One topic that I usually don’t see come up in discussions about third-party voting and “lesser-of-two-evils” voting is how the actual voting system itself discourages third parties. An essential part of third party strategy should be to update the voting system to better express voter preferences.
Every election has some discussions about the electoral college, but I think it’s generally pretty straightforward.
1. It’s silly to have an arbitrary geopolitical unit like the state of North Dakota have a disproportionate influence on the government. I don’t think we should kill things like two-senators-per-state entirely because regional discrimination is real, but we should shift it significantly in the direction of population-based. To borrow some software nerding, it’s a parameter tuning problem.
2. Winner-take-all is bad because the voices of the losing sides are essentially ignored. Maine and Nebraska’s per-district electors makes more sense.
This post became humongous D:, so the voting system discussion is after the jump.
The run-off process in most countries, which wiki explains much better, is basically you vote in rounds. If the first round doesn’t give a clear winner, then you pick the top two candidates(1) and run the election again.
Obviously, this has the strong disadvantage of running an election multiple times, which is an expensive administrative burden.
It also has a gaming problem. Wiki’s example of a recent French election illustrates this well. Here are the top candidates from the first round:
- Jacques Chirac 19.88%
- Jean-Marie Le Pen 16.86%
- Lionel Jospin 16.18%
Chirac and Jospin are from the mainstream French political parties. Le Pen is a ludicrous racist who would have no chance in a head-to-head election against any reasonable candidate. Here are the second round results:
- Jacques Chirac 82.21%
- Jean-Marie Le Pen 17.79%
Now, what does this mean? If you were a supporter of Chirac and were reasonably sure that he would be in the top two, you could vote for Le Pen in the first round. You would also have incentive to convince others to vote for Le Pen to edge out Jospin, who might compete more strongly with Chirac in a second round.
Voting should not be subject to this kind of gaming. People should vote their preference clearly.
A cousin to this voting system is instant run-off voting, which has gained some traction in the US since it mitigates (but doesn’t completely eliminate) the “spoiler effect” (e.g. what people blame Nader for in Bush v Gore). I’ll restrict my urge to nerd out on this and link to this good explanation of similar gaming problems with it.
So now what’s the solution? The article about instant run-off provides two good alternatives: approval voting and range voting (a.k.a. score voting). The wiki page for approval voting has all sorts of interesting info about voting system metrics. It is some major game theory nerding, if that is your thing.
In approval voting, you simply give a vote to every candidate that you like. The one with the greatest number of votes wins(2).
This negates the “spoiler effect” that people blame Nader for. You will always vote for your first preference.
When your first preference has pretty much zero chance of winning, you will always vote for the lesser of two evils. For example, in the past election, I would have probably voted Jill Stein, Rocky Anderson, and Obama.
However, when your first preference has a chance to win, then maybe you need to make a harder decision.
Another big advantage of approval voting is that it’s simple. In fact, it already exists in local elections like my local school board, where you can vote for up to two.
There’s little confusion or administrative overhead for counting the votes too.
I think approval voting is sufficient for the current political state of the US. If third parties start doing better though, e.g. if Jill Stein had a shot at the White House, I might not want to put an approval vote for Obama.
Range voting, which is a more general version of approval voting(3), helps with this case. In range voting, you assign a score or “no opinion” to each candidate. The highest average score wins(4).
For example, you might be allowed 0-9 or “no opinion” as options. So I would maybe vote something like:
- Jill Stein: 9
- Rocky Anderson: 8
- Barack Obama: 7
- Gary Johnson: 0
- Mitt Romney: 0
Note that a 0 is not neutral. A 0 will lower a candidate’s average. So you can effectively downvote candidates you dislike. “No opinion” will prevent your vote from affecting the average.
This allows you to put a number to how much less you like a candidate.
However, the drawbacks of range voting are as follows:
1. It’s complicated, which makes it harder to persuade people of its merits.
2. It still has opportunities for strategic voting. For example, in the fictional universe where I think the Green Party will definitely beat the Republican party but not necessarily the Democratic party, I might be tempted to vote 0 for the Democratic ticket.
3. It adds administrative and training overhead since you have to do the averaging. However, it’s not so bad since each district can more or less report results independently(5).
Hopefully this is some useful info on voting systems. This is a crucial part of better expressing the preference of the voters. In the short to medium term, voting system reform is probably the only way to guarantee getting more third party voices into our government.
(1) Usually clear winner means greater than 50%. Also, you don’t have to pick the top two, you can pick the top n-1 or something. There are lots of variations.
(2) There can be variations where you do a run-off or require a certain margin of victory, etc.
(3) Math nerds love saying sentences like this: approval voting is a special case of range voting where your only options are 0 and 1.
(4) Again, there can be variations requiring a certain margin of victory. Importantly, with averages, there must be some kind of quorum (minimum number of votes) requirement.
(5) Warning: math nerding!
Let sequence s=x1, x2, x3…xn be the values of the votes for some candidate.
Mean is sum(x1 through xn) / n.
You can replace any subsequence xi through xj with (j-i+1) copies of mean(xi through xj) and sum(x1 through xn) will still be the same.
mean(xi through xj) = sum(xi through xj) / (j - i + 1) (by definition of mean)
mean(xi through xj) * (j - i + 1) = sum(xi through xj)
That means each district only needs to report its mean and the number of voters with no information lost for mean-calculating purposes.permalink | 1 note