On Canada’s Voting Reform Survey: Part 1

13 min readSep 11, 2016

The ERRE “Special Committee on Electoral Reform” has put out a survey (“E-CONSULTATION ON ELECTORAL REFORM”) so that Canadians can express their opinions about electoral reform. If you are Canadian, I’d urge you to read this article and then go answer that survey.

I’m writing this because I think that, for the most part, the survey is confused, confusing, and poorly-designed. Still, it’s great that the government is asking Canadians for their opinion. In this article I will go over the survey and explain my beefs with it — and along the way I’ll teach you about several of the many nifty voting systems we could consider for Canada.

The survey’s problems start right on the first page.

First, while this is certainly a list of characteristics of voting systems, I would argue it’s not a list of “important” characteristics. For instance, “number of candidates per constituency” is not even a meaningful concept in some voting systems. Meanwhile, voting systems that are as unrelated to each other as FPTP and SDR don’t have a fixed percentage “threshold for determining winners”. Likewise, Condorcet and IRV may have identical ballots, but they behave differently in close races. This is the first example of the survey being confused rather than merely confusing: it doesn’t seem to have a good grasp of the whole range of possible voting systems, or what the important differences between them are.

The important characteristics that differentiate electoral systems from one another are attributes such as “which mathematical properties does it meet?”, “is it ‘proportional’ in any sense?”, “is it straightforward for voters to use?”, “can voters easily understand the outcome of the election?”, “does it enable or impair voter choice?” and above all, on the day after the election, “does it maximize the extent to which Canadians are pleased with the outcome?” (there is actually a mathematical concept for this, called Bayesian Regret.) Characteristics are important if they tell us how well a system works, not what the ballot looks like or even how many steps determine the winner — although there are some systems that are relatively simple and work quite well (or might work well, if we had the courage to try them). Some important characteristics, such as “what kind of people would run for office under this system?” are hard to know with certainty without trying a system out for a couple of decades (it saddens me to think that we will never try certain ideas on the sole basis that they’ve never been tried before, but I digress).

Mathematicians have proven that it’s impossible to have a voting system that meets all seemingly-desirable properties of a voting system, and because of that, even people who are well-informed can disagree about which system is best.

On the other hand, I don’t think anyone who is well-informed can think that the current First-Past-The-Post (a.k.a. single plurality) is the best. Dozens of voting systems have been designed to replace it, and I can say unequivocally that FPTP is the worst system. Since you only get one vote in only one district, it gathers the least possible amount of information from you. Occasionally, a candidate runs unopposed, reducing voter choice to zero! Since it is non-proportional and often causes “strategic” voting (a.k.a. voting for the “lesser of two evils”), it’s also the least democratic. I would prefer literally every other system I’ve ever seen, and I’ve pretty much looked at them all. Range Voting, Approval voting, the Condorcet methods, IRV, Borda, Party-list PR, Open-list PR, Mixed-member PR, STV, P3, Direct Representation, and my own Simple Direct Representation… they’re all better. If you have the choice to switch from First Past The Post to any other democratic system, always say yes no matter what the alternative!

Anyway, on with the survey!

I disagree with these groupings, too. Instead of “plurality or majority systems” and “proportional representation systems”, it’s better to classify them as “single-winner systems” and “multiple-winner systems”.

Single-winner systems are designed and optimized for situations in which there are single winners, such as elections for mayor or party leadership. These include Range Voting, Approval voting, the Condorcet methods, IRV, Borda, runoff voting (multi-round FPTP) and, of course, FPTP.
Multi-winner systems are designed for multi-winner situations. Multi-winner systems include Party-list PR, Open-list PR, Mixed-member PR, STV, Dion’s P3, Direct Representation, and my own Simple Direct Representation. Multi-winner systems are usually “proportional”, but don’t have to be.

Single-winner systems are only sensible for use in single-winner elections, such as for mayors and leaders. By nature, when any single-winner system is used to pick many winners (by dividing the land into “districts” or “ridings”), it always “wastes” votes, it enforces a geographic restriction, and it cannot produce a “proportional” outcome. I’ll explain this in more detail below.

The next few pages of the survey seem reasonable enough, except that there’s a textbox with a 150-character limit that freezes up completely when you reach 150 characters (workaround: reload the page to clear the box. Only works before you click Next.)

Let’s skip ahead to:

Goals of a Voting System

“This section seeks to gauge your opinion on what an electoral system should accomplish. Please indicate how important each of the following elements are to you:”

To know your local representative: in a survey about voting systems, this question makes little sense. Electoral systems do not give knowledge to voters, their purpose is to gather information from voters. Also, the question assumes that “local” representatives exist, even though some voting systems do not include that concept. Better questions would have been “How important is it to be represented by an specific MP?” and “Do you prefer that everyone be required to vote for a local representative, whether they want to or not?”
To be affiliated with a political party: again, it’s a strange question. Voting systems do not “affiliate” voters with parties; voters affiliate themselves with parties outside the electoral process. That said, some voting systems don’t allow you to vote for parties while other voting systems don’t allow you to vote for individual candidates. They could have instead asked a pair of questions: “How important is it that the voting system allow you to express a preference for (1) a party and (2) an individual?”
To be aware of political party platforms: what does this have to do with voting systems? Ballots don’t exist to raise awareness of party platforms. Are they suggesting ballots should try to inform voters about stuff? I might agree with that, if I knew what they were talking about.

“Canada’s electoral system should favour the following outcome:”

One political party holds a majority of seats in Parliament and is able to implement its campaign platform. STRONGLY DISAGREE: the voting system shouldn’t be biased against how voters actually voted, nor should it be biased against smaller parties. A single party should get a majority if that’s what the voters asked for, but First Past the Post tends to produce false majorities (e.g. there have been 3 majority governments since 2000, but each one only got between 38% and 41% of the popular vote).
No single political party holds the majority of seats in Parliament, thereby increasing the likelihood that political parties will work together to pass legislation. WEAKLY AGREE. Minority governments should not be a goal of a system, but it is a sensible outcome when the citizens disagree about who should lead them.

Current electoral system

The current electoral system adequately reflects voters’ intentions: STRONGLY DISAGREE. The current system only allows a single choice, which encourages strategic voting, it only allows you to choose a local candidate, and it doesn’t ask which party you support.
If I vote for a candidate in my riding who does not win, my vote is wasted. STRONGLY AGREE. I’ll say more on this below.

First Past the Post

STRONGLY DISAGREE: this question wrongly assumes that I want a system based on “ridings” in the first place, or that I want a system based on the traditional concept of “votes”.

Runoff Voting & Preferential Ballots

These questions are designed specifically for runoff voting. Runoff voting, whether it’s IRV (a.k.a. Alternative Vote) or Old-Fashioned Runoff, is based on a somewhat silly premise: “First-Past-The-Post is a flawed system, so we can fix it by having a series of First-Past-The-Post elections!” This strategy of “doubling down” on a flawed system has never made sense to me. I do think IRV is better than FPTP, but that it isn’t justified by anything more than tradition.

IRV uses a preferential ballot like the one above, but IRV is not the same as a preferential ballot. Preferential ballot, as you might guess, is a kind of ballot, whereas IRV is a method for determining the winner after the ballots are collected. I argue that if you’re going to have a preferential ballot, the winner should be based on the Condorcet criterion:

The Condorcet candidate (aka Condorcet winner) is the person who would win a two-candidate election against each of the other candidates using a plurality vote. For a set of candidates, the Condorcet winner is always the same regardless of the voting system in question. A voting system satisfies the Condorcet criterion if it always chooses the Condorcet winner when one exists.

IRV doesn’t meet this criterion; in fact it behaves in an unstable way in close races. This instability is demonstrated by the fact that it doesn’t meet the monotonicity criterion:

A ranked voting system is monotonic if it is neither possible to prevent the election of a candidate by ranking them higher on some of the ballots, nor possible to elect an otherwise unelected candidate by ranking them lower on some of the ballots (while nothing else is altered on any ballot).

See here for an example where ranking a candidate higher on certain ballots will make him lose.

Range voting

At first, I loved the Condorcet methods, but I now prefer Range Voting, in which you rank each candidate on a scale from 0 to 10 (or 0 to 5, or 1 to 10, or whatever) and the candidate with the best average score wins.

In other words, Range voting is exactly the same system used in most of the survey itself!

A ballot for Range Voting (a.k.a. Score Voting)

Here’s why I prefer Range over Condorcet and IRV:

Range voting lets you express how strongly you prefer one candidate over another, while Condorcet and IRV ballots don’t. Also, due to a misguided desire for “simplicity”, officials may deploy a Condorcet ballot that doesn’t allow voters to express an equal preference for two candidates, nor to abstain from judging a given candidate. So if you have an equal preference, you may be forced to lie and make up a preference that doesn’t exist. Range voting always allows equal preferences (or at least, it would be silly not to allow them). You don’t have to exercise your freedom to express nuanced preferences — you could rank candidates with only minimum or maximum scores, which is the Range voting equivalent of “strategic voting”. But I want that freedom.
Range voting is easier for voters when there are more than four candidates running, because they can simply consider how much they like each candidate in isolation; they need not (directly) construct a ranked order.
After an election, Condorcet and IRV ballots do not produce a numeric result that humans can easily understand; there is no single, simple answer to the question “by what margin did candidate A win over candidate B?” The lack of a single “best” measurement allows different observers to choose different measurement methods according to their political goals, leaving voters confused or misled. To drive this point home, there is not one Condorcet voting system but several, each of which breaks cycles in a different way (changing the winner in rare cases). Similarly, the IRV winner may not be the Condorcet winner. Clearly, there is no accepted way to pick the winner of a preferential ballot, only a legislated way. In contrast, Range voting systems give a numeric score for each candidate (the average) so the result is obvious and easy to understand.

As I mentioned, it’s mathematically impossible for a single voting system to fulfill all desirable criteria that have been proposed, but I think even when Range voting fails a criterion, it does so in a sensible way. For example, Range voting fails the “majority criterion”, which states “If more than 50% of voters consider a particular candidate to be the absolute best choice, then that candidate should win”. That sounds like a sensible rule, so why doesn’t Range voting do it? Well, suppose 47% of the people vote something like this:

And suppose 53% of the people vote more like this:

If we just go by “top choice” votes, Alvin Anderson should win. But looking at these ballots, it’s obvious that the 47% minority dislike Alvin more than the Alvin supporters dislike Bob. The Range voting method awards the victory to Bob with his average of 3.9, rather than Alvin with his average of 3.1. While Alvin’s supporters are disappointed, they are less unhappy than Bob’s supporters would have been if Alvin had won. Clearly, Alvin supporters don’t have a big problem with Bob. If they did, they would have ranked him “1” instead of “3”. Thus, Range voting has produced the best outcome in terms of voter happiness. Bob is probably the kind of candidate that knows how to get along with everybody, unlike Alvin who seems to rub 47% of the people the wrong way. (You could argue that maybe Bob’s supporters just “conspired” to give Alvin “1”s, but if they tried that, Alvin’s supporters would surely find out about the plan and retaliate by giving Bob “1”s, which could shift the winner to Alvin. In that case, Range voting could fulfill the majority criterion without improving the outcome in terms of voter happiness.)

But you know what? Forget everything I just told you about IRV, Condorcet, and Range voting, because all of that misses the forest for the trees. The problem is this:

Single-winner systems don’t pick multiple winners

That’s what multiple-winner systems are for.

FPTP, IRV, Condorcet, and Range voting were designed to pick only a single winner. To extend this to multiple winners despite their fundamental nature, officials typically divide up the country into arbitrary chunks called “districts”. Sometimes, an impartial commission picks the district boundaries, while other times the party in power picks the boundaries in an effort to increase their own power, a practice known as gerrymandering.

https://www.youtube.com/watch?v=6Ls5j5iz2eA

If you use a single-winner system to pick multiple winners, here’s what happens:

Many of the votes are “wasted”, in two different ways. First, the opinions of people who voted against the winner in a district are ignored. Those people “lose” the election and have no effect on the outcome. In Canada this “wastage” is especially large, since candidates often win on less than 40% of the popular vote, with over 60% of the people in the district voting against them. But what’s more, in a sense your vote was wasted even if your preferred candidate won. For example, if the winner wins by 1002 votes, what would have happened if you just stayed home? Then the winner would have still won by 1001 votes and the outcome would have been the same. Your individual vote only really counts if it might have changed the outcome, which almost never happens. Some variation of this analysis applies, to some extent, to all the single-winner systems.
It enforces a geographic restriction: you’re not allowed to vote for someone outside your riding. If your favorite candidate lives in the North side of the city but you live on the South side, screw you! You can’t vote for him, even though if he wins, he won’t live in the city at all, he’ll live in Ottawa.
It cannot produce a “proportional” outcome: the overall makeup of the legislature is only weakly related to the proportion of votes cast for each party, and the results are vulnerable to gerrymandering.

In Part 2, I’ll explain how single-winner systems have biases that have nothing to do with democracy, and why they inherently cannot produce proportional outcomes.

Like this article? Please follow me!