Dear friends,
Last week I asked you all to offer vote in a fun election to
pick your favorite months. This was done as an approval-style election where
you were allowed to vote for as many choices as you like. This is
an example for me how approval elections might work in practice, and an
educational example to you all to show other election methods.
For more information on Approval, see:
The best thing about approval voting is that it can
be done in political elections without any need for updating voting equipment or
ballot design.
More advanced voting methods like "Instant Runoff Voting" use
rankings (first, second, etc) which allow voters to specify the other of
their preference more accurately. [See http://www.fairvote.org/irv/ for more
information and background on Instant Runoff Voting.]
Without runoffs, our elections with 3 or more choices
often have a "spoiler candidate" where if people vote for a 3rd
party favorite, this may help their least favorite win. Therefore voters
with an acceptable compromise candidate are encouraged (through fear)
to vote for their compromise (who is among the expected top two
candidates) to have a positive affect on the election results.
Approval allow voters to vote BOTH for a most favorite
AND a compromise, if they are worried their favorite won't win. This can be very
important for 3rd parties who need a certain fraction of the vote in order to
qualify for governmental funding, and it helps determine a candidate with
majority preference in a single election.
Here is the results from the fun "Favorite Months" Approval
election: There were 58 voters.
Approval:
Month Votes Percent-Approval
Sep 29 ( 50.00%)
May 28 ( 48.28%)
Jul 24 ( 41.38%)
Jun 21 ( 36.21%)
Aug 21 ( 36.21%)
Oct 21 ( 36.21%)
Apr 18 ( 31.03%)
Nov 14 ( 24.14%)
Dec 14 ( 24.14%)
Feb 11 ( 18.97%)
Mar 10 ( 17.24%)
Jan 10 ( 17.24%)
May 28 ( 48.28%)
Jul 24 ( 41.38%)
Jun 21 ( 36.21%)
Aug 21 ( 36.21%)
Oct 21 ( 36.21%)
Apr 18 ( 31.03%)
Nov 14 ( 24.14%)
Dec 14 ( 24.14%)
Feb 11 ( 18.97%)
Mar 10 ( 17.24%)
Jan 10 ( 17.24%)
The total approval votes cast were 221, or average 3.8
votes/voter. We can also look at the median
votes/voter cast. Below is a table of the frequency of votes per voters.
The median comes out to be between 2 and 3 votes. (45% votes for 2 or less, and
55% voted for 3 or more.) This statistic suggests that
there could be an runoff election among the top two or
three candidates to allow people to refine their choices.
58 voters, 221 votes (Average 3.8
votes/voter)
Votes/ballot distribution:
Votes Voters
0 1 (No months liked!)
1 7 (Bullet voted for favorite)
2 18 (Two favorites)
Votes/ballot distribution:
Votes Voters
0 1 (No months liked!)
1 7 (Bullet voted for favorite)
2 18 (Two favorites)
[Median between
2-3]
3 12
4 4
5 1
6 3 (Half of choices are liked)
7 5
8 2
9 1
10 2
11 1 (A vote against one disliked choice)
12 1 (All months liked!)
3 12
4 4
5 1
6 3 (Half of choices are liked)
7 5
8 2
9 1
10 2
11 1 (A vote against one disliked choice)
12 1 (All months liked!)
Although Approval is traditionally considered w/o runoffs just
like plurality, Approval elections, like plurality (single vote) elections,
can use runoff elections between top choices. The advantage of an approval vote
runoff is that people can specify all their favorites in early rounds and so it
should take less runoff rounds to pick a single winner. This is important
if there are many choices.
Now with this Month election, we can look at all the ballots and only consider the top 3 months and
how they were supported. With 3 choices, there are 8
possible ballot-types: 1 type with no choices, 3 types
for one choice, 3 types for 2 choices, and 1 type for all choices. Here are the
results of each group:
15 for none of these choices:
----------------------------
15 25.9% None of
them
15 for one choice:
------------------
8 13.8% July only (Summer)
4 6.9% September only
3 5.2% May only
8 13.8% July only (Summer)
4 6.9% September only
3 5.2% May only
18 For two choices:
-------------------
12 20.7% May AND September (Spring and Fall)
3 5.2% May AND July
3 5.2% July AND September
12 20.7% May AND September (Spring and Fall)
3 5.2% May AND July
3 5.2% July AND September
10 for all of these choices:
----------------------------
10 17.2% All of themThis shows that 74.1% of voters support at least one of these
three months. In other words 3/4 of voters would be interested in at least
one of these candidates winning in a runoff election. This
also shows that there is an affinity between Spring and Fall supporters (largest
two-choice coalition with 21%), while Summer supporters represent the largest
single-choice coalition (14%).
We can also consider having a top-two runoff. With 2 choices
(Sept and May), there are 4 possible ballots:
23 for neither choice:
----------------------
23 39.7% None of
them
15 for one choice:
------------------
6 10.3%
May
7 12.1%
September
12 for both choices:
--------------------
22 37.9% May and September
This shows that 60% of voters like either May or
September, and 38% like both! In fact, September won this election over May by a
single vote (1.7%) and only 22% of voters have expressed a preference between
them! A runoff election could clearly go either way depending on the
opinion of 40% who preferred neither, and 38% who preferred both.
No election with simple voting can nicely
choose between 12 candidates, but approval voting gives voters more power
than single vote elections.
Thank you all for helping me in this election. It
is both fun and interesting and I learned a lot from it.
Here are some fun and interesting voter comments
on this month election: (Anonymous for voter privacy)
Comment #1:
After listening to all the
candidates, I've decided to cast my vote for
"none of the above". I'm sick of all the mudslinging by April and May,
and I really hate all the reruns of July and August. But most of all, I'm
sick of the stupid Christmas ads of December. January and February are so
dreary, and March is a bore. I would have voted for November (my
birthday), but November tries to make me fat with a thing called
Thanksgiving. I can't vote for September because it reminds me of the
start of school. June brings mosquitoes, and October brings winter.
I guess I'll hope for better candidates next time and just complain about
the weather until then...
"none of the above". I'm sick of all the mudslinging by April and May,
and I really hate all the reruns of July and August. But most of all, I'm
sick of the stupid Christmas ads of December. January and February are so
dreary, and March is a bore. I would have voted for November (my
birthday), but November tries to make me fat with a thing called
Thanksgiving. I can't vote for September because it reminds me of the
start of school. June brings mosquitoes, and October brings winter.
I guess I'll hope for better candidates next time and just complain about
the weather until then...
Comment #2:
Every corporation in the U.S. uses approval (cumulative) elections for
their limited number of seats on their Boards of Directors. However, it
is done the same way Russia elects officials: you "withhold" your vote
for the candidates you don't like.
So that is how I am casting my ballot. I am withholding my vote for [...]
Every corporation in the U.S. uses approval (cumulative) elections for
their limited number of seats on their Boards of Directors. However, it
is done the same way Russia elects officials: you "withhold" your vote
for the candidates you don't like.
So that is how I am casting my ballot. I am withholding my vote for [...]
*** [Tom says: Cumulative voting is somewhat different
since voters have a fixed number of votes for a multiseat election. Cumulative
voting supports Proportional Representation, while Approval voting only supports
majority rule.]
Comment #3:
Your election concept is somewhat
similar to the way that Israel votes - but
there each party has an entire list of people interested in being in the
legislature and depending on what percentage votes for that party, a certain
number of names are taken off of that list for the jobs. Your idea is
basically what happens now in multiple candidate city council races - where
several people can be voted for and the top "yes" votes win.
there each party has an entire list of people interested in being in the
legislature and depending on what percentage votes for that party, a certain
number of names are taken off of that list for the jobs. Your idea is
basically what happens now in multiple candidate city council races - where
several people can be voted for and the top "yes" votes win.
****************************
Extra credit report:
Instant Runoff with Approval
ballots!
****************************
If you're not all bored or overwhelmed yet (I can
have some hope for this possibility), I offer this second report on this
month election. Some people I've talked to don't like approval voting because
they claim that it gives some voters more power than others. This is nonsense so
long as we are interested in MAJORITY RULE because each voter gets exactly
up to one vote per candidate, but approval is different from what we're
used to.
If we like our traditional single vote
approach, it is possible to consider approval voting, but we can do it by
dividing a person's vote among the choices. Thus if I vote for 2
choices, each gets a 1/2 vote from me. This may seem unattractive, but it
can be useful by considering an "instant runoff" process. When one of my
choices is eliminated, my remaining choices get a higher fraction of my vote.
Thus I can still get a full vote on one of the top two choices if
I want.
Instant runoff voting is an alternative system
which uses ranked ballots. The highest ranked candidate on each ballot gets one
vote for that ballot. After the top ranked candidates are counted, if there are
more than 2 choices and if none have a majority, then the one with the least
votes can be eliminated. This frees the next lower ranked
candidate to be given the vote.
An approval election can be done with
approval-style voting instead of ranks. I decided
to apply this process to our month election to see what happens.
Here's the first round of that process, with votes divided.
You can notice that the top three haven't changes in order, but there is some
change in lower rankings as votes who supported more candidates have
less influence here. (In this process there is a definite advantage to
voting for a single favorite if you have one that can win.)
Total votes=58
Month votes percent(rounded to 0.1%)
Sep 8.8 ( 15.18%)
May 8.6 ( 14.90%)
Jul 6.5 ( 11.27%)
Aug 5.6 ( 9.62%)
Oct 5.6 ( 9.62%)
Jun 4.7 ( 8.14%)
Apr 4.2 ( 7.18%)
Dec 2.9 ( 5.05%)
Feb 2.9 ( 4.94%)
Nov 2.7 ( 4.61%)
Mar 2.3 ( 3.94%)
Jan 2.2 ( 3.79%)
None-of-the-above=1.0
Sep 8.8 ( 15.18%)
May 8.6 ( 14.90%)
Jul 6.5 ( 11.27%)
Aug 5.6 ( 9.62%)
Oct 5.6 ( 9.62%)
Jun 4.7 ( 8.14%)
Apr 4.2 ( 7.18%)
Dec 2.9 ( 5.05%)
Feb 2.9 ( 4.94%)
Nov 2.7 ( 4.61%)
Mar 2.3 ( 3.94%)
Jan 2.2 ( 3.79%)
None-of-the-above=1.0
Now, here's the instant runoff results, each round eliminating
the month with the least votes.
Round # 1 Sep=8.8 May=8.6 Jul=6.5 Aug=5.6 Oct=5.6
Jun=4.7 Apr=4.2 Dec=2.9 Feb=2.9 Nov=2.7 Mar=2.3 Jan=2.2 [NOTA=1.0]
Eliminate Jan
Round # 2 Sep=8.9 May=8.8 Jul=6.7 Oct=6.2 Aug=5.6 Jun=5.0 Apr=4.2 Dec=3.2 Feb=3.0 Nov=3.0 Mar=2.4 [NOTA=1.0]
Eliminate Mar
Round # 3 Sep=9.0 May=8.8 Jul=6.9 Oct=6.4 Aug=5.7 Jun=5.1 Apr=4.4 Dec=3.3 Feb=3.3 Nov=3.2 [NOTA=2.0]
Eliminate Nov
Round # 4 Sep=9.4 May=9.0 Jul=7.2 Oct=6.7 Aug=5.8 Jun=5.2 Apr=5.1 Dec=4.0 Feb=3.6 [NOTA=2.0]
Eliminate Feb
Round # 5 Sep=9.6 May=9.0 Aug=7.5 Jul=7.4 Oct=7.3 Apr=5.7 Jun=5.2 Dec=4.3 [NOTA=2.0]
Eliminate Dec
Round # 6 Sep=9.9 May=9.2 Oct=8.1 Aug=7.8 Jul=7.6 Apr=5.8 Jun=5.5 [NOTA=4.0]
Eliminate Jun
Round # 7 Sep=11.5 May=9.8 Jul=9.4 Oct=8.5 Aug=8.4 Apr=6.4 [NOTA=4.0]
Eliminate Apr
Round # 8 Sep=12.2 Jul=11.1 May=10.6 Oct=9.3 Aug=8.8 [NOTA=6.0]
Eliminate Aug
Round # 9 Sep=13.1 Jul=12.9 May=11.6 Oct=10.4 [NOTA=10.0]
Eliminate Oct
Round #10 Sep=14.8 Jul=14.3 May=13.8 [NOTA=15.0]
Eliminate May
Round #11 Sep=22.5 Jul=17.5 [NOTA=18.0]
Eliminate Jul
Round #12 Sep=29.0 [NOTA=29.0]
Winner: Sep
Eliminate Jan
Round # 2 Sep=8.9 May=8.8 Jul=6.7 Oct=6.2 Aug=5.6 Jun=5.0 Apr=4.2 Dec=3.2 Feb=3.0 Nov=3.0 Mar=2.4 [NOTA=1.0]
Eliminate Mar
Round # 3 Sep=9.0 May=8.8 Jul=6.9 Oct=6.4 Aug=5.7 Jun=5.1 Apr=4.4 Dec=3.3 Feb=3.3 Nov=3.2 [NOTA=2.0]
Eliminate Nov
Round # 4 Sep=9.4 May=9.0 Jul=7.2 Oct=6.7 Aug=5.8 Jun=5.2 Apr=5.1 Dec=4.0 Feb=3.6 [NOTA=2.0]
Eliminate Feb
Round # 5 Sep=9.6 May=9.0 Aug=7.5 Jul=7.4 Oct=7.3 Apr=5.7 Jun=5.2 Dec=4.3 [NOTA=2.0]
Eliminate Dec
Round # 6 Sep=9.9 May=9.2 Oct=8.1 Aug=7.8 Jul=7.6 Apr=5.8 Jun=5.5 [NOTA=4.0]
Eliminate Jun
Round # 7 Sep=11.5 May=9.8 Jul=9.4 Oct=8.5 Aug=8.4 Apr=6.4 [NOTA=4.0]
Eliminate Apr
Round # 8 Sep=12.2 Jul=11.1 May=10.6 Oct=9.3 Aug=8.8 [NOTA=6.0]
Eliminate Aug
Round # 9 Sep=13.1 Jul=12.9 May=11.6 Oct=10.4 [NOTA=10.0]
Eliminate Oct
Round #10 Sep=14.8 Jul=14.3 May=13.8 [NOTA=15.0]
Eliminate May
Round #11 Sep=22.5 Jul=17.5 [NOTA=18.0]
Eliminate Jul
Round #12 Sep=29.0 [NOTA=29.0]
Winner: Sep
So an Instant Runoff is actually another way to
evaluate an approval election if you want to restrict voters to a single full
vote divided among all choices selected.
The most exciting advantage of this approach is
the ballots allow two ways of counting:
1. An approval vote with full votes for each candidate. (Used
to compute federal funding for parties and such)
2. A plurality vote with divided votes. (Used to determine
elimination order in an instant runoff process.)
Going to a ranked ballot election (standard IRV)
is an even better step to allow voters to express a preference between
candidates, but that requires more from our voting
system. A ranked ballot would need, for
example, each candidate could have 5 columns which could be marked for 5
possible ranks. Voters would mark inside
a 5 column, 6 row matrix of circles for a 6 candidate election, marking up to
one circle per row and column. This is clearly more complicated for voters and a
problem for existing equipment.
IRV with ranked ballots is somewhat of an overkill system
(with two strong parties dominating) since it puts all that work to allow voters
to rank candidates and then only uses the lower ranks if the favorite is
eliminated.
Approval is a smaller reform and appears to me to fairly
address the biggest concerns of our current elections - that of spoiler
candidates, and voters needing to neglect their favorite to affect the
election.
Approval elections (with or without runoffs, with or without
divided votes) won't change how people vote, especially for those
voters who already have a clear favorite. It will most affect the choices
voters who want to vote third party but through fear would otherwise vote a
compromise among the expected top two candidates. It won't help 3rd parties
win, but it will give them proper due for their support. It also allows
dissatisfied major party voters offer a protest vote if a 3rd party supports
important issues being ignored, even if they grudgingly are also willing settle
for a top-two candidate.
If any of you have questions, I'd be happy to
respond.
Sincerely,
Tom Ruen
