Preferential Voting, Mail Ballot, Three Board Seats

[Guest Joseph Mott]

Our organization is going to conduct a mail-in, absentee vote for the first time in our board elections. Our board consists of nine people, and three are elected each year for a three-year term.

RRONR says that election by MAJORITY rather than PLURALITY is the better method (p. 404-405) and recommends preferential voting.

The question comes up around counting the votes. While disparaging a plurality vote, RRONR appears to suggest just that on page 428 when it describes counting exactly the sort of ballots our election will generate. As an example, if we have four candidates running for three seats, RRONR seems to be saying that we simply go through the process ONCE ("until all but the necessary number of candidates have been eliminated -- that is, in the example [of three members of a board to be chosen], all but three" and just eliminate the single candidate with the fewest votes. Then the other three win -- and at least two of them will win WITH A PLURALITY.

It seems much fairer to me to elect one board seat at a time, and go through this process WITH THE SAME BALLOTS in three separate, consecutive rounds, in each round simply first eliminating the name of the candidate who won a seat in the previous round. In the last round, it will be a simple race between the remaining two candidates. This way, EACH elected candidate will have received a majority.

That seems to me to be the correct way to use these ballots to replicate the majoritarian process preferred by RRONR in every other context, including in-person voting.

Does anyone have any thoughts?

Joseph Mott

[jstackpo]

First thought: Do your bylaws authorize absentee -- i.e., mail-in -- voting? If not then forget the rest.

Preferential voting: Some systems of counting do indeed allow for a plurality (non-majority) winner. But other systems require majority votes to win. It depends on the rules for your preferential system. Have you, your association, adopted any such rules? The sample rules in RONR (which you, your association, have to formally adopt to use) do require a majority to win.

[JMott]

We are revising the bylaws to authorize the absentee vote, and we are deciding on the system to use.

The issue is that RRONR seems to describe a process for preferential voting in the case of three seats, which MUST result in a plurality election. The alternative would be to run the process three separate times, each time coming up with a majority winner, instead of electing three at once. That seems fairer to me, but it's NOT what RRONR seems to say.

[jstackpo]

If after the first elimination of the low-man (woman), and you have three piles of votes, if one now has a majority of the ballots cast, i.e. more than half the total number of pieces of paper in the three piles, you are 1/3 done. If none of the three have majorities, distribute the third (lowest) pile to the other two. RONR doesn't (quite) say this but other systems I have seen (e.g., Amherst College Trustees) do. Remember you are setting out the rules so state them clearly. Now you have two piles - one is (all but) sure to be a majority.

Now pick up ALL the ballots and distribute them to the remaining three candidates, ignoring, of course, votes for the person who was just declared a winner. And so on.

Rather than all this repeated paper shuffling, may I recommend (RONR doesn't, but should) the Borda Count as the best (or least bad) system of counting and scoring ranked lists on the ballots. It has a completely unambiguous counting method and you only have to do it once to get (in your case) three ranked winners out of four, or more for that matter. Google "Borda Count" and you will get all the directions.

The Borda count has its quirks, but so does any system of selecting from more than two candidates.

[g40]

The issue is that RRONR seems to describe a process for preferential voting in the case of three seats, which MUST result in a plurality election. The alternative would be to run the process three separate times, each time coming up with a majority winner, instead of electing three at once. That seems fairer to me, but it's NOT what RRONR seems to say.

Maybe I am missing something, but if you had one vote, with each voter selecting up to 3 candidates, then three candidates could get a majority. Depending on how many select 3 and how many candidates there are, it may be very likely that 3 get a majority.

Example - if there are 5 candidates (A,B,C,D,E)and ten voters - the votes might come out as follows:

A: 6

B: 1

C: 6

D:10

E:3

X: 1 (write-in)

27 total votes (several of the 10 voters did not vote for 3)

A, C and D get a majority (more than half of 10) and are elected.

[Guest Edgar]

. . . it may be very likely that 3 get a majority.

Or they may not.

I was scratching my head too but I believe the problem here is that, since this is a mail-in vote, subsequent rounds of voting are not an option.

[Bruce Lages]

We are revising the bylaws to authorize the absentee vote, and we are deciding on the system to use.

It's not clear from this statement whether you plan to also incorporate whatever plurality system you choose into the bylaws as well, but be aware that RONR makes that a requirement if election of officers is involved (p. 405, ll.2-6). And if you're abiding by RONR, board members who are titled as 'directors', or 'trustees', etc., (rather than the named officers of president, vice-president, secretary, etc), are also considered to be officers (p.572, l. 23).

[Guest Edgar]

And if you're abiding by RONR, board members who are titled as 'directors', or 'trustees', etc. . . . are also considered to be officers (p.572, l. 23).

What RONR actually says is that they should be classed as officers, not that they necessarily are. Which is why the sample bylaws make this an explicit provision.

[JMott]

I reread the section on "Elections: Elections by Ballot" and I think I understand now. If there were ten ballots, each ballot counts as "one vote" regardless of how many names it votes for. If each ballot indicated three choices (for three seats, among four candidates) there would still only be ten "votes" and SIX would constitute "a majority." So it's quite likely that one (or more! -- even all four) candidate(s) could receive "a majority" of at least six votes. For instance, three candidates could receive six votes each, and with the other getting 12.

So in that sense, even winning with six votes isn't a "plurality" win because that name appeared on the majority of ballots.

Somehow it still seems fundamentally fairer to conduct the elections one seat at a time, though. This is still nagging at me.

[JMott]

I think I've put my finger on it. The preferential voting system counting procedure described in RRONR 426-428 says to continue the procedure only until "all but the necessary number of candidates have [sic] been eliminated (that is, in the example, all but three)." If in a 4-candidate race, with twenty ballots cast, nobody has a majority (11 votes out of 20) in the first round, the lowest-ranked candidate is dropped and the other three win -- because there are now three candidates left for three seats.

First Round:

A: 8

B: 7

C: 3

D: 2 (eliminated, ballot redistributed to second-ranked candidate).

Leaving:

A: 10

B: 7

C: 3

But even giving those two votes to the biggest vote-getter in the first round (A) doesn't leave A (let alone B or C) with a "majority" in any sense when the election is completed according to the procedure described in RRONR. The only way to achieve a "majority" in scenarios like this would be to continue the process to elect one seat at a time (in which case, one candidate WOULD eventually have to get a "majority" or at least tie, in which a tie-breaker would be used which would be a proxy mechanism for a majority).

And in the next election with A out of the race, redistributing the eight ballots in which A is ranked first might really shift the balance between the remaining three candidates. What if D had been the second choice on each of those ballots, for example? D would then start out with 10 votes in the second election and have quite a good shot at winning a seat!

It just seems that truncating the counting procedure at the very first moment where there are only three candidates left (after counting only the first-ranked choices) defeats the purpose. Even with first-to-the-post voting, all the votes on each ballot would at least be counted.

I'm sure this looks different in other scenarios (more candidates, for example, and many more ballots) but the most realistic scenario for us in the immediate future is to have AT MOST four or five candidates vying for three seats (with probably no more than 100 ballots cast).

[jstackpo]

Try the Borda Count - it will alleviate all your anxieties.

But for now, you are right that it may well require three rounds of counting all the ballots to generate three winners who get a majority (eventually, after ballot redistribution). Seems reasonable to me.

Stopping with plurality winners is a far worse system

[Shmuel Gerber]

If after the first elimination of the low-man (woman), and you have three piles of votes, if one now has a majority of the ballots cast, i.e. more than half the total number of pieces of paper in the three piles, you are 1/3 done. If none of the three have majorities, distribute the third (lowest) pile to the other two. RONR doesn't (quite) say this but other systems I have seen (e.g., Amherst College Trustees) do. Remember you are setting out the rules so state them clearly. Now you have two piles - one is (all but) sure to be a majority.

Now pick up ALL the ballots and distribute them to the remaining three candidates, ignoring, of course, votes for the person who was just declared a winner. And so on.

Dr. Stackpole,

How do you interpret the statement in RONR (p. 428, ll. 1-4) that the OP is concerned about, "The counting procedure is the same as described above, except that it is continued until all but the necessary number of candidates have been eliminated (that is, in the example, all but three)"?

Does it require the continued counting as you've described, or are A, B, and C declared elected in the second step of the example, with no further counting, even though none has achieved a majority?

I know, you say RONR doesn't (quite) say. But suppose you're the parliamentarian in this scenario, and the rules adopted are exactly as stated in RONR. What do you advise the chairman of the tellers' committee?

[jstackpo]

My "advice" to the teller's chairman would be to throw up his hands, throw in the towel, punt, or whatever version of "I quit!" you prefer.

The rules for counting in the RONR "illustration" (p. 436, line 13) as applied to multiple position elections (like boards) on p. 428 are just plain inadequate. And after the chair has thrown up his hands, he should refer to question to the assembly, via the presiding officer, for the final interpretation of how to count the ballots.

I agree with S. Gerber (I think, but I could equally well agree that the "continuing" of the procedure means working down to one majority winner [or more in the same counting pass] and then starting over for the next candidate) that as written on p. 428, once D is eliminated as the low-vote candidate A, B, & C are declared winners. This, however, results in a (possibly) gross violation of the "majority needed for election" rule -- JMott's example back in posting #10 shows this. Hence my waffle in "agreement".

RONR does say (ibid., line 8) that the counting procedure "should be prescribed in detail". RONR has not done so for us in this "illustration" (nor should it be required to - there exist many whole books on voting theory - things are not settled yet - if they ever will be - so putting one set of rules in RONR in not really appropriate).

As an example of unsettled variations, the tie breaking mechanism in RONR (tossing out both tied-for-lowest candidates at once) differs from the Amherst College Trustee election method (too long to describe). The interesting thing is that the differences in tie breaking mechanisms can lead to different candidates winning, even with the same set of ballots. When (arbitrary) rules of vote tabulation can determine the ultimate outcome of the election, that makes one very suspicious of the voting system in the first place.

Go Borda!

[Shmuel Gerber]

I agree with S. Gerber (I think, but I could equally well agree that the "continuing" of the procedure means working down to one majority winner [or more in the same counting pass] and then starting over for the next candidate) that as written on p. 428, once D is eliminated as the low-vote candidate A, B, & C are declared winners.

I didn't say that "as written on p. 428, once D is eliminated as the low-vote candidate A, B, & C are declared winners."

In fact, I'm inclined to think that when RONR says, "The counting procedure is the same as described above, except that it is continued until all but the necessary number of candidates have been eliminated (that is, in the example, all but three)", this does not mean that anyone is declared elected without having achieved a majority, but rather the procedure is actually as you described in #4. The "counting procedure ... described above" requires a majority vote for election.

[jstackpo]

Ok, sorry S.G., I read your reply too rapidly. Evidently Mott's co-members on his bylaws revision committee seemed to think that once D was out, the other three were in. This, I presume, made Mott unhappy (as well it should) and he raised the question here. Good, and welcome!

But I still don't think much of the preferential system. Nor of RONR's skimpy write-up of it.

If you want more on the inadequacies of the preferential system, just let me know.

[Guest Jim]

If this posting is still open to thoughts, I think the problem of plurality voting is clearest when you have multiple candidates for a single position.

In the case of a society to which I belong, four people ran for president, and each gained between 21 and 27% of the votes. No-one is certain that individual who was awarded the Presidency would have ranked highly (i.e. in 2nd or 3rd place) on the ballots cast by members for someone else, had the society's ballots provided for ranked voting.

I think it is harder to work through when one is faced with filling multiple positions. I do accept, in the example above, that the simple fact that "C" got 3 equivalent-to-first-place votes, and "D" got only 2, need not mean that the membership would prefer for the third seat to be filled by "C" instead of "D". I can also envision a situation where whoever had secured the first two positions could make a difference to who the membership may desire in the third position, for example for reasons of balance or diversity.

I also suspect that if the organization is conducting a mail-in ballot, it is because the membership is geographically dispersed. Seldom, in my experience, does what a candidate writes on a campaign statement provide a good measure of the person's actual capacities to be effective in the position for which they are offering themselves. This kind of challenge can be as big a challenge as the system of voting.

[JMott]

Thank you for your input, everyone.

I think I'd like to recommend to my board that it consider using serial instant run-off voting to fill these three seats. My only hesitation is that -- despite the plethora of descriptions on various websites -- I have been unable to find ANYBODY advocating that particular approach. The authorities I've looked at either advocate the multi-seat IRV method that truncates just as soon as the number of candidates is reduced to the number of seats (potentially leaving one or more plurality candidates winning), or they advocate the much more complicated STV approaches that use fractional re-allocation of votes (including redistributing the "excess" votes of winning candidates).

This makes me suspicious that I might be overlooking something obvious. Could there be a REASON that running a sequential series of three elections on the same set of ballots (which rank ALL the candidates) would not be fair?

I don't think it's because there's a violation of the "one person, one vote" principle, because in an in-person series of elections, everyone in the room would be eligible to participate and vote on the second candidate, regardless of whether the candidate they'd voted for in the first election won, or not.

[Shmuel Gerber]

This makes me suspicious that I might be overlooking something obvious. Could there be a REASON that running a sequential series of three elections on the same set of ballots (which rank ALL the candidates) would not be fair?

You seem to be overlooking the Amherst College method, as described in Reply #4. Why can't you take "yes" for an answer?

[JMott]

Well, that's what I've been looking for, I guess! I would love to take "yes" for the answer!

I can't seem to find a reference detailing their system, though. I wonder if you might have it? It would certainly make my job easier if I could refer to the rules of a group already using just such a system successfully.

Thank you!

[jstackpo]

I suspect if you wrote to the Amherst Alumni Association, Amherst MA 01002-5000, they would be happy to send you a copy. (And not even ask for a contribution!)

Or send me your postal-mail address (or even a fax number) -- use the internal mail system available here -- and I can ship you a copy. (I don't have an electronic copy - sorry.)

But... Borda is (still) better.

[jstackpo]

Bingo! Go here.

[Gary Novosielski]

This makes me suspicious that I might be overlooking something obvious. Could there be a REASON that running a sequential series of three elections on the same set of ballots (which rank ALL the candidates) would not be fair?

The method listed in RONR does assume that ALL the candidates are ranked (or can be) on every ballot. I can't tell how you're interpreting the rules, but as I read them, if voters rank all the candidates, a plurality winner is not possible.

I don't think it's because there's a violation of the "one person, one vote" principle, because in an in-person series of elections, everyone in the room would be eligible to participate and vote on the second candidate, regardless of whether the candidate they'd voted for in the first election won, or not.

If each person ranks all the candidates, then he will ultimately cast one vote for one candidate, although as he casts his ballot he cannot be certain for whom his one vote will ultimately count.

[JMott]

Thank you, everyone. Particular thanks to JDStackpole for that link to the Amherst rules!! That is exactly what I was looking for. Now I don't feel as if I'm going out on a limb at all in recommending that voting procedure.