Fun with voting

Speaking of the stupid ways we run our elections, I just saw this review of Gaming the Vote, about the stupid ways we run our elections.  I’ll try to read the actual book sometime soonish.

I do like the idea of Range Voting–I certainly don’t think it could be any worse than what we have now.  As far as I know it avoids most of the pathologies of our current plurality system (even with a pure popular vote, not worrying about the monkeywrenches due to the electoral college and whatnot), and the results are simple to understand and “obviously fair,” unlike, say, Condorcet voting and Borda count (which have problems of their own).

I think I like Approval Voting–a special case of Range Voting–even more.  I would expect it to produce generally similar results to Approval Voting when there are lots of voters (although says not); and it’s much simpler.  Arrow’s Theorem doesn’t apply to either, as they are not based on pure preference orders.


Tags: ,

4 Responses to “Fun with voting”

  1. Clay Shentrup Says:

    Range Voting produces better social utility efficiency than Approval Voting.

  2. mrlauer Says:

    Interesting point. I’ll try to make time to look into it, and experiment with the code provided, so I can comment intelligently. A few unintelligent comments in the meantime:

    –The differences, if I’m interpreting them correctly, are pretty small. The simplicity of approval voting might outweigh it. That’s simplicity both of ballot design and of the decisions voters have to make (even rating netlflix moves makes me think more than I’d like).

    –I would expect a pretty large fraction of voters to vote strategically–which I don’t think is “dishonest” in the slightest, FWIW–hence putting us in the small-difference part of the chart.

    –That result clearly depends on the underlying models of voters and utilities and candidates and strategies, and the model used is obviously very simple. Some other models could produce more similar results for range and approval voting; e.g. I would a random cutoff (rather than a mean cutoff) in approval voting generally to produce the same results as range voting, at least under the vaguely realistic assumption that large numbers of voters have similar preferences (something probably not captured in a 100-voter experiment) (I could be wrong, of course). That might be even LESS realistic, I really have little idea.

  3. Clay Shentrup Says:

    Rating is generally easier for people than a yes/no, which is part of the reason used ratings instead of a thumbs up or thumbs down “approval voting” system. The problem is that when options are right on the line, people have a hard time deciding yay or nay. But rating on a scale is far easier, because you can use a score like a 4 or 5 or 6, and know that even if you were off by a point or two from what you’d have decided if you spent all day thinking about it, that’s a pretty minor issue.

    Also the social utility efficiency difference between Range Voting and Approval Voting is pretty large actually. If you look at the Bayesian regret figures scaled and “inverted” (so that more is better, instead of worse) into social utility efficiency figures, it’s more clear:

    That large improvement in average voter satisfaction is probably worth the small cost of a larger ballot. Even if a very small fraction of voters use the sincere intermediate scores (and evidence says a lot of them will) that leads to a large improvement in overall voter satisfaction. Here’s more about honesty and strategy:

    Smith’s simulations used 720 combinations of 5 parameters (“knobs”) governing factors such as the fraction of honest vs. strategic voters, over the range from 0% to 100%, in increments. For every combination, hundreds of thousands of simulations were averaged together.

    Range Voting beat all other common methods (Approval, Condorcet, Borda, IRV, plurality) in ALL 720 models. So your “model is too simple” argument doesn’t seem to hold up to scrutiny.

    Also Smith showed that at a certain point, adding more voters makes a negligible difference, and just makes the simulations take a ridiculously long time. Hence the 100 voters. In general, any criticism you could think of, we’ve already thought of and beaten to a pulp.

  4. mrlauer Says:

    After reading Poundstone’s book–which I mean to blog more about, eventually–I was already pretty convinced of your first point.

    I still haven’t looked enough at the voting simulations to feel convinced by them. Not that I’m skeptical of the results (I’m not, actually), I just want to understand them. I’m particularly interested in the utility generators. The completely-random utilities in the code linked in your first comment really is too simple to convince me of anything; I know you have others that are more realistic, but I haven’t looked at them yet.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: