No Panaceas

Thursday, March 11, 2004
Reforming the Nomination Process II or is Instant Runoff Voting really that great an idea?: What percentage of the vote did Kerry get in Iowa? 38% New Hampshire? 39%.

So, on the one hand, Kerry won both states. But he won only in the sense that the media declared him the winner because he won a plurality of the vote. He was, as they say, "first past the post." Yet, a clear majority in both states -- 62% and 61% -- preferred someone else. These two "victories" created such a powerful snowball effect that the race was effectively over.

My point is not to pick on Kerry, but rather it is to illustrate the extreme disparity between democratic ideal and democratic reality. It is not just that voters in two completely unrepresentative states have far more influence than voters in other states, it is that a tiny fraction of voters in those unrepresentative states, usually far less than a majority, have far more influence than voters in other states.

It is true that delegate allocation is more proportional, but what really matters is the perception in the press and among voters about who is a winner and who did well and who did poorly. These perceptions shape subsequent races and the ultimate allocation of delegates.

We know from Arrow's Impossibility Theorem that no election system is perfect. Yet, some are better than others and a system that uses plurality voting with many candidates is among the worst. This is one of the ways that the California recall was flawed and it is the way most of our primaries are flawed.

Let's imagine an electorate with 8 voters and 4 candidates, Red, Blue, Green and Brown. Suppose the voters preferences were as follows:

Voter 1: Red > Blue > Green > Brown
Voter 2: Red > Green > Blue > Brown
Voter 3: Red > Green > Blue > Brown
Voter 4: Green > Brown > Blue > Red
Voter 5: Green > Blue > Brown > Red
Voter 6: Blue > Brown > Green > Red
Voter 7: Blue > Brown > Green > Red
Voter 8: Brown > Blue > Green > Red

Straight up plurality voting yields the following:

Red: 3 votes
Green: 2 votes
Blue: 2 votes.
Brown: 1 vote.

So Red wins a plurality despite the fact that he was the least favorite candidate for a majority of voters. In fact Red could have been not just the least favorite but completely unacceptable to a large group of primary voters, yet won anyway. I am certainly not saying that this was the case with Kerry, but it is easy to imagine a situation where a candidate --perhaps an extremist candidate -- has a strong devoted following but is unattractive to everyone else. McGovern in 1972?

The reform receiving the most attention is Instant Runoff Voting. In more international settings it gets called the Alternative Vote or Preferential Voting. It's the way Australia elects MPs to its House of Reprentatives.

With IRV, voters ordinally rank candidate. If no one gets a majority of first preferences then the candidate who came in last drops out and then the second preferences of all the voters who voted for the last placed candidate are distributed to the other candidates. The process continues until a candidate breaches 50%.

For the above example, IRV would work as follows:

First Tally
Red - 3
Green - 2
Blue - 2
Brown - 1.

Brown drops out and her voters' second preferences are redistributed:

Second Tally
Red: 3
Green: 2
Blue: 3

Green now drops out and his voters' votes go to their highest ranked remaining candidate. For both Voter 4 and Voter 5 that is Blue. So Blue wins the election 5-3.

The good thing about IRV is that a candidate must demonstrate a breadth of support to win. In my opinion some supporters of IRV overstate its value. I'm thoroughly unconvinced, for example, that IRV will increase voter turnout. There is a big downside. If there are a lot of candidates then IRV can be a major pain for voters. Imagine if IRV were used for the California recall with its 135 candidates. Don't forget too that more than nine candidates were on the ballot in the some of the primary states. A total of twenty-three candidates were on the Democrat's ballot in New Hampshire.

It is possible to modify the balloting so that voters only rank the candidates they care about. But even still the process of finding and ranking candidates is tricky. Oops I meant Gary Coleman to be sixth not that watermelon guy! And what if you have preferences for the bottom of your list and the top of your list and then you are indifferent among everyone else? You would have to rank the entire list.

I think the above quibbles are not that big of deal. IRV is no panacea -- imagine that -- but it would be an improvement on first-past-the-post. I won't go into how delegates would be allocated, but that's not a difficult problem.

There is another alternative and that is Approval Voting. With Approval Voting you vote only for those candidates you like. So, let's imagine that in the above ranking all the voters disapproved of their fourth choice. The vote tally would be:

Red: 3
Blue: 8
Green: 7
Brown: 6

So Blue wins because she is acceptable to the largest number of voters. On a range of normative and practical criteria I think Approval Voting is the better choice, but both alternatives are reasonable. You could even construct an argument -- perhaps a naive one -- that either approach would reduce negative campaigning because candidates have to worry about driving up their own negatives.

Wednesday, March 10, 2004
Reforming the Nomination Process I: Now that the nomination races are over it is time to return to that quadrennial favorite: How should we reform the nomination process? The problem with the process is well known. Mainly because of an historical accident two states -- Iowa and New Hampshire -- enjoy an inordinate amount of influence over who gets nominated. It's not so much that either state directly chooses the ultimate nominee, as much as the states directly influence who loses, thus limiting the choices available to voters in subsequent primaries and caucuses. This would be less of a big deal if the party electorates in these two states reflected the racial, ideological, and geographic diversity of the respective party's national electorate. But of course they do not. For example, the national Democratic electorate is far more urban and racially diverse than Democrats in Iowa and New Hampshire. Likewise the core of the modern Republican party is in the South and non-coastal West.

In response to the Iowa and New Hampshire advantage, other states have moved their elections earlier and earlier in the process. This has created a dramatically compressed schedule where voters have little time to focus on the campaign -- you can't expect all voters to pay attention much more than a week before an election.

The question is how to reform the process? (Assuming political conditions allow you to reform it, an issue I'll address at the end.) There are two criteria that need to be addressed. First, the voters who have the most say, i.e., those who go early in the process, should be diverse and they should change from election year to election year. Second, underdog (i.e., underfunded) candidates should have a chance to compete. One of the very positive attribute of both Iowa and New Hampshire is that their size, and the fact that they go first, makes so-called "retail" politics not just possible, but necessary. This puts a premium on organization and small group interaction between voters and candidates.

There are any number of ways a new system might be installed, but the best one I have seen is the Graduated Random Presidential Primary System, better known as the California Plan. Thomas Gangale's article in the recent PS does a nice job of explaining it. (Full cite: Gangale, Thomas. 2004. "The California Plan: A 21st Century Method for Nominating Presidential Candidates." PS: Political Science and Politics 37(1): 81-87. PS is a subscription periodical. It should be available in any university library and in most good municipal libraries.)

Without going into much detail, here is the basic plan. First, all states are indexed by their number of U.S. House seats. U.S. territories and the District of Columbia each are defined as having one seat. On a bi-weekly basis a set of primaries are held. The first set consists of randomly selected states whose index number sums to eight. Thus, for example, this first group of primaries might consist of Alaska (1), New Mexico (3), and Kansas (4). Or it might include just Minnesota (8). The second group of primaries include states that sum to sixteen, then followed by twenty-four and so on until the primary group that sums to eighty completes the full list of states and territories.

As Gangale notes, however, the above penalizes the big states, especially California (52). Under such a system California can never vote earlier than the seventh group. Thus the need for a modification, one that gives the big states a reasonable shot at voting early. Here is the sequence Gangale lays out:


Thus California has a chance to be in the fourth, sixth, eighth, or tenth groups.

This idea has a number of possible advantages. First, retail politics remains possible in the early stages of the campaign because the states will tend to be small. Yet chances are good that the early states will be rather diverse. DC and Vermont, for example, are both small but strikingly different in terms of race, location, economic base, etc.

In my view, a fundamental reform such as this should coincide with later primaries. That is, we need to shorten the election period. Why not have the first group start in April (just in time for Tax Day) rather than January?

Will anything like this happen? Probably not. At the end of the day the national parties get to make their own rules, so it is possible to overcome the Iowa and New Hampshire opposition. But those two states have done a "good" job of blackmailing the parties and presidential candidates into supporting Iowa and New Hampshire preeminence. (As Gangale notes, if one of the two parties went first, the other will follow, as was case with the McGovern-Fraser reforms that created the modern nomination system.)

What really is needed is for one of the party leaders, i.e., Kerry or Bush, to put his weight behind reform. But why would they? After all, the current system worked just fine for both of them.

Tuesday, March 09, 2004
Wacky, but credible: I've started a context with my students to see who can come up with the wackiest "credible" VP candidate for Kerry.

Here's the criteria: The person has to be "credible" as in -- for a few seconds -- "Yeah, I can see that. Hmmmm." The person also has to be wacky as -- after a few seconds -- "No way, that's nuts!"

Paris Hilton is wacky but not credible. Dick Gephardt is credible but not wacky. The guess is even better if the candidate is mentioned by a major media outlet. So here is the list so far:

1. Bill Clinton
2. Tom Brokaw
3. John McCain.

All of the above have been discussed in major outlets: Clinton, Brokaw, McCain.