[EM] Condorcet candidate losing in Peru?

[Jan Kok]

Thanks for alerting us to this interesting election, Rob.

I confirm that Flores is rather convincingly the Condorcet Winner,
according to the pre-election polls, and yet lost in the actual
Plurality with top-two runoff election.

Would Flores have lost in an IRV election?  This is harder to say (but
the bottom line is that I think Flores would have lost an IRV
election).  Here is an analysis:

* How Flores might have _won_ an IRV election:

The results of the first round Plurality election were:
30.848% Humala
23.534% Flores
24.348% Garcia
21.270% Others (the next three candidates got 7.452%, 5.758%, and
4.379%, followed by   14 candidates who got less than 1% each.)

Flores and Garcia were nearly tied in Plurality votes.  So, if the
Others votes flowed more to Flores than to Garcia, then Flores would
have made the the runoff and then would likely win.

The Wikipedia page says that Flores is "the leader of the
conservative, right-wing Unidad Nacional coalition", while Humalia and
Garcia are both leaders of "left-leaning" parties.  As the only major
"right-wing" candidate, it is possible Flores might attract more
second-choice votes from the Other voters, while the two major leftist
candidates would split the Others' preferences.

* How Flores might have _lost_ an IRV election:

Other than the left/right designations, there is no evidence that I
could see that the Other voters would prefer Flores over Garcia.  So,
if the Other's preferences would not upset the ranking of Humala >
Garcia > Flores, then Flores would indeed lose under IRV.

There are two bits of evidence that Flores was a centrist candidate,
and that she would have fallen victim to the infamous center-squeeze
effect of IRV.

The first bit of evidence is that Humala is a revolutionary - he led
an uprising against former president Fujimori.  Meanwhile, Garcia was
also a former president but his administration "was marked by...
economic crisis, as well as widespread terrorist activities..."  Thus
Garcia might be seen as "old establishment" in sharp contrast with the
revolutionary Humala, and thus Flores might actually be seen as
somewhere in between!

The second bit of evidence comes from assuming that the Others'
preferences would flow to the three main candidates in the same
proportion as the votes in the Plurality election.  Thus, with just
the three main candidates remaining, the votes would be:

39.182% (round to 39%) Humala
29.892% (round to 30%) Flores
30.926% (round to 31%) Garcia

If we combine those percentages with the Condorcet poll results, we
see that the partial orderings are:

11 H>G>F
28 H>F>G
12 F>H>G
18 F>G>H
25 G>F>H
6 G>H>F

(I find it helpful to visualize this by drawing a pie chart with six
sections in the order shown.)

There are very few H-first and G-first voters giving their second
choices to each other.  And the F-first voters divide their second
choices more evenly between H and G.  Thus, Flores seems to be the
centrist candidate in this election.

I have an ...intuition... (not sure there is much factual basis for
it) that minor candidates -- the low vote-getters -- tend to be
outliers on the political landscape.  If that is true, then we could
expect that the Others' preferences would flow more to H and G.  This
would increase the probability that Flores would fall victim to the
center squeeze effect in an IRV election.

Cheers,
- Jan

On 4/17/06, Rob LeGrand <honky1998 at yahoo.com> wrote:
> The current election for president of Peru may give an excellent real-
> world example of top-two runoff (and thus probably IRV) failing to elect
> a Condorcet winner.  According to the Wikipedia coverage of the election
> at http://en.wikipedia.org/wiki/Peruvian_national_election,_2006 , the
> official results of the vote that took place on 9 April 2006 will be
> released in around ten days.  The "partial official results" place
> Ollanta Humala ahead of Alan Garcia and Lourdes Flores, in that order,
> meaning that the runoff election to be held in around a month will likely
> include only Humala and Garcia.  However, the last polls before the
> election showed that Flores would win a runoff against either Humala
> (55%-45%) or Garcia (58%-42%), making her the likely Concorcet winner,
> since all other candidates were well behind the pack.  (Those same polls
> showed that Humala would be favored over Garcia 51%-49%.  My brother is
> in Peru right now with the Peace Corps and is hoping that Humala, a
> friend of Venezuela's Hugo Chavez, doesn't win--he says that Humala would
> be likely to kick Peace Corps out of Peru.)
>
> --
> Rob LeGrand, psephologist
> rob at approvalvoting.org
> Citizens for Approval Voting
> http://www.approvalvoting.org/