In Some Elections, Second Best Might Be Good Enough

On Tuesday, Nov. 8, Portland, Maine will hold its first mayoral election in 88 years. (The mayorship previously rotated between city council members.) But it’s going to be unusual for another reason: voters will use a ranked choice system, which means they have to list the 15 candidates in order of preference. An image of the ballot appears below. Here’s the AP’s David Sharp reporting on the complexities:

The ballot is too complicated to be understood by the city’s voting machines, so only first-place votes will be announced on the night of the election, said Caleb Kleppner, vice president of TrueBallot Inc. The final outcome of the race won’t be known until the following day when the ballots are scanned and all of voters’ rankings are extrapolated, Kleppner said.

Portland voters won’t be the only ones making multiple commitments on election day. A handful of municipalities across the country use similar systems. San Francisco has been using ranked choice voting (RCV) since 2004, and voters in next week’s mayoral election there will be allowed to rank their top three candidates.

(Digital Vision)

In San Francisco, it works like this: if no single candidate gets more than 50 percent of first-choice votes, the lowest-ranking candidate will be eliminated from the pool. The ballots that listed the eliminated candidate as first choice will have their second-ranked choice tallied up instead. Since voters’ second choices are built-in, the city avoids having to hold a costly runoff.

Steven Hill, a political consultant and the architect of San Francisco and Oakland’s systems, estimates that San Francisco has already saved $7.2 million using RCV and will save another $3 million by avoiding a runoff this year. The city’s Department of Elections wouldn’t provide numbers to confirm, but they did note that the last city-wide runoff election, held in 2003, cost $3.5 million.

One of the strange things about these elections, though, is candidates don’t necessarily need the most first-choice votes to win the election.

Sound confusing? There has been plenty of griping about the complicated system, including criticism that the ballots alienate minorities. But Steven Hill sees evidence that the system has actually increased minority representation. For instance, the number of racial minorities on the Board of Supervisors has doubled from four to eight since the institution of RCV. Because minority communities don’t need to throw their weight behind a single candidate to avoid a split vote, Hill thinks they end up with more of a say.

Another possible consequence of RCV is less mudslinging. Because a vote for another candidate isn’t necessarily a vote against any other, politicians don’t need to engage in as much cutthroat smearing to sway voters away from other candidates. Hill says, “with ranked choice voting, incentives are different. The winner may need the second rankings from the supporters of some of his or her opponents. So that means you have to be more careful about what you say about those opponents.” So on election day, you can (theoretically) play nice.

In our latest Marketplace podcast, Stephen Dubner talks with Kai Ryssdal about how special interests have greater influence when turnout is low. Because turnout in runoff elections is typically low, RCV may even help curb special interests’ influence and increase voter participation in the eventual outcome.

Do any readers out there live in municipalities that use vote ranking? Have you seen it change the incentive structure for politicians or voters?

Leave A Comment

Comments are moderated and generally will be posted if they are on-topic and not abusive.

 

COMMENTS: 30

View All Comments »
  1. Mike says:

    Ranked choice is also used in Oakland, CA, where last year Mayor Jean Quan won due to rank choice – and as an Asian in one of the cities that is commonly seen as an African American stronghold, this was quite a feat.

    This year in the SF election, since Mayor Newsom has vacated and the interim Mayor Lee hasn’t had time to build a strong base, the ballot choices were off the wall, something like 10 or so candidates for mayor (Lee, Dufty, Alioto, Avalos, Baum, Herrera, Adachi, Ascarrunz, Rees, etc) so there’s no chance of avoiding a run-off without ranked choice. A lot of candidates stand for similar positions as other ones, so instead of being even more cutthroat to differentiate themselves, they are instead listing similar candidate as their second choice – for example, “Vote for me, Lee as your first choice, and make Herrera your second.”

    Well-loved. Like or Dislike: Thumb up 12 Thumb down 1
  2. gb says:

    Australia has used this system for a long time, and our political parties are just like those in the US, each one is absolutely correct while the other one is a bunch of dribbling idiots.

    Federal and state elections use preferential voting. You can just put a 1 beside your preferred candidate, but if they are eliminated in the process, your vote is not counted in subsequent rounds. For the senate, you can either vote in order for the ~50 candidates or you can tick they political party of your choice and the order that they have proposed will apply to your vote.

    Thumb up 4 Thumb down 0
  3. Dale Sheldon-Hess says:

    Hidden due to low comment rating. Click here to see.

    Disliked! Like or Dislike: Thumb up 6 Thumb down 14
    • Rob Richie says:

      In Dale’s zeal to promote another voting method, he goes out of his way to muddy the waters of the ranked choice voting story in San Francisco. For example:

      * On racial minorities, every seat on the Board had been elected in district elections with runoffs before RCV was adopted. Racial minorities lost some key runoff races to white candidates in that period.

      * In the city’s last 14 runoffs in 2000-2003, ten of the races had voter participation that declined more than a third from the first round, and in eight races the December winner had __fewer__ votes than the November first-round leader. Meanwhile, no RCV winner of course ever saw a decline in votes, and participation dipped below 74% of the first round count only once.

      * The SF Ethics Commission strongly backed moving to RCV as quickly as possible because of the impact of money in runoffs. For instance, in the 2002 runoffs, independent expenditures quadrupled in the runoff round over the first round.

      * The SF Department of Elections head has testified that SF has saved money even with a number of unique one-time costs for being the first California city to use the system. Costs have risen in other elections for other reasons. There’s not any doubt it’s saving the city about three million dollars this year.

      * As to outcomes, the Bay Area has had dozens of RCV elections. The “Condorcet” winner (the candidate who defeats all other candidates 1-on-1) has won every single time, including in four races last year where the first-round leader did not win the race. You would think Dale would grudgingly admit RCV did a good thing here — but no. Instead, he promotes an approval voting system that even one of his allies Jameson Quinn admits would have elected the Condorcet __loser_ in the RCV election in Burlington in 2009 — e..g, the candidate in the top three who would have lost to each of the other two candidates 1-on-1.

      Is RCV perfect? No. Is anything perfect? No again. Is RCV making a positive difference in a lot of elections, including races with three or more candidates in five cities next week? Yes!

      Well-loved. Like or Dislike: Thumb up 17 Thumb down 2
      • Jim Riley says:

        There were 15 runoffs in San Francisco between 2000 and 2003. What is a “key runoff”? Mabel Teng lost a “key runoff” in 2000, but won a non-key runoff in 2002, because she was Asian?

        There were 8 at-large supervisors who attempted to switch to a district seat in 2000. One had been appointed, and never run for office and finished 6th. Under the at large slating system she would have been an “incumbent” and elected because voters had “heard of her” and she would have been in all the slate advertizing.

        Those doing the slating would provide racial balance because it meant that there were less likely to be challengers. But supervisors elected at large might not make good district supervisors or good district campaigners. The other 7 at large candidates made it to the 2000 runoff. The 2000 election was on the bottom of a presidential ballot, and it would be hard for a candidate for a district of 60,000 to afford media in a multi-million person TV market (San Francisco-San Jose-Oakland). Name recognition would get a candidate to the runoff.

        The runoff was separate, and you would have actually campaign in the district and get your supporters to vote. Four incumbents who had been elected at large, were defeated by hard working challengers. This simply won’t happen under IRV, so what you need there is name recognition, especially since you are going up against presidential or gubernatorial elections.

        There were 3 White v Asian runoffs. Two of the Asian candidates were at large incumbents.

        Aaron Peskin had close to the total vote for the 2nd, 3rd, and 4th, combined and easily won the runoff. He would have been elected under IRV. But because the second place candidate was Asian, this may have been a “key” runoff. Peskin was re-elected under IRV, defeating an Asian candidate. After he was term-limited he was replaced by an Asian.

        Mabel Teng lost to Tony Hall. There has not been any apparently Asian candidates in that district since. So it appears that she was simply an at large supervisor who happened to live in the district, and might never have been elected from the district. She was later elected to a citywide office in a conventional runoff.

        Jake McGoldrick defeated incumbent at large supervisor Mathew Yaki in a runoff. But he was easily re-elected in an IRV election. After he was term-limited, he was replaced by an Asian.

        The 4th Asian on the board is from a district that had no history of apparent Asian candidates before 2010.

        So you might be able to make a case that at large elections were better for minorities if they could gain token positions on the slate, and that the change to district elections was harmful (that argument was made at the time). And you might be able to make a case that the public name recognition achieved by slated candidates, might have got them elected in an IRV district elections, and helped preserve the slated candidates. But that is hardly a ringing endorsement of IRV, is it.

        Perhaps, Ed Jew would not have been elected under a conventional runoff, particularly if the fact that he did not live in the city, let alone his district, had become an issue. But maybe that is the sort of negative campaigning that IRV relieves us of. Does the cost of the special election in his district get charged to the use of IRV?

        And it is quite likely that it takes a while for the political activity fostered by district elections to take hold. If the first elected office you can aspire to is to be slated as an at large supervisor in a city of 700,000; or spend a whole lot of cash to defeat the slate; or not raise much cash and finish 13th place in an election for 5 places; most sane persons aren’t going to run.

        It is the district elections, which provide an entry level at a much smaller area, and term limits that provide the opportunity. It is quite unlikely that IRV had anything to do with it.

        Thumb up 1 Thumb down 0
    • Richard says:

      I don’t see why RCV would be a waste of time. I don’t care that much whether turnout is affected. The cost issue seems to be a minor one. The effect on mudslinging would be difficult to quantify, but a non-binary system would affect candidate behavior in some way because being a voter’s second choice would be more valuable under RCV. But I don’t care that much about mudslinging, either.

      I am interested in the effect on outcomes. Under the current system for US House races, candidates have to win a primary election in which they generally appeal to an extreme and then they pursue a 50-percent-plus-1 strategy. The general voting public is then left with a choice between two extreme and often unappealing candidates. So my question is: What data is there on the type of candidates who get elected under RCV? I would think that a centrist candidate would have a much better chance here, but I suppose there could be some reason that’s not true.

      Well-loved. Like or Dislike: Thumb up 6 Thumb down 0
      • Dale Sheldon-Hess says:

        Good question!

        Unfortunately, your suspicion that RCV would help elect more-moderate candidates is mostly not true. Perhaps an example is in order:

        35% of voters vote A > B > all others
        32% of voters vote B > all others
        33% of voters vote C > B > all others

        Even though 67% of voters prefer B over C, and 65% of voters prefer B over A, B will not win this election under RCV. Candidate B is what’s called a “Condorcet winner”, and while there are a number of rank-order-ballot based election methods which always elect the Condorcet winner (if one exists, which it doesn’t always), RCV is not one of them, which is one of the reasons I say that RCV is the worst possible election method–other than plurality.

        The method I advocate for, approval voting, has been shown via computer election modeling to be more-likely to elect the true Condorcet winner than any rank-order-ballot based election method, when voters try to vote strategically.

        The graphic on this page: http://rangevoting.org/BayRegsFig.html shows the computer-simulated results of candidate-quality under various election methods; RCV is shown under one of its previous marketing names, instant runoff. (And yes, range voting is theoretically better than approval, but approval is a bit simpler.)

        Thumb up 5 Thumb down 1
      • Rob Richie says:

        Dale won’t accept that competitive campaigns with approval voting and range voting would be subject to massive discussion of its practical flaw: you can’t indicate support for a second choice without that counting directly against your first choice. We know from all our work in ranked choice voting elections that this is a HUGE issue — we have to spend a great deal of time reasssuring people that indicating support for a second choice won’t hurt your first choice. If we couldn’t say that, the system would collapse into a morass of tactical discussion.

        In contrast, Dale attacks the theory of RCV and can look at results after they happen and say versions of “ha, if Republicans had nominated Romney instead of McCain, then Obama might have lost”, but her’s the thing: none of it affects RCV campaigns as they happen. See http://www.approvalvoting.blogspot.com for a lot more detail on why this WOULD happen with approval voting and range voting.

        So the practical choice for electing one person: plurality voting systems (subject to split votes and much talk of “spoilers” whenever more than 2 candidates run ), traditional runoffs (narrow the field to two in one fell swoop, then have a second election between those two) and ranked choice voting. I’d take either runoffs or RCV over plurality voting any day. But within the choice of RCV and runoffs, there are a lot of sensible arguments for RCV: more gradual reduction of the field during the count, decreased impact of big money (which can have its greatest impact in one-on-one races, where negative attack ads are most effective), equalized turnout (runoffs often subject to very different levels) and so on.

        Thumb up 5 Thumb down 1
      • Richard says:

        Examples and computer-simulated results of this type are not data. But I do get that the middle candidate in my three-candidate race would have to be first or second in first-place votes or be eliminated. A weak moderate candidate isn’t going to beat two strong extreme candidates.

        On the other hand, I would guess that adding candidates would make the extreme positions less likely to win. And I notice that the Portland race has 15 candidates. Do RCV races tend to attract more candidates?

        Thumb up 2 Thumb down 0
      • Makfan says:

        I do think that RCV is very different. I don’t particularly find it appealing when compared to a runoff where I have two candidates and get to make a selection between them. It may save money, but it is quite different. I felt like I had to look at polls to form some sort of strategy in picking which candidate to mark 2nd and which to mark 3rd.

        Thumb up 0 Thumb down 1
    • David Cary says:

      The analyses published by electology.org and “The Center for Election Science”, an advocacy for approval and range voting, are flawed, incomplete, and misleading.

      The voter turnout analysis is self-described as crude, has significant calculation errors, and conveniently stops its self-improvement as soon as it starts to find greater turnout under RCV. What does “statistically determinable” mean anyway? They’re comparing full populations, not random samples. A more complete analysis shows a much larger increase in voter participation under RCV.

      The wide-spread consensus in San Francisco is that mudslinging has been reduced under RCV. Even politicians who dislike RCV concede that point. The Sheldon-Hess criterion, that all negative campaigning has been eliminated, is overly severe. More sloppy analysis or disingenuous rhetoric?

      The cost analysis that electology.org pushes is not credible. They fail to account for the many other factors, other than RCV savings, that determine the SF Department of Elections budget. Notably, they chose not to check whether any differences were “statistically determinable”.

      Thumb up 3 Thumb down 0
    • Jeremy Faulk says:

      Take a look at your financials again.
      2001: $9M
      2003: $9M
      2005: $10M
      2006: $11M
      2002: $14M
      2004: $15M
      The order in which you list them is slighted at best and manipulative at worst. ’01, ’03, ’05, are local election years I assume. They are the lowest cost. 2006 and 2002 are midterm years which typically have a higher turnout than presidential years, and 2004 was a presidential election. Based on the numbers you provided, the cost of a midterm election decreased by $3M while the cost of a municipal election grew by about $1M over 4 years (’01-’05). You provide no comparable for a presidential election year. Stop trying to manipulate people and be honest.

      Thumb up 1 Thumb down 0
  4. Andrew says:

    The System of Voting you describe is the Preferential System of voting and is use in all Australian States and the Federal for Lower house single member seats. This system is also used in local municipality elections in most local government elections. Further it is also used with variations in the upper houses of state and federal governments using a proportonal preferential system. Please consider consulting http://en.wikipedia.org/wiki/Electoral_system_of_Australia . Many times in Australian elections gaining the highest number of first choice votes does not guarantee that the candidate will win; minority parties preferences can change the course of the election.

    Thumb up 3 Thumb down 0
  5. Paul D'Agostino says:

    RCV (or preferential voting as it is known in Australia) is the norm for all levels of government in Australia. We don’t have run offs. The advantages and disadvantages have been torn apart in detail. Like the USA we tend to have two major parties with the occaisional independent but periodically we have a third small party. Currently The Greens are that third small party and yes – running second for the ALP (= Democrats) often still results in elections. Three horse races sometimes emerge when there are strong – usually rural based – independents. It gets even more complicated for the Federal Senate elections – 6 senate positions per election (half senate) per state with the option of party list voting with the party nominating preferences (you put a one ina box) or putting preference numbers in boxes on a really large ballot paper. The result can take a couple of weeks for the last position to be declared as there are lots of micro parties and other nutters. The deals with the major parties can result in a micro parties accumulating preferences from other micro parties and getting enough preferences to be elected. A quota for a senate seat is 1/(n+1) of the total voter turnout +1 vote where n = the number of senators (=6). There have been instances of senators being elected with 2% of the primary vote and accumulating preferences to being elected.

    Well-loved. Like or Dislike: Thumb up 5 Thumb down 0
  6. Laffer Curved says:

    Another Australian here…
    Looking at US and British media, I think our politicians do manage to be more civil to each other, both during elections and in parliament. We may still be dominated by two parties, but they pay attention to the third party voters. In plurality voting, a vote for an independent is a vote lost, while in preferential voting it’s worth trying to convince that independent voter to preference you second.

    The preference breakdown following elections also gives interesting insight into what people think – much more nuanced than simple plurality voting.

    As for people understanding it – a while back, my town had a mayoral election with four candidates. The incumbent (a very polarizing figure) scored about 46% of the first preferences, while the other three were astonishingly close to 18% each. However, it turned out that the votes for them were “anything but the incumbent” – preferences flowed to each other until the last one had over 50%. Enormous numbers of people complained that this was unfair, and plurality would have been better.

    I’m disappointed in my fellow citizens, since I think this election shows the benefits of preferential voting – the incumbent will always have a huge advantage in name recognition, so preferential is fairer than plurality. On the other hand, the outcome between the other three candidates was almost random – although if any of them had run against the incumbent in a two-candidate plurality election, they would have won.

    I see the mathematical advantages of Approval Voting / Score Voting / Range Voting. But, why do supporters of that method only turn up when preferential voting is discussed? Most elections in the US are pluralities, and many have clearly nonsensical outcomes. Your Republican Presidential Primary is looking to be especially ludicrous. So why not promote a more nuanced vote whenever that is discussed?

    Well-loved. Like or Dislike: Thumb up 10 Thumb down 0
    • Dale Sheldon-Hess says:

      “I see the mathematical advantages of Approval Voting / Score Voting / Range Voting. But, why do supporters of that method only turn up when preferential voting is discussed? Most elections in the US are pluralities, and many have clearly nonsensical outcomes. Your Republican Presidential Primary is looking to be especially ludicrous. So why not promote a more nuanced vote whenever that is discussed?”

      Because then I’d *never* get any work done! ;)

      So, better (I think?) to jump in when people are already primed to think about election-method reforms. Although I have tried such “cold calling” when local elections get a nonsensical result (happened for school board recently.) Apparently though, you don’t live in my local :)

      Thumb up 2 Thumb down 4
  7. Carl says:

    I don’t know about changing incentives for politicians, but a system like this would certainly change the incentives for voters. And in the process, it could actually make the outcomes worse. Here is my logic:

    Over the years I have often voted for my “second best” candidate, because I expect a close election and my first choice has little chance of winning. This system would give me the option to vote for both.
    Of course, there is a flaw in my thinking here. This system does NOT mean that a vote for my 2nd choice will actually count. Dale Sheldon-Hess presents an excellent example of the problem in his second post.
    However, this systemic flaw is likely to be lost on most voters (including me – I didn’t notice the flaw until Dale pointed it out). Many will mistakenly believe that their 2nd & 3rd choices will actually matter. Thus many voters could CHANGE their vote to a candidate with lower chances of winning.
    This in turn could negatively impact the electoral chances of my “second best” candidate. From my point of view, the electoral outcome could be made worse.

    Thumb up 2 Thumb down 2
    • David Leppik says:

      As someone who works with opinion surveys professionally, it strikes me as odd that our electoral system purports to weigh opinion, and yet is so out of whack with how psychologists measure opinion. For example, a well-written survey might ask: “How satisfied are you with X? Very Satisfied, Satisfied, Neither Satisfied nor Dissatisfied, Dissatisfied, or Very Dissatisfied?” Similarly, if you want to measure opinions about candidates, you might ask (for each candidate):

      How strongly do you support or oppose Barack Obama (Democrat, incumbent) for President of the United States?
      Strongly support
      Support
      Neither support nor oppose
      Oppose
      Strongly oppose

      Now, how you would count the votes I’m not sure: each method would have its own advantages and disadvantages. You might, for example, decide that avoiding strongly opposed candidates is more important for a democracy than electing strongly supported candidates. But at least you’d be starting with an accurate measure of voter opinion and intent.

      Thumb up 3 Thumb down 0
      • Dale Sheldon-Hess says:

        Indeed, and there is something of a debate on these issues.

        In approval voting, there are only two choices: approve, or disapprove. The winner is whomever has the most “approves”.

        Range voting has more choices; building off your example, “strongly oppose” would be worth 0, and “strongly support” would be worth 4. The winner under range voting is the candidate with the highest average (mean) score.

        And then there’s majority judgment, where the winner is the candidate with the highest median score.

        There are a lot of options, but one thing does seem clear, and that is that any method based on ratings, like you describe, would probably be a better choice than any method based on rankings, like RCV is.

        Thumb up 1 Thumb down 4
  8. Michael Lewin says:

    The issue was debated ad nauseum in Britain this year (under the name Alternative Vote) as we had a referendum on it (the result was not to adopt it)

    Thumb up 2 Thumb down 0