07
Dec
happyemergency asked: I'm not sure of a better way to post a response to your comment, so here it is! Approval Voting is a very poor choice, in my opinion, because it doesn't allow for a complete expression of voter preference, and can be highly Tactical. I also wouldn't consider it to meet the Condorcet Winner criteria. Philosophically it fails because it doesn't capture voter intent, where as ranking candidates relative to each other is 100% intuitive and an integral part of human psychology.
I appreciate the response. To be fair to readers I’ll explain Condorcet and Approval Voting.
Condorcet: Rank all the candidates from best to worst. Calculation creates pairwise comparisons between all candidates to determine a round-robin winner. When there is no round-robin/beat-all winner, then a complicated algorithm is used to pick a winner.
Approval Voting: Pick (no ranking) as many candidates as you wish. Most votes wins.
I can understand your initial position. Looking at it, the expression between these systems is completely different. One is ranking showing different positions (though not degrees) between candidates. The other (Approval) allows you to express opinions on all the candidates but only at two levels—approve or not. These expressions are indisputably different.
But what’s more important than how your vote is expressed? Perhaps more important is how your vote is calculated to determine the winner. And Approval Voting is actually quite good at electing the Condorcet winner when one exists (Reasoning and proof here). Approval Voting even elected the Condorcet winner in a large-scale study looking at the 2007 French Presidential election (beautiful comparison data set). That’s even when IRV and Plurality failed to do so.
And Approval Voting gives you these results with super simplicity. Approval Voting is about as simple as it gets. No special computing software is needed, can be easily counted by hand, ballot spoilage rate is the lowest of any voting method, and the results are extremely easy to understand. All these hit obstacles with a Condorcet method. Plus you have to worry about cycles with Condorcet (A>B, B>C, C>A). There is no such worry with Approval Voting and no complexities to address these problems like with Schulze’s method.
As far as tactics, Approval Voting is one of the very few voting systems that ALWAYS lets voters choose their honest favorite. Condorcet, IRV, and Plurality all fail this. A simple explanation of Approval Voting tactics is to choose who you want out of the frontrunners and then to vote for everyone else you like. Sometimes, you may choose more than one out of the frontrunners (tactics on Approval Voting). Or, you could be more exact about it. Further, vulnerability to strategy is not the end-all-be-all. Random ballots are immune to strategy, for example, but we surely wouldn’t advocate that method.
I would strongly recommend that if you read anything that you read this article: www.electology.org/approval-voting. You’ll find a lot of information you’re looking for there if you’re interested in different ways to analyze voting methods and why Approval Voting may not be so bad. Plus, it has a cool video by one of its co-founders—Steven Brams.
Finally, if you’re extra ambitious, consider whether it’s more important to see how a voting system performs over many elections on average. How might a system perform using voter utility as a measure while factoring in different levels of voter strategy? Using Monte Carlo simulations, that has an answer. And given strategic voters, Approval Voting does pretty well—much better than IRV, Condorcet, and Plurality with strategic voters. This measurement tool is called Bayesian Regret. (Bayesian Regret).
Thanks for the question and feel free to follow up on anything I’ve missed.
