR Voting is that folks can pretty quickly and easily tell you how it works. From a single sentence - “STAR Voting elects the majority favorite between the two highest-scoring candidates” - to multi-paragraph explanations, STAR is always and simply Score Then Automatic Runoff.

The final statement of Section 2 is truly telling and inadvertently prophetic. FairVote writes, “With STAR voting, the voter's rankings are used in two ways - both to determine which candidates advance to the runoff and which candidate to support in the runoff. As discussed below, this combination can be problematic.”

This feature of STAR Voting is problematic — for voters trying to act strategically. To see why, read on.

Section 3: Why FairVote is neutral on STAR Voting in governmental elections

We were doubly thrilled with this article because, despite their significant concerns with the system, FairVote was willing to put forward a “neutral” rating. This tells us two things: one, FairVote understands rating systems, and two, at least they like STAR Voting better than Approval Voting, an alternative voting system they dump on regularly.

FairVote’s fundamental criticism in this section is that STAR Voting is new and has never been tried in public elections, so it probably should never be tried in public elections.

We disagree. Our very democracy was founded on the notion of innovation in governance. In our current times, the need is apparent to many that we must fundamentally reform our democracy if it is to continue, and STAR Voting is an invention born of this necessity. To say that we shouldn’t bother giving promising new voting systems a try is like saying we shouldn’t bother to adopt smart phones and should stick with the telegraph – the cutting-edge of communication technology at the time IRV was created.

Section 4: STAR Voting fails the majority criterion. And the mutual majority criterion.

A portion of FairVote’s analysis relies on particular rigid voting system criteria for support. This is the first case in point. In STAR it is hypothetically possible for a majority of voters to strictly prefer one candidate over all the rest and still have that candidate lose. However, what FairVote fails to disclose in this analysis are the conditions under which this can actually happen. With STAR Voting, a majority bloc of voters would have to support at least two candidates beyond their actual favorite in order for their favorite to have even a chance of losing.

Unlike with Score or Approval Voting, because of the runoff step, a majority of STAR voters can give a top score of 5 to a favorite candidate and a 4 to a second favorite candidate, and still ensure that their true favorite wins, no matter what the other voters do. In practice this means that voters can support secondary candidates without overly worrying that support would knock out a majority-preferred candidate.

This feature of STAR Voting has serious implications for the entire rest of FairVote’s analysis, and was key in the discovery of the Relaxed Majority Criterion, a more reasonable criterion for voting system efficacy.

Section 5: Easy understanding of how to vote tactically

On to the fun stuff!

FairVote Assertion 1: “STAR voting advocates maintain that tactical voting is unlikely with their system, but we strongly disagree”

We don’t argue that tactical voting is unlikely with STAR Voting. Some voters try to be tactical in every system we’re aware of. In the case of our current system, voters often tactically vote for the “lesser evil” because to honestly vote for a third party wastes their vote and increases the likelihood that the greater evil will win. This same problem also exists in IRV.  in Australia, where IRV has been used most widely for the longest time, sophisticated voters also vote tactically.

What we actually say is that STAR voters can try to be tactical, but that effort won’t get them much, if any benefit, and will just as likely backfire and create a worse outcome. We do think this will be an incentive, for both straight forward and strategic voters, to cast their ballots honestly.

FairVote Assertion 2: “Unlike RCV, STAR voting doesn't uphold the later-no-harm criterion”

This is entirely true, and actually desirable. STAR balances two competing criteria - Later No Harm (LNH) and Favorite Betrayal (FB), and minimizes the impact of both. LNH is essentially about how willing a voter will be to support candidates beyond his or her favorite. In a system that passes LNH a voter can always give 2nd, 3rd place, and additional rankings without worrying that that support could hurt their favorite. What LNH does not address, at all, is whether it is safe for a voter to give maximum support to his or her actual favorite candidate in the first place.

And this is critical. The fundamental problem with our voting system today is that it can incentivize voters to vote against their actual favorites, and instead place a “lesser evil” in first position. LNH is all about secondary choices, but the real problem, and what creates our false two party duopoly, has everything to do with the expressed first choices of the voters.

In the case of Instant Runoff Voting, voters who put their actual favorites in first place can be punished by the voting system in very predictable ways

In STAR you can harm your favorite’s chance of winning by supporting other candidates, but only if your favorite had little chance of winning in the first place.

In STAR, the two highest-scoring candidates advance to the automatic runoff.

If you think your favorite is in first position (i.e. the top scorer overall), no problem. You can offer secondary support without worry. Or, you may be so confident in your favorite that you “bullet vote” — that is you give your favorite top marks (5 of 5 stars!) and you give everyone else a zero. Bullet voting won’t help your favorite actually win, since there’s a runoff step, but it’s a simple way to fill out the ballot. On the other hand, maybe you aren’t supremely confident. In this case STAR lets you hedge your bet. You can give your second favorite(s) a 1… just a hair more than zero, to ensure that if your first choice doesn’t make the runoff, at least you still have a voice in the final outcome.

But what if you think your favorite is vying for second position in the runoff?

In this case you have a very strong incentive to make sure that the strongest choice you like makes the runoff. The strongest (highest-scoring) will be the most likely to beat the candidate you fear in the runoff. If you think your favorite is vying with your backup choice for second position, you are strongly incentivized in STAR to accurately support both.

FairVote Assertion 3: “Under systems that violate later-no-harm, we see that significant numbers of voters feel pressured to bullet vote, so as to not dilute their vote for their favorite. High-levels of bullet voting mean that results do not differ much from a plurality election.”

The runoff in STAR Voting entirely mitigates this “bullet voting” criticism, which FairVote has been lobbing at Score and Approval voting for decades.

In the STAR runoff you vote goes to the finalist you preferred so there is a very strong incentive for voters to show their preferences, especially in the elections where it matters.

Showing a more nuanced opinion in the runoff ensures that even if your favorite can’t win, your vote will help prevent your worst case scenario.

Since the publication of FairVote's article, Rob Richie, its coauthor and FairVote's Executive Director, conceded that, "I don't see bullet voting as a strategic incentive with STAR voting.” Glad you've come around at least that far, Rob!

FairVote Assertion 4: “We believe voters would be incentivized to tactically "bury" the strongest opposition candidates to keep them out of the runoff.”

Burying is a workable tactic in some voting systems, but it’s an unwise move in STAR. Burying, also known as “weak insincerity” is where you promote a weak 3rd choice candidate over a strong 2nd choice, in the hopes that will help your first choice candidate win against your weak 3rd choice.

Here’s the problem with that entire approach in STAR: if you think your first choice can’t win head-to-head in the runoff against your second choice, that means you think your favorite is vying for the second seat in the runoff.

If you think your favorite is vying for second seat in the runoff, the worst thing you can do is add points to a candidate you like less than both your first and second favorites. The most “strategic” you’ll ever be is to bullet vote in this scenario, but in reality, if you aren’t even sure your favorite has a shot at the runoff, you’re very likely to give your strong second choice at least one point.

FairVote’s Rob Richie sent us the article "The Effect of Approval Balloting on Strategic Voting Under Alternative Decision Rules" to support the “burying” thesis, but in fact this seminal paper does the opposite.

The paper concludes that burying is only effective in a multistage score process if the voter holds three key beliefs, (1) and (2) with great confidence: (1) that his favorite can make the runoff, (2) that his favorite can beat a third or lower preference candidate, (3) and that he is not sure his favorite can beat his second favorite in the runoff.

As illustrated above, in STAR Voting, the voter who believes (3) cannot believe (1) with great confidence: if you aren't sure your first choice can beat your second choice, your first choice is in play for the second spot in the runoff. Adding any support to a less preferred candidate in this scenario simply increases the likelihood your own favorite will be squeezed out and that your runoff vote will go to someone you didn’t actually like.

FairVote tries to support their thesis with the example of Macron’s election in France, but we invite them to make their case using a scenario a little closer to home. Imagine an election between Bernie Sanders, Donald Trump, Hillary Clinton and John Kasich. FairVote essentially suggests that Clinton voters would promote Trump and Kasich in STAR over Sanders, just to make sure that a weaker opponent makes the runoff. We think it’s quite obvious that very few real human Clinton voters would willingly risk promoting Trump and Kasich over Sanders in STAR, for obvious reasons.

Fundamentally, FairVote’s burying “strategy” can only work in STAR for one faction out of the many challengers, yet it requires all the factions to use it in order for it to work in the first place. Thus, if all of the contender factions use this approach, all but one will see a worse outcome than the compromise choice. Because we know that voters today regularly “lesser evil” vote in order to prevent their worst outcomes from coming to pass, we find it highly unlikely that voters on the whole will adopt a “strategy” that will burn most of them.

Scenario 5: FairVote writes: “Imagine any election in which a relatively strong incumbent was facing more than one challenger. Suppose the incumbent was likely to be the first choice of 40% of voters, but short of a majority - yet very tough against any challenger head-to-head.”

An incumbent with only 40% support of the populace is vulnerable to losing his or her seat in STAR Voting. This is a Good Thing™. STAR is about finding winners supported by a wide swath of the population, and 40% isn’t a majority. The good news is that STAR will place the minority-supported incumbent head-to-head with the candidate otherwise most supported by the populace, and therefore most likely to win against the incumbent in the automatic runoff. If the other 60% of the population have an honest “anyone but the incumbent” viewpoint, then indeed, that 60% of the population can ensure that the incumbent doesn’t even make the runoff.

However, if the incumbent is indeed “tough” against any challenger head-to-head, that means the incumbent is actually the second choice of many voters. As described in Scenario 4 above, voters in STAR will be on the balance punished if they adopt the “burying” strategy FairVote suggests.

FairVote Assertion 6: FairVote writes, “in the heat of real campaigns, many voters are more interested in electing their favorite than keeping their least favorite out office”

This assertion flies in the face of reality. As we have all witnessed in our current system, people vote defensively. They pass over their actual favorites to choose a “lesser evil” in order that a “greater evil” doesn’t win.  We see no support for this assertion when looking at actual voter behavior.

FairVote Assertion 7: “In the real world, [Instant Runoff Voting] allows voters to avoid that tradeoff [between electing their favorite and keeping their least favorite out of office], but STAR voting would likely not.”

This is actually false on both counts. A recent IRV election demonstrated how honest IRV voters can be punished for not voting strategically. In Burlington’s 2009 Mayoral race, the candidate who was preferred over all others lost unfairly. This kind of election failure had been predicted by mathematical analysis: voters ranking honestly in IRV can, in predictable ways, cause their least favorite candidate to win by putting their most favored candidate in first position.

STAR voters, on the other hand, know that two candidates advance to the runoff. They can give maximum support to their favorites, and non-zero support to their second favorites with very little risk of an undesirable outcome.

Section 6: Advantage to voters who understand the system

In this case, the section title is correct — STAR Voting gives an advantage to voters who understand that in STAR, an honest vote is a strong vote. The strategies FairVote has put forward in the article, on the balance, disadvantage the voters who use them: the “bullet” voter loses the ability to express any runoff preferences should his or her favorite not make the runoff, and the “bury” voters are more likely to create a bad outcome for themselves than a positive one.

Again FairVote repeats the fallacy that “In contrast, all voters who vote as the RCV [Instant Runoff Voting] ballot suggests are almost certainly voting just the same as fully informed voters” - despite their own article that suggests that in Australia, where IRV has been used most widely for the longest in public elections, “politically aware voters” often strategically promote major party candidates over minor party favorites in order to avoid vote-splitting.

Section 7: Tactics could lead directly to undemocratic outcomes

FairVote writes, “While it was perfectly logical to cast a ballot with 4’s for those other candidates, now they aren’t differentiated. Scores of zero for the two candidates going to the runoff would also nullify their runoff ballot -- even if giving those scores might have made tactical sense for trying to keep both those candidates out of the runoff election.”

Here’s the crux: in STAR you can’t have your cake and eat it too. It is not at all “logical” to cast the dishonest STAR ballot that the article suggests precisely because the ratings are used in both steps of the counting process. If your ballot makes “tactical sense” for the scoring phase, but then turns into an obvious nonsense vote in the runoff, that ballot doesn’t actually make “tactical sense” in STAR.  What FairVote is actually pointing out in this section is precisely that any attempt at strategic voting might backfire in STAR Voting. We agree - an honest vote is actually the best bet.

We are also glad that FairVote brings up the importance of democratic outcomes. A system that produces results that match the collective will of the electorate is critical for a functioning democracy. Recent simulation work by Harvard Statistics PhD candidate Jameson Quinn, one of the first such projects to include STAR Voting (called SRV in Quinn’s study), concluded that STAR “is undeniably a top-shelf election method, and arguably the best out of all the ones I tested.” IRV, while more accurate than our current plurality method, falls well below STAR on the measure of representation accuracy.

Section 8: Inconsistent translation of preferences into scores and runoff votes

On a STAR ballot voters show relative degrees of support. They can offer that support on the basis of whatever factors are important to them, but at the end of the day, they are putting forth an objective support level - not a subjective rating.

FairVote attempts to support their narrative here noticing that YouTube and Netflix have abandoned 5-star ratings. There are three problems with this: first, YouTube and Netflix have a ton of data on what people actually watch (and therefore “like”) making a rating level largely unnecessary. Second, from the linked article about YouTube, TechCrunch writes, “As you can see, there are some 1-stars and a huge amount of 5-stars, and then basically no 2, 3, or 4 stars.” From data we’ve already gathered on http://star.vote, and elsewhere we can see quite clearly that voters in STAR highly differentiate their votes. You can download the ballot data from when Unrig Summit attendees were asked to STAR vote in a hypothetical 2020 Presidential election here: https://star.vote/poll/csv/n9rtkpdr/ – in contrast to the data from YouTube reviews, STAR voters regularly use the spectrum. Finally, YouTube and Netflix have no “runoff” mechanism in which a user’s scores determine an outcome that affects them; there’s no incentive to differentiate nuanced preferences.

FairVote writes further “In addition, it's a challenge for voters to have their scores be used for two purposes at the same time: determining which two candidates advance to the runoff and then which candidate you prefer in the runoff.”

The reality is that this isn’t a challenge for voters at all. Voters we connect with regularly on the street have NO PROBLEM understanding that their preferences will be used to determine who gets their final vote - i.e. if they gave candidate A a higher score than B, that their runoff vote should go to A if the runoff is between A and B. That is, in fact, entirely logical.

The real “challenge” is for tactical voters and organizations like FairVote to come up with actually viable ways to game the system. So far they’ve come up empty.

Section 9: Less likely to elect “beats all” Condorcet winner

Because this section is based on a flawed notion of viable strategy, it comes to an entirely erroneous conclusion. For a thorough review of a textbook case where Instant Runoff Voting failed to elect the Condorcet candidate, as well as an analysis of what might have happened were STAR Voting used instead, check out http://equal.vote/Burlington.

Conclusion:

We are very glad FairVote spent the considerable time needed to run thought experiments for how STAR Voting might be gamed by strategic voters. As we’ve described above, neither the “bullet voting” strategy that FairVote paints on Approval Voting, nor the “burying” strategy they propose are viable tactics for STAR voters. This gives us additional confidence that STAR is ready for trial in actual elections, starting at the local level.


Showing 1 reaction

Please check your e-mail for a link to activate your account.
  • commented 2018-05-21 12:39:45 -0700
    I can’t find a place on this site to ask questions. I’m hoping someone can response to these questions. My apologies if these questions are answered somewhere already and I missed it.

    I’m a Lane County, Oregon voter. I’m interested in this idea, but don’t believe in signing petitions to put something on the ballet unless I truly believe in the idea. With that in mind:

    1) If the idea passes and Lane County gets STAR voting, what would the cost be? Is the software that prints out the voter ballots currently able to handle STAR voting? What would it cost to configure the software or buy new software? As broken I think our current voting system is, I would like an answer to this before I can sign the petition.

    2) Just trying to understand: Can you abstain from voting for a candidate? For example, what if there are 5 candidates and you fill out a star rating of 1 thru 5 for 4 of the candidates. For the 5th candidate, you don’t have enough information and you don’t want to vote for that candidate. You are happy to let the majority of voters make the decision about that candidate.

    I ask this question because I saw something above on this page about zero stars. To me, that means that you rate the candidate so low that you give them less than even one star. That’s not what I would want to have happen. I would just not want my vote to count for that candidate.

    There are sometimes races where I just don’t have an opinion. I can’t responsibly vote. I wouldn’t want my two zero stars to count as zero stars. I would want to be an abstention.

    Thank you for any additional information you can provide.