Tuesday, February 05, 2008

Two Ideas for the Primary System

Having watched the primary season unfold from a very nice vantage point, I think that the nomination processes would have been better served by approval voting. From a potentially long list of candidates, voters simply vote for as many of them as they find acceptable. The candidate with the most votes wins.

The main advantage of approval voting is that it allows voters the opportunity to express a preference for more than one candidate. The drawback to approval voting is that it does not provide voters with the opportunity to rank candidates within the set that they find acceptable. With approval voting, we wouldn't see primary voters having to worry about "wasting a vote" in expressing a preference for a candidate who has little chance of achieving a plurality. There would be less pressure on candidates to drop out of the race if they don't "win" one of the early states. This is particularly important given how much weight the early primaries seem to have.

I also find it odd that the national parties award delegates in proportion to a state's total population rather than some other measure that considers the distribution of voters across the two main parties. Consider the case of New York. There is almost no chance that it will go for the Republican candidate in the general election, regardless of the nominees from the two parties. So Senator Clinton's victory there tells us nothing about whether she is a more viable Democratic candidate in the electoral college than is Senator Obama. Ditto for Senator McCain.

It seems like a party could get a more viable candidate by awarding more delegates to states that are expected to be more competitive in the general election and fewer delegates to states that are expected to be less competitive in the general election. There is a limit to how much downweighting could be done in a state in which the party is strong--that would encourage state party leaders to discourage turnout, which is unhealthy to say the very least. But some movement in this direction could be helpful.


Jason F said...

Actually it is odder than you think, at least for the Republican primary. They do not award delegates proportionally to population. Instead they use a formula that includes population but also gives bonus delegates to states that voted Republican in 2004, have Republican Senators or a majority of Republicans in the Congressional delegation, and other such metrics.

David said...

re: they use a formula that includes population but also gives bonus delegates to states that voted Republican in 2004

That's a step towards what the Prof. is saying - if a state votes overwhelmingly Democrat in the last election, it's likely to do so again in the coming one, so the Republicans don't want to weight that state's delegates as heavily - that state's voters aren't going to make the big difference in the general election, so their preference in the primary is less important.

The thing is that it's only a half step. As Prof. Samwick said, they should overweight competitive states, not previously Republican ones. Just as underweighting overwhelmingly Democrat states makes sense, so, too, does underweighting overwhelmingly Republican ones. Texas, for example, will go to the Republican presidential candidate almost entirely regardless of which Republican wins the primary - so there's as little reason to overweight Texas as there is to overweight Mass.

Florida and Ohio, however - those states are always close. If giving their choices a bit more weight in the primary process will help increase their turnout (by choosing a candidate those states like better) in the general election, it makes sense to do that.

Tom said...

I'm not sure why you don't put a preference on the votes...at least "first choice" and "other acceptable choices"...as the number gets narrowed down to 2, many people would vote for both, kind of defeating the purpose of a primary.

Andrew Samwick said...

Once the number gets down to 2, in the last states to vote, then voters can choose only 1 of the candidates if they like without changing the overall system.

The main reason to avoid rankings, even your very reasonable modification, is that they make the ballot more complicated when there are multiple candidates. And even with that complication, there is still a need in your system to specify how to weight an "acceptable" relative to a "first choice" in determining a winner. (Approval voting specifies an equal weight. The current system specifies a weight of zero on the "acceptables.")

See this wikipedia article on the Borda count for a good description of electoral methods.

Tim Dreisbach '71 said...

Andrew: I am not sure if you wish to weigh in on Dartmouth politics or not.

There is a desire by trustees and many alumni to have a "clear majority" winner. This is simple in a two candidate race. The last constitutional proposal failed, in part, because it gave a small appointed committee powers to "manipulate" the ballot in order to create head-to-head situations (for example if there were 4 candidates for two seats).

Even if the nominating committee puts forward fewer candidates, an open petition process assures that multi-candidate elections are likely.

Would you recommend keeping the current approval system, or moving to a more complicated instant-runoff form of preferential voting?

Some in the Dartmouth political sphere claim that approval voting is too difficult to understand, people only vote for one or two they want, rather than for all except those they do not want, and that this misunderstanding of the rules results in split votes. If it is too complex for Dartmouth graduates, why do you feel it will work for elections by the general public?

(Full disclosure to Andrew and readers-- I am a member of the Alumni Association executive committee who prefers approval voting but is willing to consider other preferential systems.)

Biomed Tim said...

Here's Eric Maskin on Rank Order Voting.

Tom said...

Tim - Our ballots in Vermont are full of "Vote for up to X" where we can fill in one name or many depending on our preference. It's standard for our state house elections. It doesn't strike me as much different than Approval voting, except there's only one position...right?

Andrew, I actually trust people to digest information better than we give them credit for. We declare a "winner" based on number of first place votes, but then add the approval votes in for another tally.

Andrew Samwick said...


I don't think I have much to add to all the ink that's been spilled on Dartmouth governance elections. Whatever method is adopted, it would be nice if it aimed to bridge divisions rather than widen them.

Good luck,


Lord said...

I thought it odd as well, as it does overweight the states in which the party is a minority. As these minorities evidently tend towards moderation though, it may have the desired effect.

A Red Mind in a Blue State said...

How about ditching the primaries altogether? Or go back to a handful. Seems like primaries have grown in number about as fast as college football bowl games. How about letting the pros pick the candidates? Smoke-filled rooms gave us TR and Wilson and FDR and Eisenhower, and even to a great extent, JFK.

Primaries have given us Carter and Bush the Elder and Bill and W (tho Reagan does balance out that group somewhat).

To use another sport's metaphor, when they allowed the fans' vote to be the only method of choosing, baseball's All Star teams never put the best players on the field.