In Canada the population is divided into ridings for the purpose of voting, and each riding sends a single candidate to office by simple majority vote (whoever gets the most votes wins). This is a nice approach because it's simple and you get regional representation (the person in office represents his/her region).
The problem is that each candidate also (usually) represents a political party which has its own agenda. The relationship between the party and its members is symbiotic: each member gets free advertising by being affiliated with a party and, in turn, he/she must agree to support the party's stance on certain issues. So the strength of the party's platform is represented by the fraction of seats it holds in parliament. To represent the public's opinion this fraction should be close to the total fraction of people who voted for the party -- proportional representation.
However, it frequently occurs that these two numbers are drastically different. To see how this occur, imagine the following scenario: there are three parties and in every single riding party A wins 40% of the vote and parties B and C each receive 30% of the vote. Then A will hold every seat, with complete control over the government, even though they only 40% of the public voted for them. Of course, that's an idealization. But this kind of thing happens frequently in Canadian politics--in the 2000 federal election the Liberal party won 57% of the seats even though only 41% of the votes cast were for them.
I think a good system must have these three properties:
- Regional representation. A member of parliament should be a representative of a particular riding, standing up for the region's ideals and needs. And the member should be elected directly by the riding's constituents.
- Proportional representation. Overall, the proportion of the seats held by each party should be on par with the popular vote (proportion of votes for said party).
- Each voter should cast a single vote. Some reform suggestions have voters casting multiple votes or ranking their preferences. While they may lead to fairer systems, we've already started down the road of "one person, one vote" so I think it would be a bad idea to change it now.
The Solution
Is it possible to reconcile all three of these requirements? I believe so. Here's my idea:
The Two Seat System
- Each riding is represented by two seats in parliament.
- If a candidate wins more than half of the votes cast, he/she wins both seats (ie. represents two votes in parliament).
- Otherwise, the top two candidates in each riding are elected.
Simple, no?
Everybody still has one vote, satisfying condition (3), and we still have regional representation. The only uncertain point is whether this improves proportional representation.
To show that it does, I have collected data from recent Canadian elections and compared the actual outcomes to what would have happened if the above rules had been followed.
I calculated the popular vote (nationwide) for each party as well as the percentage of seats won by each. Then I simulated the Two Seat System by pretending to send the top 2 candidates from each riding to office if the actual winner didn't get more than half of the votes. Of course this doubles the number of seats so I will work in percentages. The results are summarized in the table below. (Click on an election to get detailed results.)
Caveats
Of course, this new approach raises a few issues. For instance, to preserve the total number of seats the neighbouring ridings would have to merge. This means somewhat reduced regional representation because the politician now represents a larger region.
This new system would also change the nature of strategic voting in ways I couldn't possibly predict. However, I suspect it will alleviate the problem of vote splitting somewhat. For example, imagine a scenario where 60% of the voters in a riding lean toward two parties, A and B, with similar positions. The remaining 40% prefer party C with a strongly conflicting platform. Come election it frequently occurs that parties A and B split the vote (let's say 35% and 25%) so that party C wins. With the Two Seat System we can expect a candidate from party A or B to win the second seat, thereby more accurately representing the interests of the riding. The whole point is that the only way a single party can win both seats is if it has the support of over half the voters.
Surprisingly, I think this can even be extended to a two-party election. The trick would be to also count, for example, blank ballots. If you didn't count them then the Two Seat System becomes meaningless, because the winner by definition has over half of the votes. But if you do count blank ballots then disillusioned voters could cast a protest vote by dropping a blank ballot in the box. If there were enough that the winner didn't have over half the votes then both candidates would be elected, with one seat each, to keep each other in check.
One problem which may (rarely) occur is a perfect balance of power. By definition there would be an even number of seats in parliament (currently, I think there are always an odd number) and therefore a possibility that a legislative vote be split exactly 50/50. But I'm sure that possibility already exists (say, if some members abstain from voting) and they've already devised a way to break the symmetry.