[EM] Get Real, Richard Moore:

Richard Moore rmoore4 at home.com
Mon Apr 2 21:10:46 PDT 2001

I Like Irving wrote:

> - - - - - - - - - - - - - - - - - - - - - - - - - - - - 04/02/01
> Mr Richard Moore,
> You wrote:  "Why is the particular majority selected by IRV entitled to be
> the majority that rules, when there are so many other possible majorities?"
> Davison: Yes, you are correct, there are other possible majorities than D
> 11 of 17.  Had you ran one more runoff cycle you would have gotten a
> majority of D 17 of 17,  but that is not what you had in mind was it?  You
> are fishing for a majority that will elect one of the lowest three
> candidates like A or B or C.

I'm not fishing for any majority. Only IRV fishes for majorities. Approval
the most widely approved candidate, and Condorcet searches for the one
who is favored by majorities over every other candidate, or who has the
most significant majority win if that candidate does not exist.

>      In order to do that, you will need a different method.  Approval
> Voting is currently in vogue on this EM list.  We should try that method
> before it goes out of fashion.  Approval Voting will give the following
> results to your example:
>               17 A,   17 B,   17 C,   17 D,   17 E

Don't insult my intelligence with stupid arguments. Why do you think this
particular voting population is so dumb that every voter approves every
candidate? If you want to create pro-IRV examples based on a stupid
electorate, go right ahead.

>      Opps, looks like Approval Voting is unable to give you a `better'
> winner with a different majority.  In fact, Approval is useless, for it has
> yielded a silly result.  That's Approval for you.

Approval didn't get the silly result. You did.

In fact, if you had only stopped to think, you would realize how your
argument only supports the proposition that Approval is better than IRV.
How? Because anyone reflecting on your absurd statement will realize
that, in fact, there is no way to tell what the Approval outcome will be
from the given information (rankings). As I went on to observe in my
original message, there are two extreme possibilities (and everything in
between): "...the 11 voters who preferred some permutation of ABCD
over E may actually have very close ratings of those four", and "5 of the
11 dislike D almost as much as they dislike E." The inability to
distinguish these extreme cases is another failure for IRV. Approval will
give different outcomes for the two extreme cases, however. That is
Approval's strength: It emphasizes the preferences that matter most to
the voters.

>      Bucklin will give us: 17 A,  2 B,  3 C,  6 D,  6 E
>      So, it is candidate A, the lowest candidate of top choices, that you
> have determined shall win the election in your concocted example.  Why did
> I not see that before, yes indeed, why not?

OK, Don, from the original example, how did you determine (1) that I have
determined that A shall win, and (2) that A is "the lowest candidate of the top


Here's the example:

6    EABCD
1    ACBDE
2    BACDE
2    CABDE
6    DABCE

Without knowing the actual Approval votes, there is no way to determine who
would win in Approval. Suppose everyone approves their top three choices
except the B voters, who vote for their candidate only? Are you surprised at
this result? You shouldn't be.

What are the top choices, anyway? Certainly not D and E, unless we assume
most of the voters rank their top four choices closely, and even then that
be debatable. Let's not be guilty of intellectual dishonesty by proclaiming
winners and losers and then criticizing methods for failing to give our

>      What is wrong here is that Bucklin violates the Golden Rule of
> Preferences, that is, later preferences are not to harm nor help earlier
> preferences. The Bucklin solution is a clear case of lower choices helping
> to defeat top choices.

I'm not advocating Bucklin so I don't understand why you bring it up. Maybe
it's your favorite method to pick on when IRV is under fire.

>      Besides, your example is unrealistic on three accounts.
>      One, voters of the top two candidates, D and E, would not march
> lockstep and vote for the same second choice A, and then the same third
> choice B, etc.

You can permute the rankings of A, B, and C withing each faction any way
you like, keeping D and E fixed, without affecting IRV's silly outcome, so
this objection is not relevant.

>      Two, few of the D and E voters would choose A or B as a second or
> third choice when we consider that the voters of A and B chose D and E dead
> last, the dislike would be mutual.

D and E are (most likely) polar opposites, so supporters of D will rank E last
vice versa. A, B, and C appear to be centrists, and who's to say which of
three centrists an extremist party would prefer to align with. There might even

be cynical reasons for this alignment, such as D and E voters alike feel that A
B would be ineffective as leaders and therefore would be easy to subvert.

If that's not enough for you, see my earlier observation about permuting A, B,
and C.

>      Three, there is no reason for the voters of the two front running
> candidates, D and E, to make any lower choices. Their two candidates are
> head and shoulders above the three lowest candidates, it is of no interest
> to these voters which of the three is to be the last candidate.

Is that how voters vote in IRV? I'm pretty sure that, strategically, IRV
truncation. Isn't it the whole alleged point of IRV for voters to express all

And, strategy aside, supposing the D and E voters do not make any lower
choices. How does the outcome of this particular election change?

>      A more realistic example would be:
>         6 E,  1 ACBD,  2 BACD,  2 CABD,  6 D
>    Irving: candidate D still wins with 11 of 17

That's not "more realistic". It's the same example truncated.

>  Approval: candidate D wins, 5 A,  5 B,  5 C,  11 D,  6 E

Again you are inferring Approval votes from rankings, and doing a poor
job of it at that.

>   Bucklin: candidate D wins, 5 A,  5 B,  5 C,  11 D,  6 E

Again I don't care about Bucklin.

> Condorcet: candidate D wins all D pairings.

Which spanks the E voters very thoroughly for truncating their votes
in Condorcet. Stupid, stupid E voters!!! (Not very bright of the D
voters either, to put so much faith in the centrists to help their cause.)

>      The jury is in, candidate D is the winner, but we already knew that.
>      You were merely hoping against hope that your bottom feeding candidate
> would somehow win - it's not going to happen.  We will not allow you to
> leverage one vote into a win.
>      To see this, all you need to do is to `Get Real' Richard Moore.

Sorry, you've only shown a win for D in IRV, which Is No Good, and in
Bucklin, which I don't care about. For Approval you incorrectly inferred
the votes, and for Condorcet you assumed the D and E factions aren't
very bright.

Now tell me, which candidate is my "bottom feeding candidate"? You don't
even know.

 -- Richard

More information about the Election-Methods mailing list