Ken’s Take: This is from my archives – one of my favs. The original article was inspired by Clinton’s win over elder Bush (the Perot factor), younger Bush’d win over Gore (the Nader factor), and Jesse Ventura’s gov win in Minnesota.
There’s current news in the article since the independent in NJ may allow Corzine to sneak thru, and the Conservative may prevail in NY 23 as the party cadidates split the liberal vote. It’ll be interesting to watch … and (I think), the article is a fun read.
* * * * *
Excerpted from WSJ: How Beef-Hungry Voters Can Get Tofu for President, March 14, 2003
Those odd ducks who scrutinize returns, calculate how each additional candidate affects the others’ chances and analyze strategic voting are hard at work. I refer, of course, to mathematicians.
Yes, there is a mathematics of elections.
Research has identified various voting systems world-wide in which, paradoxically, becoming more popular can make a candidate lose, abstaining gives your preferred candidate a better chance, and picking a winner means accepting someone a majority of voters don’t want.
This last paradox characterizes the U.S. system of plurality voting (vote for one; the top vote-getter wins). It works fine when there are two candidates, but with three or more, plurality voting can come up short.
For a democracy, the mathematicians’ most robust result is chilling. “It’s surprisingly difficult to identify a voting system that accurately captures the will of the people”.
* * * * *
The Election
So as not to inflame passions with current political examples I’ll illustrate his point with food.
You and two colleagues are planning an office party, and the caterer offers chicken, steak or tofu. You poll 17 invitees:
5 people prefer chicken to steak to tofu.
2 people prefer chicken to tofu to steak.
4 people prefer steak to tofu to chicken.
4 people prefer tofu to steak to chicken.
2 people prefer tofu to chicken to steak.
One organizer tallies the ballots by the plurality method, counting only first-place votes. Chicken wins (7 votes), while steak is last (4 votes).
A second organizer uses “approval voting,” in which voters mark all acceptable choices (everyone’s top two choices are acceptable). Now steak wins with 13, tofu gets 12 and chicken is last with 9.
The third organizer uses a point system that gives their first choices 2 points, second choices 1 and last picks 0. Now tofu wins with 18, steak gets 17, chicken 16.
The ‘winner’ changes with the choice of election procedure … An ‘election winner’ could reflect the choice of an election procedure” rather than the will of the people.
* * * * *
It gets better. Thanks to a mathematical property called nonmonotonicity, in some voting systems, ranking a choice higher can defeat it.
In a plurality-with-runoff system, the two candidates with the most first-place votes face one another in round two.
This time, we invite other departments to our office party, and get this first-round result:
27 prefer chicken to steak to tofu.
42 prefer tofu to chicken to steak.
24 prefer steak to tofu to chicken.
Chicken (27 votes) and tofu (42) reach the runoff. Assuming steak fans maintain their preference and give their second-round votes to tofu, tofu wins the runoff.
That seems fair.
But what if four people in the group of 27 chicken lovers are last-minute converts to vegetarianism and, in round one, prefer tofu to chicken to steak, like the group of 42?
Now steak (24 first-place votes) and tofu (46) make the runoff, in which steak beats tofu 47 to 46. Tofu’s late surge turned its win into a loss.
* * * * *
Such paradoxes tend to occur under specific but far from unusual circumstances.
With plurality voting, the most common is when two centrists face an extremist. The majority splits its vote between the centrists, allowing the fringe candidate to squeak in. In Minnesota’s 1998 governor’s race, Hubert Humphrey got 28% of the vote, Norm Coleman 34% and Jesse Ventura won with 37%, even though most voters ranked him last.
* * * * *
Thanks to such outcomes, scientists say what’s most needed is “a way for voters to register their second and third choices … especially in primaries, where there tends to be a large field.” Both a ranking system (give candidates 4, 3, 2 or 1 point) and approval voting accomplish that.
The U.N. chooses a secretary-general by approval voting. “It is particularly appealing in elections with many candidates … If your favorite candidate is a long shot, you can vote for both him and a candidate with a better chance without wasting your vote on the long shot. Approval voting would do a lot to address the problem of presidential-primary victors not being the choice of most voters.” Approval voting could well make more people (especially supporters of long shots) feel their ballot matters.
Still, no system is perfect. As Nobel-winning economist Kenneth Arrow proved mathematically in 1951, no voting system is guaranteed to be free of paradoxes in a race with three or more candidates, except one — a dictatorship.
The Homa Files 