Tactical nomination


Bryce Carmony argues on Twitter that Score Voting is more vulnerable than Approval Voting to tactical nomination. His justification for this is:

My evidence is getting 10% of a vote is easier than getting 100% of a vote.

To make this claim clear and coherent, what I think he’s really trying to say is that changing the normalization boundaries is more likely to change relative scores than approvals. E.g. if you change my score for X from 0 to 2, that has no effect in Approval Voting because it’s a “disapproval” either way (assuming a threshold of 2.5 let’s say).

But the unproven assumption here is that this makes any difference on candidate behavior. If a tactical nomination is only half as likely to work with Approval Voting, so what? That might still be plenty of incentive. Or put another way, why would the incentive be big enough with Score Voting but not big enough with Approval Voting?

There are other factors. In a competitive election, where we already have a variety of candidates, and thus it becomes harder for tactical nomination to have any effect. You have to nominate increasingly more extreme candidates than yourself to convince voters you’re a safe choice. And if they’re using probability of winning in their approval threshold, then this won’t have much impact anyway, as those extremists are going to be deemed no-hopers.

I could go on.

I don’t think bullet voting is a very big issue, but to the extent some voters are irrationally optimistic, and thus willing to vote for their favorite, IoIA failure (and thus, vulnerability to tactical nomination) might actually be worse with Approval Voting. Let’s say 5% of voters might fall victim to this. That could easily be enough to wipe out any gains Bryce is speculating about here.

The larger point is that this is a higher speculative and complicated angle, and Bryce isn’t providing any arguments that pass the muster that most election theorists would put forth. Even for people who already prefer Approval Voting to Score Voting, and doubt many would find this particular argument compelling.


He also thinks that approval voting is immune to gerrymandering and he is vehemently paranoid of proportional representation. He is a silly person.


Gerrymandering works by exploiting the ideological bottlenecks created by partisan primaries. The winners of primaries are fixed into place by having a 100% partisan electorate regardless of how many votes you win a general by. 50%+1 republican votes buys you a 100% republican candidate.

Approval voting frees us from partisan primaries, we only adopted them because of vote splitting. With votes no longer split there is no reason to maintain primaries that solely polarize the electorate and reduce choices for voters. If you draw your district 55/45 you no longer get a 100% pure candidate you get a candidate who appealed to both the 55 and 45. Cracking districts is ineffective.

But what about packing districts to deplete rival voters. This is impaired on the practical effects of approval voting. Without primaries the need to register with a party is removed so fewer people will do it. With more centrist candidates on general ballots more people vote making the gray space you could use to Gerrymander disappears. Finally, with voters able to vote across party lines party affiliation is no promise of sole party support. Essentially, packing is harder to do and even if you succeed your reward is giving your rival ideology a powerful voice in a room full of consensus seeking centrists who are likely to listen to good ideas.

Can you Gerrymander in approval voting? Of course you can you could draw district lines to spell your name out purely for vanity. You could use computers to maximize surface area to make campaigning difficult for less sophisticated candidates, or any other crazy things. Approval voting still let’s you draw bad lines but it at least enables us to draw good ones should we choose to. Using first past the post voting there is no satisfactory way to district regardless of who draws the lines. Independent councils and computer algorithms can’t make FPtP distracts satisfying.

Proportional voting creates small winning coalitions which incentivizes representatives to use bribary and special interest to stay in power not general welfare. Approval voting requires large winning coalitions to stay in power which incentivizes general welfare over special interest.



I just wanted to say, welcome! I appreciate your sharing your thoughts, and I expect we’re largely in agreement with what you’ve said here. I look forward to some interesting feedback as we continue to transition off the old forum system (Google Groups).



…no, Gerrymandering works by concentrating and/or distributing voters of a particular ideological leaning. You would still have Safe Seats without a primary, they would just be safe for the party rather than safe for an individual.


Ciaran, I’ve argued with Bryce about this for hours on twitter and it went nowhere


Ciaren, individuals hold office parties don’t. Two different people can belong to the same party and be completely different. Bernie Sanders and Hillary Clinton are both Democrats but are idealogically quite different.

In approval voting without primaries 55/45 districts will elect substantially different candidates than 100/0 will. It’s easy to be hyper partisan and say “all X party is the same” but that’s not reality. Partisanship wasnt always a limitus test for identity in the past people would vote across party lines and the two parties were even more overlapped than they are today.


Electing candidates who are each great representatives of a special interest is more inclusive than electing a bunch of centrists who are each decent representatives of the entire population. The average distance from each voter to the ideologically nearest candidate is lower. More inclusive systems increase voter satisfaction and reduce conflict and divisiveness.


1928 Germany had a super inclusive proportional voting system that led to warfare not welfare. Centrism outperforms coalitions of fringes because consensus has the virtue of progressing in the same direction and that direction Has been a steady stream of progress.

Centrism is democracy, if you dislike that plenty of people hate democracy.


“Centrism” (relative to the electorate, not an absolute scale) is the democratic choice for single-winner elections, not multi-winner elections.


Seriously? Godwin’d already?

You’re ignoring the fact that the Nazi’s never won a majority, and had democracy not already been subverted, they likely never would have come into power

But don’t worry about minor things like facts…


I don’t think there’s any proof of this. You’d have to see countries with centrist-favoring systems like Score Voting and Approval Voting, and compare their success to similar countries with PR.


Yeah, I don’t think invocation of the NAZI’s serves this argument, or really any argument.


The point is without proportional voting Hitler never rises to power. He tried the Beer Hall Putsch and it failed. But when 2% of the vote gives 12 seats to a Fringe movement you have a problem. When you have 15 parties chasing 15 directions the chaos created by that creates the perfect conditions for the base common denominator taking over.

The Treaty of Versailles created massive stressors that some European conflict was likely inevitable. But the form it took under Hitler is absolutely a direct byproduct of party list voting. I’m not calling anyone a Nazi because they’re authoritarian in how people order soup at their restaurant I’m looking at arguably the greatest failure of Democracy and studying why it happened.

Clay cares about neat and tidy math in simple and elegant models with his computers so complex messy hazy subjects like history and human behavior don’t really get his motor going but why Democracy was such a train wreck in the Weimar republic is worth looking at.


This is a case of cherry-picking. You can look at thousands of election cycles under PR governments around the world, which never resulted in anything like the Third Reich. You can look at the USA under Plurality Voting, and see Trump, who bears striking similarities to historical authoritarian leaders.

The arguments you’re referring to were between Score Voting and Approval Voting, and the models are certainly not “neat and tidy”. Smith’s calculations used 720 different permutations of five “knobs”, including “ignorance factor”, and ratio of tactical voting. There were different utility generators. Jameson Quinn performed his own independent tests under a quite different model, and found strikingly similar results.

His findings are buttressed by a considerable amount of empirical data, such as:

The irony here is that your own argument, about tactical nomination, is completely theoretical.

Another long-standing retort by you was that intermediate scores are “half-assed”, which to any serious student of social choice theory, where we want to collect the most accurate information possible, is just ludicrous, purely emotional language.


Well, since there haven’t been any other countries that have used proportional representation in the history of the world, I guess that one example settles it!


I don’t think there’s any proof of this. You’d have to see countries with centrist-favoring systems like Score Voting and Approval Voting, and compare their success to similar countries with PR.

I don’t think this is something that can be “proved”; it’s just a definition. The average ideological distance between a voter and the nearest candidate is smaller in PR than in Bag of Centrists, so voters are better-represented by their government.

If you want to talk about some kind of practical outcome of different voting systems, then yes, you’d need empirical evidence from countries that have legislatures of centrists to examine. But the empirical evidence of PR vs majoritarianism says that PR works better because it is more “inclusive”; better at representing each ideology, at making their members feel included in government. I’d conjecture that centrist legislatures would fare somewhere in between majoritarian and PR, since they aren’t as inclusive in this sense.


psephomancy, while I generally support PR, this is where I have to disagree with you. I don’t believe that representation should be the ultimate goal of a voting method unless, by representation, you mean the ability for a group of voters to influence legislation. When using this definition of representation, the idea that groups of voters are better represented when the average ideological distance between a voter and the nearest candidate is closer is simply wrong. A multi-winner bayesian regret simulation is one good measure of how well voters are able to influence legislation and according to one such simulation, PR methods do seem to occasionally outperform single-winner methods such as approval voting.

That simulation does not prioritize the representation of smaller groups of voters so it isn’t really measuring proportionality. However, it could be transformed into a proportional satisfaction efficiency simulation by converting regret to its inverse: satisfaction utility, and rather than measuring the average satisfaction, such a simulation could measure the average logarithmic satisfaction, which I call proportional satisfaction because maximizing the log of everyone’s satisfaction is the utilitarian definition of proportionality. The difference between whether you believe it is better to maximize average satisfaction efficiency or average proportional satisfaction efficiency just comes down to how important equality is to you when measuring a voting methods performance.

However, I would avoid relying on multi-winner bayesian regret simulations and proportional satisfaction efficiency simulations too much because they often do a poor job of simulating how different ideologies are distributed in real life, and because alot of the factors that determine how ideologies are distributed in real life are determined by a variety of factors ranging from how districts are drawn, to how minorities are distributed, to how polarization affects ideological distributions, to how a diversity of perspectives in a legislative body can help promote a diversity of ideas rather than a bunch of groupthink, etc. I believe that any theoretical model of how ideologies are geographically distributed will be flawed in some big way, but for the time being, this oversimplified model may be the best possible comparison between single-winner methods compare to multi-winner methods available when electing legislative bodies.



That’s… an interesting interpretation of representative democracy.

unless, by representation, you mean the ability for a group of voters to influence legislation.

Can you clarify what this means?


Do we prepare for the best or the worst case? People like clay built the fukushima power plant because you could look at tons of reactors working just fine. What are the odds an earthquake hits just right it has to be low right???

There are great dictators in history even today some people would say Singapore is thriving under a benevolent dictator while democracy produces Donald Trump so let’s just get s dictator.

If you’re unwilling to plan for the worst case that’s fine just be honest that you aren’t and say you’re a gambler.