Critique this before I publish it on Blogger

#1

Background

Fed up with the lack of resources that show the problems with RCV and the benefits of Score without being written by PhDs for PhDs (that’s scorevoting.net), I decided to write my own blog post…
…except that I feel like I went on too long and it may still be too long for an average person to understand (; didn’t read).

(The bolded colors in the election examples are rendered in their own color, with a black background on yellow, but it does not show on the forum.)

This one is directed at people who know of RCV already.


Voting Reform: Tricks and Traps

If you live in the US, chances are you probably at some point thought that the government is broken. Politicians lie, lobbyists buy off Congress, no one can agree... But perhaps you heard about people wanting to change the system to make it better. For example, the organization "RepresentUs" posted a video about a strategy to break down corruption here. Almost everything in the video is spot-on, but there is one mistake that could undermine the benefits.

That one mistake occurs at 5:48: "Create ranked choice voting so third parties and independents can run and win." Unfortunately, this bold claim does not hold up to scrutiny. They are correct that our current voting rule, where you name one candidate only, is problematic, but ranked choice voting (RCV) suffers from many of the same problems.

For those who have never heard of RCV, the system goes as follows: When you vote, instead of naming one person, you list the candidates in order of preference. For example, you may vote:
  1. Alice Ideal
  2. Bob Good
  3. Carol Meh
  4. Dave Otherparty
  5. Evil N. Corrupt
(Usually, you are not forced to rank every candidate.) Initially, your vote only counts for Alice. The first-place votes are counted up, and the candidate with the fewest votes is eliminated. Anyone who voted for that candidate gets their vote transferred to their second choice (or the next choice that is still in). So if Alice dropped, your vote goes to Bob. (If Bob had already dropped, your vote goes directly to Carol, and so on.) This continues until one candidate has an absolute majority of the votes.

(The name "instant runoff voting" has also been used to refer to RCV, because RCV simulates a series of runoff rounds, while only requiring one physical election.)

This may seem like a pretty good system. Supporters claim that the use of RCV will:
  • Elects a true majority winner (as opposed to in our current system, where in a five-candidate race, Bob may win with 21% of the vote).
  • Prevent the "spoiler effect" where a minor party draws votes away from a frontrunner and flips the result.
  • Discourage "tactical voting", because voting your true favorite is supposedly always safe because your vote will transfer to your favorite frontrunner.
  • Allow third parties and independents to win.
Do not get me wrong. Those are all massive defects with the current system. My point in this blogpost is that RCV also falls into those four traps (although perhaps less often). There exist other reforms like Approval Voting and STAR Voting that do much better, but I am getting ahead of myself.

RCV Example

Before we begin: I am using small elections (3-4 candidates) to make it easier to see what is going on. However, these problems largely persist with more candidates. More evidence is given at the end of this section.

Suppose that a town has been dominated for decades by the Purple Party and the Orange Party. The Oranges generally win the mayor's office 55-45, but occasionally an independent shows up, steals votes from Orange, and Purple wins.

One day, the town decides to switch to Ranked Choice Voting. They send out papers showing people how to use the new system, and encourage third parties to run. The Yellow Party runs as a stronger version of the so-called "dull" Orange Party, and the votes come in...

45% vote Purple, 35% vote Orange, 20% vote Yellow.

Yellow is eliminated, but their voters all put Orange second, and Orange wins 55-45. Spoiler averted! So what is the problem?

Well, the Yellow Party is not interested in getting 20% every time. (That would not solve the problem of two-party domination.) They decide to push harder next election, and hopefully try to win.

40% vote Purple. (Their second choices are irrelevant.)
28% vote Orange:
    14% voted Yellow second.
    8% voted Purple second. *
    6% gave no second choice. **
32% vote Yellow:
    26% voted Orange second.

*These voters are swing voters who faced a close decision between the main rivals. Yellow was regarded as too extreme.
**These voters did not care too much about politics but liked the Orange Party and simply voted for that party. Both of these two categories of voters will exist in real elections, even though giving a second choice cannot hurt Orange.

OK, so there is a lot more going on here. (RCV is not simple to count once there are more than two competitive candidates.) But Orange drops out. The second round looks like:

Purple: 48% (40 first-ranks, and 8 from Orange transfers)
Yellow: 46% (32 first-ranks, and 14 from Orange transfers)

...Purple won? Wait a second. I thought RCV was supposed to eliminate spoilers! The truth is, RCV only partially solves the spoiler problem. It is true that RCV is immune to "perfect" vote-splitting (where two or more candidates split votes entirely amongst themselves). However, in this case, there were some moderate Orange voters who may have been deciding between Orange and Purple, but thought Yellow was too extreme and ranked Yellow last. In real elections, "perfect" vote splitting will almost never occur.

Let's look at that election in more detail:
  • RCV did not elect a majority winner for two reasons:
    • Purple won with 48% of the vote, because some Orange voters neglected to give a second ranking. (These voters probably only care about the Orange Party, and neglected to research the other two candidates. This will happen.
    • Orange would have defeated either Purple or Yellow in a head-to-head race. Orange only lost because of not having enough first-place votes.
  • Yellow was a "spoiler", because Yellow knocked Orange out of the race too early. Yellow was not strong enough to beat Purple, and lost the second round.
    • In particular, note that Yellow's votes never transferred to their second choice. Every single Yellow voter specifically wrote that Orange was the second choice. None of those second choices were counted; however, Orange voters' second choices were counted. (Seems unfair? That's the spirit.)
  • Yellow is never going to be able to defeat Purple. After this election, Yellow voters will tactically rank Orange first, for fear of another disaster.
  • In this case, the third party was thwarted, and voters will return to two party domination.
OK, so this tears RCV to shreds. Except perhaps you feel like the last point is not sufficient. Yellow was too extreme to win; perhaps a centrist who draws votes from both parties will fare better. In fact, that is what the Red Party decides. They mount a campaign that is between Purple and Orange. Then Crimson, an independent moderate, decides to join in as well. We have the following:

2500 vote Purple.
1999 vote Red.
2001 vote Crimson.
3500 vote Orange.

The apparent tie is broken after days of checking, recounting, and hand-counting again. (The problem of ties is intensified under RCV because there are more rounds and probably more candidates.) Red drops by two votes, and her votes transfer all over the place. The second round looks like this:

3010 Purple
2990 Crimson
4000 Orange

(Another near-tie? Murphy's law...) Unfortunately, Crimson still does not make the cut. This is called the "center squeeze" effect. Presumably, many Purple and Orange voters put the moderates second and third. But in general, moderates have a hard time getting first-place support by the very nature of being a moderate: they tend to gather many second ranks, but those mean nothing unless they get enough first-ranks to upset a major party.

RCV's selling points, like the stereotypical telemarketer, do not hold water. However, all is not lost. RCV is only one of many ways to fix our voting system, and there are three that look more promising.

If you think 3-4 candidates are not enough to get a full picture, I have some evidence here:
This takes some explanation. Each small circle is a candidate (14 total). Each pixel is an election in which the voters are centered on that point (with more voters closer to the pixel). Voters rank candidates in order of distance (closest first).


What actually does work?

There is one big flaw with ranking candidates: it lets you state your preference for Alice over Bob, but not how much you prefer Alice over Bob (compared to, say, Bob over Carol). However, a system involving ratings, say, from 0 to 10 allows you to vote something like:
  • Alice = 10
  • Bob = 8
  • Carol = 5
  • Dave = 1
  • Evil = 0
(If Bob and Dave were frontrunners, you may be tempted to vote them 10 and 0 instead. However, once the minor parties pick up steam, this will stop being necessary.)

In fact, we just came across one of the main proposals, known as "Score Voting". Score Voting just elects the candidate with the highest total score. This system treats every candidate independently, which means it truly eliminates spoiler problems. It also allows third parties to grow without affecting the major parties at all (until they win, of course).
For example, in the color election from earlier, voters may vote something like
  • Yellow = 10
  • Orange = 10
  • Red = 6
  • Crimson = 4
  • Purple = 0
Purple voters, in the meantime, may vote
  • Purple = 10
  • Crimson = 6
  • Red = 4
  • Orange = 0
  • Yellow = 0
(Some people have argued that people who give out more points have "more power" than those who do not. The first voter gave out 30 points and the second gave out 20. However, these two votes canceled each other out. This means that neither one can have "more power" than the other!)

And even if Orange ends up winning in the end, the minor parties are now a force to be reckoned with. The town newspaper will start featuring them more often, and eventually they will win.

(By the way: Score voting can be proved to eliminate spoilers, so it truly is safe to put your favorite first.)

But there are others...

One criticism of Score Voting is that it can degenerate into people voting only max and min and not using the middle of the scale. They advocate for a slight tweak to the Score rules to a new system called "STAR voting", which stands for Score Then Automatic Runoff.

With STAR, the two candidates with the most points are declared "finalists". The candidate scored higher on more ballots wins. (So if you score Alice 3 and Bob 2, and those are the top two, then your vote counts as a full Alice vote in the artificial runoff.)

Although I prefer plain Score due to simplicity, STAR should also break two-party domination. (Unlike RCV, STAR allows candidates to "grow in the shadows" like Score, and if many alternative candidates run, the chances are high that two will make the runoff.)

With some clever engineering, Score Voting can be counted on existing voting machines. (STAR may work, but it is trickier.)

Another method that has been proposed is Approval Voting, which is basically just Score but with only two levels (Approve and Disapprove). Approval is nice because the ballot looks like the current one, but it is less expressive than Score. (Approval can be used as a stepping stone to Score if a particular district has trouble switching.)

Score voting and STAR voting should also reduce the influence of money "naturally" (as opposed to laws specifically designed to do so), because of the wide array of candidates. Multiple candidates from the same party can run, and if one is evil and corrupt, the voters can give that one a 0 and their party still has a chance to win.

If you would like to research further, here are some links:
Equal Vote Coalition: www.equal.vote (promotes STAR voting)
Center for Election Science: www.electionscience.org (promotes Approval, supports Score but not as prominently)
The website www.scorevoting.net is a somewhat detailed and somewhat outdated website explaining the flaws with RCV (called IRV) and the benefits of Score Voting.

Summary

Ranked Choice Voting is a false cure. The problems it claims to solve are real and serious, but RCV is not the answer. Score Voting and STAR Voting are better ways of ending two-party domination and reducing corruption.






#2

Not only do they persist, they get worse. (e.g. ‘the four corners of a square’ Yee diagram scenario versus my profile picture).

I’m guessing you’re referring to cloning, but it took me 5 minutes to realize this, so it definitely needs to be reworded. I’m trying to think of a non-technical way to describe cloning, but I am struggling.

Maybe it’s just me, but it seems weird to think of the voters as two party dominated rather than the electoral system. Maybe say, “voters will refuse to back third parties.”

For some reason I don’t like this wording; maybe because you are declaring IRV torn to shreds before your argument is finished. I don’t like the second sentence because I think comparing the example you just described to the one you are introducing (as you did in the third sentence, which could possibly be reworded to replace the second) works better as a transition than addressing the reader.
Also, replace ‘that is’ with “in a race where the major candidates are”, replace “join in as well” with “run”, and replace “We have the following:” with “The count after the first round is:”

I don’t see an actual tie, so “Apparent tie” is a little confusing. Maybe say “This close finish between the bottom two candidates requires a hand-counted recount.” for the first sentence. Also, replace “drops by” with “loses by”, and strike “all over the place”. For the following table, perhaps indicate how many transfers each candidate received, e.g. 3010 Purple (+510).

Strike “still”, also, you might want to explain why Crimson being eliminated is unfortunate before saying that it is so. Also, I’ve always explained noted something along the lines of ‘moderates face vote splitting from both sides whereas extremists only face vote splitting from one’ when writing about center squeeze.

This telemarketer seems to come out of left field.

I can’t figure out whether this is a good introductory summary of what a Yee Diagram. I understand what you’re getting at, but then, I already know what a Yee Picture is. I’d ask a friend or someone who doesn’t post on CES to look at it and see if they can figure out what you are getting at.

Strike the "somewhat"s. I’m guessing you mean the web design is outdated moreso than the content, and it might be worth specifying that somehow. Maybe “is a detailed (albeit minimally designed) website…”

Also, Australia is a good empirical example of RCV being 2-party dominated. I would recommend citing it.

#3

I’d change the language to specifically attack the IRV method of counting. The re-brand to “Ranked Choice Voting” is potentially a good thing because it allows switching out the back end to a Condorcet method. “Ranked Choice Voting” only describes a ballot, and that much is fine. I prefer score ballots too, they’re more expressive, but RCV is fine.

1 Like
#4

See Simple Ranked Voting.

It’s not as good as score by quite a bit, but would work and could be a stepping stone to score.

With SRV, voters should be allowed to leave rank-places empty if they so prefer.

#5

Also known as a Borda count. Like IRV, simple to explain but not as good as other ranked methods.

#6

It is far too easy to overlook small distinctions in voting systems, probably because so many systems have been proposed. Few people can consistently keep track of all the variations.

Seemingly harmless modifications of election systems often result in major pathological differences in their results. These are called dark patterns. (“IRV”, for example, is such a dark pattern.)

At first glance, simple ranked voting looks like Borda method, but it turns out to be very different. See:

#7

@Marylander: Thanks.

Yes. Let me know if you come up with something.

I had that election as showing “20%”, but then I decided to drop the percents. Now it looks more artificial, but you cannot make something on a 10,000 voter scale easy to follow and non-artificial. (What’s 3000 + 150? OK, now what is 3387 + 134?)

Yeah, good observation. I thought that part was harder to argue.

Perhaps I can drop the image entirely and just link to the rangevoting.org article.

Actually, I do mean the content. Examples:

  • The Australia 2PD example uses the 2001, 2004, and 2007 house races (0/450 third party wins). Those are now over 12 years old.
  • The website suggests an “Iowa '08 caucus” plan. That is 12+ years out of date. http://www.scorevoting.net/IowaExecSumm.html
  • Many of the IRV debunking centers on old stuff. A search of “201” in Irvtalk shows one hit for “2011” and one for the number 201 appearing in an example. Meanwhile, searching “200” hits 23 year numbers.

The format does also add to the “by PhDs for PhDs” feel of the website.

…has already been done, and I get the sense that using “IRV” will only confuse people.

However, the middle word “choice” seems to actually apply more to IRV (absolute order) than Condorcet-type (relative order) methods. With IRV, your vote counts as your first choice. When they are eliminated, the next surviving choice on your list gets your vote. It is more important for A to be first than to be preferred over B (that is, a vote of C>A>B has no effect on a crucial A vs B near tie for last).
Condorcet methods are more about preference order (it matters more that A>B than that A is first).

In rkjoyce’s version, there are exactly 10 ranks and you can leave blanks in the middle. So you could vote something like A, skip three slots, B, C, skip two, D, E. A gets 10, B gets 6, C gets 5, D gets 2, and E gets 1.

Basically, this is Score but with a more confusing ballot (since “1” gives the most points and “10” the least), and you cannot give two candidates the same number of points.

(Also, in the quote, it says that SRV and Borda are “only the same if there are exactly 10 candidates” – but in Borda you are not allowed to skip ranks; in SRV you can.)

#8

I decided to split my article into two: one attacking IRV, and one promoting cardinal methods.
Can I get a link to someone in EVC claiming that Score is liable to one-sided strategy problems?

So you think our current choose-one method is broken.

Studies show the US is not a democracy. Average citizens' support for a bill has near-zero impact on its chance of passing. People joke that politicians never get anything done, lie all the time, and so on, and those jokes are often reality. Billionaires buy everyone off. And it is not getting any better.



You know how you can rate restaurants on websites like Yelp? Have you ever noticed that Yelp does not force you to vote for one favorite, and then say absolutely nothing about all the other places out there? Instead, you can rate each one individually from 1 to 5. You can express more complex opinions, like "I like A a little better than B, but C was terrible". The average rating is displayed for all to see: would you go to a place averaging 1.2 out of 5?

What if we applied that principle to the ballot box? In fact, let's extend the scale to 0 to 10 (starting with 0 so you are not forced to give any "support" to a candidate). You might vote something like this:
  • Alice = 10
  • Bob = 8
  • Carol = 5
  • Dave = 1
  • Evil = 0
(If Bob and Dave were frontrunners, you may be tempted to vote them 10 and 0 instead. However, once the minor parties pick up steam, this will stop being necessary.)

This system, known as "Score Voting", may seem too simple to be any good, but it has some really nice properties. Score Voting treats every candidate independently, which means it truly eliminates spoiler problems. (You give your true favorite a 10, and the frontrunner anything from 5 to 10 depending on how much you like them.) It also allows third parties to grow without affecting the major parties at all (until they win, of course). That is, voters can give scores to Greens and Libertarians, and

As an example, a conservative voter in 2016 (if the election were done with Score Voting) may have voted something like this:

10 - Trump
8 - Rubio
5 - Johnson
1 - Stein
0 - Bernie
0 - Hillary

Similarly, a liberal may have voted

10 - Hillary
10 - Bernie
9 - Stein
5 - Johnson
2 - Rubio
0 - Trump

(You may have noticed these votes are exact opposites. Notice the conservative awarded a total of 24 points, while the liberal gave out 36. This debunks the myth that Score Voting gives some voters "more power" than others, because even though the liberal awarded more points, the two votes canceled each other out.)

(If my other article made you skeptical of claims to preventing spoilers, do not worry. With Score Voting, it is always safe to give top score to your favorite candidate. Proof: Either this move elects your favorite, or else your favorite had no chance, but then their score is irrelevant.) Also, with some clever engineering, Score Voting can be counted on existing voting machines.

The best thing about Score is that third parties have a path to victory. Initially, only the two major parties will win, but voters will know that they can safely give third parties non-zero scores. Over time, those scores will increase as awareness rises. Eventually, the (best) third parties will be able to compete with major parties, and the media will have to give them coverage. Finally, one-party majorities in Congress will be a thing of the past.

If that sounds like a bad ending, remember this: Multiple candidates of the same party (or a similar ideology) can run as allies. This means that if some of them are corrupt evil liars, you (and everyone else) can give them zeroes without dooming your favorite party from winning. As long as such people run, politicians will stop being selfish greedy moneyed slackers, and start working together across party lines to get things done.

But there are others who say...

Some criticize Score Voting as being susceptible to strategic voting. Although you can always score your favorite maximum safely, there may be more nuanced strategies that voters may exploit. In particular, the "Equal Vote Coalition" (www.equal.vote) argues that Score Voting yields unrepresentative outcomes if tactical voting is tied to ideology (e.g. if richer people are more likely both to support conservative candidates and also to vote tactically). They propose a slight modification to Score to balance things out. It is called Score Then Automatic Runoff (STAR).

With STAR, the two candidates with the most points are declared "finalists". The finalist scored higher on more ballots wins. (So if you score Alice 3 and Bob 2, and those are the top two, then your vote counts as a full Alice vote in the artificial runoff.) Also, most STAR implementations use 0-5 (because of 5-star ratings), not 0-10 as I described for Score.

STAR voting is a little harder to count than Score, though it can be done with a "precinct subtotal". (Each precinct publishes the total score and the winning margin of every possible runoff. The scores are added up and the appropriate pair margin in each table is used to determine the runoff winner.)

Score voting and STAR voting should also reduce the influence of money "naturally" (as opposed to laws specifically designed to do so), because of the wide array of candidates. Multiple candidates from the same party can run, and if one is evil and corrupt, the voters can give that one a 0 and their party still has a chance to win. (This last property is partially true of RCV, although with RCV it is only guaranteed if all the same-party candidates have identical policy positions.)

Although I prefer plain Score due to simplicity, STAR should also break two-party domination. (STAR allows candidates to "grow in the shadows" just like Score, and if many alternative candidates run, the chances are high that two will make the runoff.)

If you would like to research further, here are some links:
Equal Vote Coalition: www.equal.vote (promotes STAR voting)
Center for Election Science: www.electionscience.org (promotes Approval, supports Score but not as prominently)
The website www.scorevoting.net is a detailed website explaining the benefits of Score Voting and the pitfalls of another common. Some of its content is a bit outdated, though.
#9

This part should probably go into an FAQ.

But the voters may not.

Maybe start with the intuition that voting for more than one gives more power than vote for one (which it does, if the voter likes more than one candidate)

Your favorite isn’t the only one who could spoil the election, either from your perspective or society’s.

Can you really guarantee that? I can see much more variation in each member of Congress, but you seem to be saying more that it won’t be extremist, polarizing majorities in Congress anymore.

I’m not sure who the article is addressed to, but it seems that you’re assuming a lot of criticisms that the reader probably hasn’t even thought of.

Take the “I” out, if you can. It takes away from the neutrality of the message.

That’s an understatement :slight_smile: but WDS got a lot right too.