In my previous post I proposed a bargain between the voters of Howard County and those who want to extend the time Howard County Council members can serve: Tie the proposed charter change to allow four council terms to other changes that scrap the way we currently elect the council and replace it with something better.

My first proposal is to abandon the use of council districts and return to the days when council candidates ran on a county-wide basis. However we can’t simply return to the old at-large scheme, which had its own problems.

That leads me to my second proposal: Instead of a conventional at-large voting scheme, let’s elect council members using a proportional representation scheme that takes into account voters’ preferences among the candidates and makes it more likely that the composition of the council will truly reflect the composition of the Howard County electorate— ideological, demographic, geographic, and otherwise.

For some reason proportional representation voting schemes have never taken off in the US, but they’re widely used in other countries. The details of the schemes can get a bit complicated, but the basic outline of the scheme I’m suggesting is relatively straightforward:

Rather than voting for up to five county council candidates, a voter would rank candidates in order of preference. In effect each voter would have one vote, which they would give to the candidate who’s their first choice. If that candidate didn’t need that vote to be elected, or if that candidate didn’t receive enough other votes to be elected, then that vote would be transferred to one of that voter’s other choices. (Hence the formal name of this particular scheme: the “single transferable vote” or STV.)

For example, consider a hypothetical 2014 county council race in which all current council members except Courtney Watson (who’s instead running for county executive) are on the ballot in the general election, along with some additional candidates from the two main parties and perhaps from one or more smaller parties.

A liberal Democratic voter in Columbia might vote for Mary Kay Sigaty as his first choice, and then add Calvin Ball, Jennifer Terrasa, and Zaneb Beams as his second, third, and fourth choices respectively, declining to name anyone else. Likewise a conservative Republican in western Howard might vote for Greg Fox as her first choice and then indicate Robert Flanagan as her second choice, stopping at that point. Every other voter would go through a similar exercise, some ranking a full list of five candidates and others expressing preferences for as few as one or two.

When counting the votes we’d start by looking at voters’ first preferences. If any candidates were selected as the first choice of more than one-sixth (16.7%) of those voting then they would be automatically elected. The one-sixth number is chosen to ensure that at most five candidates can be elected, since it’s impossible for six candidates to all receive more than one-sixth of the first preference vote. If we assume that 102,000 people cast valid votes (about the number of people who voted for a county council candidate in 2010) then the minimum number of votes needed to be elected (the so-called ‘quota’) would be 17,001 (102,000 divided by 6, plus 1).

If the first choices of voters in our hypothetical election matched the votes in the 2010 council election exactly then Greg Fox would receive 17,424 first preference votes, 423 votes above the quota, and would be automatically elected. We’d then take Fox’s 423 excess votes and redistribute them to other candidates based on the expressed preferences of the voters for whom Fox was the first choice. For example, suppose Robert Flanagan received 10,427 first preference votes, and 60% of Fox voters indicated Flanagan as their second choice. Flanagan’s vote total would then be increased by 253 votes (423 times 0.6), giving him a new total of 10,680.

Since this would not (yet) be enough for Flanagan to meet the quota, and since in our example none of the remaining candidates met the quota either, we’d eliminate the candidate with the lowest vote total and redistribute their votes. For example, suppose Zaneb Beams received 8,732 votes (same as in 2010), and that that were the lowest vote total of any candidate. Further suppose that 50% of Beams’s voters indicated Mary Kay Sigaty as their second choice. If Sigaty were the first choice of 14,333 voters (matching her performance in 2010) then we’d allocate to her another 4,366 votes (8,732 times 0.5) to give her a new total of 18,699. Since this is above the quota of 17,001, she’d join Greg Fox in being elected to the council, and like Fox would now have excess votes (1,698 to be precise) that could be allocated to other candidates based on the preferences of Sigaty voters.

We’d continue in this manner, redistributing the excess votes of winning candidates and reallocating the votes of the least successful candidates, until all council seats were filled. The details to make the math come out right can get a bit hairy, especially when calculating how votes get transferred in later rounds and having to account for voters who don’t provide a full set of candidate preferences, but that’s what computers and election officials are for. All voters would have to worry about is ranking candidates in order of preference, and that’s pretty straightforward.

But why should we bother going to all this trouble? I’ll address that question in my next post.