[EM] No evidence that IRV doesn't fail. Reasons why it must.

MIKE OSSIPOFF nkklrp at hotmail.com
Thu Jan 22 02:06:02 PST 2004

David Gamble--

I'd said:

>Is that what happened in Australia? And don't say that IRV hasn't committed 
>its failures in Australia. It wouldn't show up in the data that are 
>recorded  and published.

You asked:

What are you saying here Mike?

I replyi:

I'm saying that the IRV faillures that we discuss wouldn't show up in the 
data that are recorded and published. I thought that was pretty obvious from 
what I wrote :-)

You continued:

That even though you can find none of IRV's
failures in recorded and published data you nevertheless KNOW that they are

I reply:

Did I say that IRV failures were in the recorded and published data? On the 
contrary, I said that they wouldn't show up there.

You continued:

You give the impression that your opposition to IRV has elements of
quasi-religious belief about it- I don't need evidence, I know these things 
to be true
without any need of evidence.

I reply:

Now you're revealing yourself to be a complete twit.

If the recorded data aren't of a kind that can show how often IRV fails, 
that doesn't mean that the data have shown that IRV doesn't fail. It means 
that evidence, one way or the other, isn't available.

I don't know how you define "need evidence".  Evidence would be nice, if it 
were available. Regrettabley, it is not available, because the data recorded 
and published aren't the kind of data that could show how often IRV fails.

IRV opponents aren't to blame for that fact that IRV results aren't recorded 
in a way that could show how often IRF is failing.

But, lacking data that could give evidence one way or the other, it's 
evident that the IRV failures that we describe will happen for sure. That's 
obvious and undeniable just from the definition of IRV.

They happened so frequently in Merrill's spatial simulations that they 
caused IRV to have significantly poorer social utility than Approval. 
Merrill refered to it as the "squeeze effect". Median candidates tend to get 
eliminated by vote totals accumulating on other candidates, as extreme 
candidates get eliminated and transfer.

That's a robust scenario that requires only a gradual tapering of 
favoriteness from the median candidate. The squeeze effect can and often 
will eliminate a CW who is also the voted favorite of more people than any 
other candidate.

Then, in a differnt failure scenario, there's the least-avorite CW, who gets 
eliminated immediately.

With 3 candidates, and a 1-dimensional issue-space, all it takes for the 
middle candidate to be CW is for him to have favoriteness support greater 
than the difference between that of the extreme candidates. A sufficient, 
but not necessary, condition for that is for all the candidates to be within 
a factor of 2 of eachother in terms of favoriteness. If the candidates' 
favoriteness order varies randomly, then 1/3 of the time that middle 
candidate will be the smallest in terms of favoriteness, and will get 
immediately eliminated by IRV.

In other words that familiar IRV failure scenario will happen often.

Squeeze-effect example:

100: CDBEA

Let's simplify this by only showing the preferences that IRV actually looks 
at and counts:

50: AB
51: B
100: C
53: D
49: ED

It looks rather sparse, doesn't it, when we leave out people's voted 
preferences that IRV doesn't count.

C is favorite to about twice as many people as anyone else is. C is the CW. 
But IRV fails to elect C.

In thiis example the non-CW candidates are about equal in size, to show how 
popular a CW can lose in IRV. As I said, this scenario happens under 
widely-varying conditions. A less special-looking example:

(In sparse form, for simplicity)

43: AB
68: B
100: C
71: D
35: ED

Obviously these numbers could have almost any values, requiring only that 
the candidates to each side of the CW add up to more than 100, and that the 
sums of the 2 sides don't differ by more than 100.

These aren't rare special conditions for this IRV failure.

Least-favorite CW example:

40: ABC
25: BAC
35: CBA

Anyone claiming that those plausible & ordinary situations won't happen 
often should explain why they believe that.

Though there aren't recorded and publislhed data that would show how often 
these failures are happening in existing IRV elections, it's a sure thing 
that they'll happen regularly.

I"d said:

>Anything worth doing is worth doing right. Doing something right the first 
>time is better than doing it wrong, and then hoping to fix it later.

You replied:

Rejecting any option because it is not 100% of what you want is a usually a
very poor strategy in any situation. Very rarely in life can we have 
that we want. We have to compromise. Some improvement, 50% of what we want 
better than nothing. Any improvement is better than no change.

I reply:

You make it sound as if IRV's inadequacy were somehow unavoidabale. But it's 
merely the avoidable result of the fact that the busiest promoters won't 
take the trouble to educate themselves.

Mike Ossipoff

Rethink your business approach for the new year with the helpful tips here. 

More information about the Election-Methods mailing list