Optimal voting

Elections for the parliament are approaching again in Finland, and it’s time to think on how to cast your precious vote. The popular method seems to be to try out one or more of the voting aid websites (such as YLE Vaalikone or HS Eduskuntavaalit 2015 vaalikone), which typically present the voter with ~40 Likert-style questions on various contemporary political issues, and then rank the candidates and parties by the degree of match between your opinions and theirs.

Should you then vote for the candidate closest to you? Absolutely not!

At first sight, it would seem to be logical to choose a representative whose opinions match with your. However, this approach ignores the actual mechanism of politics and voting, and results into sub-optimal voting in situations where more than one candidate is being chosen (ie. all elections except for presidency and yes/no questions which are rare). Since your vote is your most valuable democratic asset, it would be wasteful to use it in a sub-optimal way. Here are two alternative approaches I’ve been developing and using during past elections.

Maximum leverage voting

voting

In this approach, one tries to maximize the impact on the current political consensus. This means to have as large as possible effect on the politics practiced by the elected. The approach is as follows: Imagine politics as a n-dimensional space defined by different axes (e.g., right/left, liberal/conservative, green/not-green). Vote for the candidate whose opinion is as far from the average opinion as possible, towards the factor defining your own opinion. However, if the candidate chosen in this way is too extreme / outlier to be considered seriously, choose a candidate closer to the political average.

For example, if the voter thinks that on a scale of 1 to 10 for right vs. left wing politics the current political consensus is around 5, but the voter would like to see politics around 6, he or she would like to the consensus shift to 6. The voter has the options of

  • Voting for a candidate proposing score 6 politics (Closest candidate, proposed by the voting aid pages), however the impact on the consensus is just a minimal shift towards 6.
  • Voting for a candidate proposing score 10 politics, where the impact is a larger shift towards 6.

Within party optimization

The second approach is for situations where D’Hondt method is used for tallying the votes. In the system, the each vote cast is primarily given to the party or list, and then each candidate is given a quotient based on his or her rank within the party. The most voted candidate on the list gets a quotient equal to all votes, the second gets 1/2 of the votes and so on.

So what your vote is actually doing is foremost supporting the party, and secondly determining the relative rank of your candidate. To get optimal results under this system, you should

  • Determine your party based on maximum leverage method (above)
  • Estimate the number of seats (marked as n) that party will get (e.g., based on previous elections or most recent polls) in your voting district
  • Estimate the top n+1 candidates within the party by expected votes they will get (again, using e.g., previous election data, candidate-level polls if available, Google Trends search frequency data or any other means)
  • Select a candidate from within the top n+1 using the maximum leverage method.

So for example, you choose that Party X is best suited for driving the political consensus to your direction. The party has received 2 seats from your voting district for the past few elections and no great change is expected. You estimate that candidates A, B and C are the top-3 likely candidates. Based on maximum leverage, you determine that candidate C is the maximum distance towards your direction from the political center and thus receives your vote.

Leave a Reply