Ranked Choice Ballot Example

Elections are surveys and ballots are a form of questionnaires.  So, as a survey designer and trainer, I look at elections and the new push for “ranked choice voting” through a bit of a different lens than most, dating way back to the infamous butterfly ballot in Florida in 2000.

A movement is afoot to change our election process to what’s called “Ranked Choice Voting” (RCV).  It’s a different voting process with some purported benefits – and some not-discussed shortcomings.  Thankfully, some actual experiences highlight the pros and cons of RCV.  And in a separate article, I address how RCV will change election strategy.

The Benefits of Ranked Choice Voting

RCV’s stated goals are:

  1. Fairer elections – though “fair” is not defined
  2. More efficient election processes since multiple rounds of voting would be unnecessary
  3. Reduction or elimination of “strategic” voting.

How Ranked Choice Voting Works

RCV adds a twist.  In a nutshell, here’s the concept.

Today when you cast a ballot, you vote for one person for an office.  (For the moment let’s put aside some elections where multiple people are elected.)  Generally, whoever wins the plurality of votes wins the election.  In some situations, a majority vote is required to win.  If no one gets a majority in the first round of voting, a second runoff election is held among the top two finishers.

With RCV that second round is unnecessary.  Under a ranked choice voting system, a ballot would have multiple columns representing a 1st choice, 2nd choice, and 3rd choice.  The voter selects candidates for the various choice levels.  For example, I might select Cynthia as my 1st choice, Sam as my 2nd choice, and Pat as my 3rd choice.

Let’s say the vote totals in percentages are:

  • Sam: 42%
  • Pat: 37%
  • Cynthia: 21%

In today’s typical voting process, either Sam wins or there’s a runoff between Sam and Pat.  But under RCV, ballots that voted for Cynthia, who got the lowest 1st choice vote total, would be examined for their 2nd choice votes.  For example, let’s say:

  • Sam: 42% + 5% 2nd choice votes from Cynthia’s ballots = 47%
  • Pat: 37% + 16% 2nd choice votes from Cynthia’s ballots = 53%

Pat would be the winner, which presumably is how the runoff would have gone.  Maybe.  That presumes that voters seriously thought about their 2nd choice votes and that Cynthia’s voters would have shown up for the final election.

Elections force people to make choices, trading off the benefits of one candidate over the others.  But is rank choice the best method to engage in trade-off analysis?  That’s debatable.

The Design Problems with Ranked Choice Voting

As noted on the Ranked Choice Voting website, voters have some important instructions to follow:

Ranked Choice Ballot Voter Instructions

Instructions for voters in Ranked Choice Ballot

These instructions point out some of the shortcomings of ranked choice voting.

As a survey designer, I never use so-called forced ranking questions.

This question type tries to measure the rank order importance of factors to some outcome.  We might ask respondents:

  • What drove your decision to purchase product X?  Please indicate the importance of each of the following 5 items by selecting 1 for the most important, 2 for the second most important, etc.
  • What are the most important issues in this election?  Please indicate the order of importance by selecting 1 for the most important, etc.”

Why don’t I use this question type?  Forced Ranking (Rank Choice) question type is prone to:

  • Respondent Annoyance. If the list is long, ranking every item can be tedious, and the difference in preference after the first few is likely trivial.  I recently saw a survey that asked to rank order 19 (sic) items!!  You can be certain that after the first 3 or 5 were ranked by respondents, they were then just clicking on buttons to get it done.
  • Respondent Error. Think about the Florida butterfly ballot from the 2000 election depicted nearby.  No matter how clear the instructions, some people would screw it up.  Fact is, instructions on surveys or ballots are like safety instructions at the start of an air flight: no one pays attention to them.

Same here.  On a paper ballot, some people will vote for two 1st choices or vote for no 1st choices –  maybe as a way of saying they don’t really like any candidate.  Such ballots would likely be declared invalid for that office election.  Even voting for the same candidate for 1st, 2nd, and 3rd choices could invalidate the ballot.  Is that fair?

  • Administrative Dilemma.  Given that some ballots will be filled out incorrectly — and many will — what does the election bureau do with those ballots?  The State of Maine has an 11-slide powerpoint deck to help educate voters.  The last few slides show how the state has decided to handle incorrect ballots.  Someone has to decide this, but at some point aren’t we looking for dimpled ballots and counting chads?  Would we want an election decided by those “redeemed” ballots?  And how were these rules decided?  Did the people through their legislative representatives vote on them?
Maine RCV Instructions

Maine RCV Instructions

A friend of mine in Maine, which uses RCV, told me that he screwed up his ballot for the first time in his life.

Electronic voting systems may have safeguards that will prevent these mistakes, but then we increase the annoyance factor.  The error messages will lead some to just skip the voting.

I will suggest the annoyance and error issues will be particularly true for older voters whose faculties have been lessened by time.  My dad was an electrical engineer with 25 patents to his name.  I saw his logic skills deteriorate later in life.  Very sad.  He would have had trouble with this system in his later years.  Does this system discriminate against such voters?

And remember that the voting method has to be ADA compliant!  (Americans with Disability Act)

Rank Choice Usability History

You might think I’m raising a non-issue.  I’m not.  I have followed electronic survey tools since their inception in the late 1990s.  The format for ranking questions constantly has been  in flux as designers try to find a way of presenting the question to respondents that avoids error and annoyance. Formats include:  a matrix of radio buttons, drop down boxes, drag and drop, among others.  SurveyMonkey I believe has changed this question format three times in the past 5 years.

I have tried rank choices in surveys and have had to throw out the question’s data because there was too much error in the responses, making the findings meaningless or, worse, misleading.

Transparency in Ranked Choice Voting – What If There’s a Bunch of Candidates?

The simple example of three candidates for an office makes ranked choice voting seem easy to understand and transparent.

Transparency is important.  If an electorate can’t understand how someone won an election, faith in the democratic process is undermined.

What if there’s a bunch of candidates, which certainly is the case for most all major offices.  The Florida butterfly ballot had ten candidates for President.

The concept, according to the Ranked Choice Voting website, is to iteratively eliminate the “candidate with the lowest number of votes.”  So, in a field of, say, 10 candidates where no candidate gets a majority in the first round, the candidate who got the fewest 1st place votes would have his/her ballots examined for those voters’ 2nd place votes.  Those votes would be allotted to the remaining candidates as indicated.

The process would repeat until a candidate gets a 50% majority.  Note:

  • The order of candidates could change with each allotment of 2nd place votes.
  • The total count of votes used to decide the election outcome will go down with each round.  Why?  Not all voters will have made 2nd place selections or their 2nd place choice will have been previously eliminated, known as an “exhausted ballot”.  The result: the 50% majority threshold  will drop with each round.

How transparent is this process?  For an actual example of the transparency of the process, I’ve explained the RCV process in the context of the Maine 2nd Congressional District race in 2018.

Imagine election nights where our esteemed news anchors try to explain how someone got elected!!  This will make explaining the US Electoral College seem like a piece of cake.

A suggestion for news bureaus: A reporter should not cover an RCV election if they can’t explain the difference among the four data types – categorical, ordinal, interval, and ratio.  (Think Jennifer Lawrence could describe the difference?)

Transparency Involving 3rd Choice Votes

And what about the 3rd choice votes?  How do they come into play?  The Ranked Choice Voting website is silent on that.

I would humbly suggest that if an election outcome relies on 3rd place votes, it would be better to flip a coin — or have a runoff!

Next time you take a survey with a forced ranking question, listen to yourself as you formulate your response.  You will think (perhaps) strongly about the 1st ranked item, a little less for the 2nd, and even less for the 3rd.  Why?  They’re not important!

Would we really want elections decided by 3rd place votes?

Fairness in Ranked Choice Voting – Again with a Bunch of Candidates

Here’s a ranked choice voting example.  (Candidates and their votes are color coded to guide the reader through the vote-allocation logic.)  Imagine these votes totals with only five candidates and 1000 initial total ballots:

  • Candidate A: 420
  • Candidate B: 230
  • Candidate C: 160
  • Candidate D: 120
  • Candidate E: 70 – lowest vote total

Candidate E gets eliminated and the 70 ballots that chose E as their 1st choice are examined for their 2nd choice candidates, resulting in:

  • Candidate A: 420 + 7 = 427
  • Candidate B: 230 + 8 = 238
  • Candidate C: 160 + 5 = 165  — lowest vote total
  • Candidate D: 120 + 50 = 170

(Note: 7+8+5+50 = 70)

Candidate D vaults into 3rd place with 170 votes.  So, Candidate C is now eliminated and the 160 ballots that had C as a 1st choice are examined for their 2nd choice candidates, resulting in:

  • Candidate A: 420 + 7 +10 = 437
  • Candidate B: 230 + 8 + 20 = 258  — lowest vote total
  • Candidate D: 120 + 50 + 130 = 300

(Note: the 5 ballots that had Candidate E as a 1st choice and C as 2nd choice are “exhausted” and will not be used in the final talley.  Thanks to reader Kathy T. for catching a mistake I had originally.  See, I told you this was confusing!)

Candidate D jumps into 2nd place with 300 votes.  So, Candidate B is now eliminated and the 230 ballots that had B as a 1st choice are examined for their 2nd choice candidates, resulting in:

  • Candidate A: 420 + 7 + 10 +40 = 477
  • Candidate D: 120 + 50 + 130 + 190 = 490

Candidate D wins with 50.6% of the “vote” !

(Note: 33 ballots were “exhausted” in the process and I have assumed that there were no “undervotes,” meaning that every ballot had a 2nd choice selection, which is highly unlikely.)

~ ~ ~

How fair is this process? (not to mention transparent!)

The above may be an unlikely and uncommon scenario, but could it happen?  Sure!  Would it be “fair” for a 4th place candidate to win an election because he or she was the second choice of a lot of voters?  Only if you’re a supporter of Candidate D!

Doesn’t it also feel that the people who voted for marginal candidates are getting a “do over”?  Should their 2nd place votes count the same as someone else’s first place vote?

Is that fair?

Do We Have a Winner with a Majority?

Also note that the winner did NOT get over 50% of the original 1000 votes cast, but only topped 50% when undervote and exhausted ballots were eliminated from the denominator.  So is it a true majority or a manufactured majority?

The Shortcoming of Ranked (vs. Interval) Choice Data

Ranked choice voting has a conceptual problem.

We know the voters’ order of candidate preference, but we do NOT know the relative preference.

To know relative preference a consistent unit of measurement – think of a ruler – must be used to measure the difference in voters’ preference for each candidate.  (That’s the distinction between ordinal and interval data types, as we learned in our college stats class.)

Why is this difference important?

Let’s go back to the above example.  Candidate D was put over the top by the 2nd place choices on ballots with Candidates B, C, and E as 1st choice.  But what if those 370 (50+130+190) voters viewed Candidate D as a barely palatable distant 2nd while the 57 (7+10+40) voters who chose Candidate A as a 2nd choice viewed Candidate A as really close to their 1st choice?  (See this article of mine for a further discussion of such a phenomenon that I experienced.)

If this were the case, it’s now even less fair for Candidate D to win based on those lukewarm 2nd place votes.  You can expect losers to raise such issues – legitimately.

Someone will argue we can just take a mean rank of the ranked choices.  Mean rank is an oxymoron.  It’s not a legitimate calculation and would not magically tell us relative preference.  (Calculating a mean requires the data have interval properties.  Sorry about bringing back memories of stat class.)

To measure relative preference for each candidate – versus rank order preference – we’d need to use rating scales for each candidate.  We’ve all seen those on surveys; you know, the 1-to-10 scales.  Would we want to choose candidates based on the highest rating?

Alternatively, we could use a fixed-sum format where we have voters allot 100 points among the candidates based on their relative preference.  As a survey professional, I much prefer the fixed-sum approach over rank choice for measuring importance or preference.  However, I can’t imagine using fixed-sum or ratings in an election — but I’d take it over ranked choice voting.

But you can expect attempts to “fine tune” RCV, which will probably make elections less transparent.

Here’s an article that discusses different ways of measuring preferences.

Now, About the Strategic Voting Claim…

I was at an event recently where the Ranked Choice Voting folks had a booth.  A young ardent person button-holed me to pitch RCV.  He didn’t understand ordinal versus categorical data so our conversation didn’t last long.  He was frustrated that I would challenge the alleged superiority of ranked choice voting on conceptual grounds.

One claim he made was that RCV would eliminate “strategic voting” — meaning people won’t vote for the lesser of two evils to avoid having a deplorable candidate get elected.  Candidates would be incented to appeal beyond a narrow base.  Strategic voting will continue with RCV; the strategy in voting — and fielding candidates — will simply change.

Ranked choice voting will likely lead to wholesale changes to elections.  That may be the whole point.  I think I understand why some are pushing this approach under the guise of “fairness.”  They understand how RCV will change election strategy and how to use RCV to win elections.

Candidate Expansion — A Strategic Decision

RCV will likely a vast proliferation in the number of candidates. Democracy at work!  Let people select their first choice rather than the lesser of two evils — then select the lesser evil in the 2nd choice — or 3rd choice.  No more “spoiler” candidates.

With numerous candidates the likelihood of a majority vote in the 1st “round” drops, meaning the decision will be made through multiple rounds of murky 2nd (and 3rd) choice voter allotments.

Some of these candidates may simply be foils to deny a majority to anyone and get to the RCV runoff.  Remember JFK’s first congressional race.  His father, Joseph P. Kennedy, got a Joseph Russo to run whose name was identical to someone viewed as a chief opponent, splitting his vote.

Transparency and fairness issues will mount.

  • Who’s going to audit the opaque vote allocation process?
  • Will the electorate, which will never understand the allocation process, have faith that the most popular candidate actually won?

Personally, I’d rather see a runoff between the candidates with the top two vote totals.

The runoff focuses the mind on the choice for those who voted for the fringe candidates — if they bother to vote.

And it’s transparent.

Predictions for an RCV System

I suspect that campaigns will quickly tell their ardent supporters to not cast any 2nd place votes.  That will reduce the possibility of completing the ballot incorrectly, which is the last thing a campaign wants – and why would you help out another candidate?  Remember, ranked choice voting says voters don’t have to provide a complete ranking of all candidates.

However, campaigns will try to coax lukewarm supporters “in the name of fairness” to “cast a complete ballot.”  That might help get their candidate over the top should 2nd choices be examined.

I do have an insidious prediction.  In some election we will see robocalls placed on election eve to likely voters for a certain candidate instructing them on how to complete their ballot.  The instructions will be intentionally wrong — say, telling the voter to select their candidate for all three ranks — to get the ballot disqualified.

Think I’m being paranoid?  Then you’ve never worked in a campaign.

With RCV Why Would You Bother with Primary Elections?

Ranked choice voting may eventually lead to calls to eliminate primaries and move to what are called “jungle elections” with no party affiliations.  California is now doing that with its primaries to be followed by a general election between the top two winners, but with RCV that second stage becomes less necessary.

Maybe that will be good, but we should be aware of possible unintended consequences, especially as it affects political parties.  Let’s remember how campaign finance “reform” has reshaped our election processes, pushing money to less accountable, non-party groups to wage the election battles.

We know from experience that loopholes will be found in all well intended systems.  And the law of unintended consequences will rule forever.

But Who is the Ranked Choice Voting Resource Center?

Now pardon me for being suspicious, but who’s funding this movement?  The RCV website lists 4 members of a “project team,” but that’s it.  This is an extensive movement.  As of this writing, the website does not ask for donations, but someone is funding this.  Who?  Certainly not the 4 people on the project team.

As I explored the RCV website, I noticed a startling omission: there’s no street address in the footer, which is a big no-no in Google search engine rankings.  From where does this project team operate?

RCV Organization Info

There’s only a phone number, and it’s a phone number “shared” with FairVote.org.  Why the obfuscation?  Why do they hide the connection between FairVote and RCV?  By not being straight about the connection, they are inviting questions.

If you look at the biographies of the people at FairVote, you’ll see positions at campaigns but without naming the campaigns.  Even their LinkedIn profiles have been cleaned of candidate affiliations.

FairVote Typical Bio

If this were a bi-partisan group, then showing contrasting affiliations would demonstrate that bi-partisanship.  FairVote has all the trappings of a partisan effort disguised as non-partisan.  What’s the end game?

I go back to the campaign finance “reform” issue.  Someone figured out that so-called “527 committees” could fund elections and be in compliance with McCain-Feingold.  Did McCain-Feingold reduce the influence of money on our elections?  Hell no.  It just pushed the money into the shadows.

What will be the unintended consequences of ranked choice voting?  Let’s let a few states be our guinea pigs before we implement what appears to be a partisan ploy.

Again, in an article I wrote after this one, I think I figured out the strategy to win RCV elections.