[EM] IRV Failures

[Bart Ingles]

Eric Gorr wrote:
> In a recent conversation with an IRV supporter I asked the question:
> 
>  What cases would you accept as failure of IRV?
> 
> They answered:
> 
>  Where the general public (or a significant fraction of it) failed to
>  accept the results as legitimate, or at least beyond question.  The
>  2000 and 2004 Presidential elections are examples of failed elections.
>  San Franciso's election was heralded as a success.

I didn't know that 2004 was widely regarded as illegitimate.

> They also believe that IRV has never failed to produce a fully 
> satisfactory result. Can anyone provide evidence to the contrary?

I'm sure "satisfactory result" is already defined as "IRV winner", but 
here goes:

(1) Australian lower house:  District elections are two-party races, 
with Labour on the left and Liberal/National on the right, despite 
whatever multiparty system exists in the PR-elected Senate.  Although a 
two-party race doesn't provide much opportunity for an IIA or 
monotonicity failure, some would view the absence of true multiparty 
competition as the real failure.

Also, the prevalence of "how to vote" cards in Australia is often cited 
as an IRV-induced problem.

(2) San Francisco 2004:  Of the four Supervisorial races tallied using 
IRV, none elected a candidate with a majority of votes.  In one of the 
races, the winner only received about 30% of the original vote.  The 
main cause appears to be that voters were only allowed to rank three 
choices, even though there were many more than three candidates.

In all four cases, the IRV winner agreed with the first-choice Plurality 
tally.  Although IRV supporters are claiming that this is "positive"-- 
that IRV didn' produce "unexpected" results-- it is probably a 
predictable result of the three-choice limitation (the fewer choices 
allowed, the closer the result will be to Plurality).  Also, the three 
choice restriction would likely have masked any 
Condorcet/IIA/monotonicity failures.

(3) California gubernatorial race, 2002:  Although IRV wasn't used (it's 
difficult to show actualy IRV failures when it hasn't been adopted), the 
primary/general election was similar enough to a runoff to show what 
would have happened under IRV.  The Republican primary featured Simon 
and Riordan, while Davis ran opposed as the Democratic incumbent. 
Riordan was eliminated in the primary, even though he was viewed as a 
head-to-head favorite against either of the other two candidates.  Davis 
won, setting  the stage for a successful recall election a year later.

Ironically, a similar situation arose in the 2003 recall election, with 
McClintock and Schwarzenegger replacing Simon and Riordan.  Since this 
was a single-round election, the need for strategy was obvious and 
Republicans overwhelmingly backed the moderate candidate.

(4) Louisiana 1998(?) governor's race-- similar to California 2002.