Thursday, November 19, 2015

Best Methods for Electing a Single Person

OpaVote offers 23 different counting methods that you can use for counting the votes in your election!  Most other election providers give you just one option.  Here, we help you navigate the options for electing a single person (president, vice-president, treasurer, etc.) We'll have a follow up post to discuss electing a group of people (committees, councils, etc.).
First off, if you insist on having a boring, traditional, and unfair election where voters pick one candidate and the candidate with the most votes wins, then select the "Plurality/FPTP/SNTV" option and move on.  On the other hand, if you really care about electing the best people to lead your organization, then keep reading.

Instant Runoff Voting

Generally, our top recommendation to most organizations is to use instant runoff voting (IRV).  IRV provides good results and is easy to understand. In particular, we recommend San Francisco RCV because the rules are well defined.

With IRV, a voter ranks the candidates in order of preference (e.g., 1, 2, 3, etc.). The first step in the counting is to count the first place votes. If a candidate receives a majority of first place votes, then he or she is the winner. If not, the candidate with the fewest first place votes is eliminated and those votes are transferred to their second choices. This elimination process is repeated until a majority winner is obtained.

The results of an IRV election are also easy to understand, and here are results of our favorite ice cream flavor poll with IRV.

Condorcet Voting

For organizations that are tolerant of more complex vote counting methods (e.g., mathematicians and computer programmers), then Condorcet voting is also a good choice.  Condorcet voting elects the candidate who wins all pairwise elections among the candidates (the "pairwise winner"), if such a candidate exists, and if not, uses another selection criteria.

In some instances, Condorcet voting may elect a "better" candidate than instant runoff voting.  I put "better" in quotes because there is enormous and subjective debate about what a "better" candidate means.  If you are curious, I recommend reading about Arrow's impossibility theorem, for which Kenneth Arrow won a Nobel prize.  For reals.

The results of a Condorcet voting election are more difficult to understand.  A relatively simple example of results can be seen for our favorite ice cream flavor poll with Condorcet.  In that example, the pairwise winner exists.  When the pairwise winner does not exist, the results are much more complicated.

Approval Voting

Another good option for electing a single person is approval voting.  Approval voting is REALLY simple.  Each voter just says yes or no for each candidate, and the candidate with the most yes's is the winner.  Here are results for our favorite ice cream flavor poll with approval voting.

I personally don't like approval voting, because I think the yes or no option for each candidate can be difficult decision to make, and many voters are tempted to approve only their favorite choice.  Approval voting does have strong supporters, however.

Borda Count

I'm throwing in one more because I recently recommended the Borda count to a customer.  The Borda count is a good solution when the number of candidates is large compared to the number of voters.  For example, if there are 5 voters selecting 1 winner among 20 candidates.

The Borda count is even easier to understand than IRV.  Suppose there are 20 candidates.  The voters rank the candidates, and a candidate gets 19 points for every first choice, 18 points for every second choice, 17 points for every third choice, and so forth.  The candidate with the most points wins.  With the Borda count, voters should be required to rank all of the candidates.

With IRV, when the number of candidates is large compared to the number of voters, it is likely that there will be lots of ties that need to be broken.  For the example above, it is likely with IRV that 5 candidates will receive 1 vote and the other 15 candidates receive 0 votes, and you then need to randomly select candidates for elimination.  Because Borda computes counts from all of the rankings, ties are much less likely.

You could use Borda for other elections as well (i.e., when the number of voters is large compared to the number of candidates), but I don't recommend this and instead recommend one of the other counting methods discussed above.