On Ranked Ballots (aka Alternative Vote): A Conversation with Michael MacKenzie

This is another “good parts version” of a conversation I had while writing a recent op-ed—this time my analysis of the Ontario election and its potential implications for the cause of democratic reform in Canada.

Initial pleasantries and irrelevant bits—irrelevantries?—have been deleted, and I’ve done some editing for flow. I provide clarifying comments as necessary in brackets. [Not parentheses.] See? Just like that.

My email interlocutor was Michael MacKenzie, a former grad student at UBC now at Harvard’s Ash Center for Democratic Governance and Innovation, where he specializes in what you might call applied democratic theory. You can read about his current project, Participedia, here.


Michael MacKenzie:Nicely done on the article. Importantly, as you point out, AV [Alternative Vote, another name for ranked ballot/instant runoff systems] is not a proportional system – STV [Single Transferable Vote—the system recommended by the BC Citizens’ Assembly, and voted on in the province’s two referenda] is.

They are two very different systems that both use preferential voting. The AV system is very simple, and it works as you explain in your article. STV is more complicated (but not THAT complicated) and it is designed to produce a result in which the percentage of the vote gained by each party in the election is near equal to the percentage of seats the party gets.

For those who support proportional representation, however, an AV system would not get them what they want.

One advantage of the AV system (and it may be a very important advantage, depending on your perspective) is that candidates have to obtain a majority of the vote before they are elected. This, it seems to me, is important when we are selecting a mayor or a president because in these positions one person is supposed to represent everyone and being elected by a majority can help enhance one’s political legitimacy, but it’s not clear to me how important AV is when we are electing MPs or councillors – many of whom win by majority margins (or near majority margins) in their districts in any case. Also it is usually (although not always the case) that the person with a plurality of votes on the first count wins the seat on subsequent counts in an AV system.

It can be difficult to say just how large the change in outcomes might be following a shift from our current first-past-the-post single member plurality system (SMP) to an AV system. I’m sure they did simulations for the BCSTV system. You can check the technical documents posted online, on the BCCA [BC Citizens Assembly] website. [Materials have been archived here.]

The difficulty with running simulations is that people vote differently under different electoral systems. As you note in your piece, AV helps prevent (or reduce) strategic voting. So people tend to vote differently when faced with an SMP election, and it is therefore difficult to simulate results for different systems based on actual election results. Although I remember reading a piece that used polling data about voters’ second choices to simulate AV results after an SMP election. I can’t remember where that was published. [Here’s a relevant, but unfortunately gated and not-Canadian focused paper on the subject.]

It’s an interesting debate as to whether AV would advance the cause for electoral reform. I think you’re right: it would advance the cause in some respects—but it is rightly difficult for those who support PR to promote a system that is not PR simply because it is different than the one we have.

Also, it’s worth mentioning that BC, Alberta, Manitoba, and a few other jurisdictions have used ranked ballots in the past. I think the system was dropped in BC because it was introduced for partisan reasons – through a hyper-political process – and thus never gained legitimacy.

In contrast, the BCCA was designed to circumvent the partisanship that will typically cloud discussions of electoral reform.



SP: Yes, I looked briefly into both Manitoba and BC cases; seems like in both, partisanship played a role in the system’s demise. This is not surprising; if Canadian politics in 2014 has taught us nothing else, it’s that partisanship and electoral reform REALLY don’t mix well.

In the present context [of municipal reform], the current broad basis of support for AV in Toronto would seem to allay that concern.

Personally, I tend to increasingly lean back on what you might call an “Aristotelian” approach to democratic reform–i.e. it’s a practice, like other political endeavours, one that Canadians nowadays quite frankly suck at. AV is not perfect, but it’s better than the current system. Seeing it come off well ought to embolden reformers, and open the minds of voters, to the broader possibilities out there.

The next time reform comes up, at whatever the level, and whatever the proposed new system, Canadians will be a little more comfortable with considering a change simply because they’ve seen one happen (and in all likelihood work out just fine) in the recent past here in Canada. Canadians are a timid lot when it comes to democratic reform, so I think a little learning-by-doing is an important part of any broader reform movement.

As for the distortionary effects on outcome created by AV, from what I understand, there are two major situations in which AV might produce an “unsatisfactory” democratic outcome by some definition.

We saw one in the Alberta Conservative party leadership election that Stelmach won, which used an actual, rather than “instant” runoff. Stelmach won the leadership (and therefore premiership) essentially as 3rd choice candidate who came up the middle between Dinning and Morton to win. Redford’s situation was a little different, but also produced what is now seen as a disappointing outcome.

Basically, the argument is that AV risks promoting weak compromise candidates over strong factional leaders.

In the second general “failure” situation, a strong second choice of a clear majority of voters doesn’t earn enough first place votes to get through to subsequent rounds. The result is thus sub-optimal in the sense that it doesn’t consider the full ranking of preferences of all voters in deciding who to eliminate.

More generally, I’m curious whether you think AV has any positive effect on proportionality. I had assumed that it could be seen, in effect, as a partial remedy, since no single candidate could get in without at least half-hearted support from a majority at the district level. That ought to push the “final round” vote numbers for parties closer to their actual seat totals; at the same time, such systems  still work against non-geographically based fringe parties like the Greens. So just how much does AV count for on the proportionality front? I suspect one’s answer depends to an extent on the degree to which one accepts the argument that Canada’s elections are really 308 (soon to be 338) mini elections, rather than one national one.


MM: Yes. Those are possible (unattractive) consequences of AV.

On the question of whether AV produces some greater degree of proportionality, I think the answer is no – at least substantively.

1. When we calculate disproportionality in preferential systems we usually use first preference votes, which is appropriate because in a PR system you only ever cast your (supposed) first preference vote.

2. In an AV system it is still possible for a party to win 100 percent of the seats with 50%+1 of the final vote in each district. In this scenario nearly half of the votes are still “wasted”.


[Time passes]

MM: Your thoughts on whether (or how much) an AV election might improve proportionality got me thinking. I’ve created a scenario to illustrate that AV does not substantively improve proportionality:

AV Election Scenario:

Simplified election: Four Parties. 5 Districts. 100 votes in each district. Showing: Round one results and final count (with subsequent preferences redistributed).

AV election MM scenario


As you surmised, if proportionality is calculated using the final vote, disproportionality decreases somewhat. Using the Gallagher Index the disproportionality score for Round One = 29.30. The disproportionality score for the Final Round = 25.36. This is still very high. A perfectly proportional system would achieve a score of zero; many proportional systems regularly achieve Gallagher Index scores of less than 2.

Party B’s results are slightly more proportional after the final round. Party B wins 60 percent of the seats (i.e. 3 seats) with 28 percent of the vote in the first round, but 60 percent of the seats with 39 percent of the vote in the final round. Party D’s results are also slightly more proportional – i.e. they win their one seat with 12 percent of the vote, instead of with 29.2 percent.

Party A’s disproportionality score is not affected much. Party C, however, actually does worse in the final round, winning more of the popular vote once subsequent preferences have been distributed, but still winning no seats over all.

More importantly, it is possible to win a majority of the seats with a majority of the votes in each winning district without winning a majority of the overall vote. If this is case the legitimacy problem remains: it is possible to win 100 percent of the power with less than 50 percent of the votes in an AV election.

This scenario does not make AV look like a very proportional system.



SP: Nice (and quick) work! I think you’ve convinced me on the substantive point: whatever gains AV produces in terms of proportionality are not significant in comparison with the gap that exists between it and SMP on the one hand, and the various proportional systems on the other. It’s on the wrong side of that gap.

There might be some secondary effects if you were to iterate elections, since one interesting thing about AV is the signalling quality of even unsuccessful votes. Even if they fail to capture seats, new parties can get a boost simply by showing better than expected on first preferences in an election. That can heighten visibility and credibility in subsequent votes. In your example, if Party C were a political upstart, they might have a better shot in district 5 in a subsequent election given how well they did here.

In addition, if per-vote subsidies were at attached to first preferences, one could build in a kind of monetary proportionality. All that said however, I don’t think it changes the basic conclusion that AV is not proportional.

Accordingly, any nice things I’d still want to say about AV relate to other potentially desirable qualities. In particular, it seems like it’s a nice stepping-stone in the present Canadian context. Most significantly it is better than, but also very similar to, what we have now. That seems to make it somehow “safer” for Canadians to talk about and potentially embrace. I’ve become quite convinced that most of the democratic virtues depend heavily on practice, and that democratic reform is no exception. Seeing one change come off will make it easier for Canadians to think about other, bolder changes.

I’m sure that if and when Canadians (or a sub group thereof) ever decide to try on a more radical shift to STV or MMP (well, the former, at least. I’m somewhat less confident in the latter at present due to the role that parties play in putting together lists, which really do seem to be an unpopular approach here, due to the total centralization of nominating power) they would come to embrace it. AV is the idea with traction now, however. There are some other qualities of AV that are still worth mentioning, I think, mostly related to democratic values other than representativeness. The fact that politicians have to build broad coalitions is nice, and something you wouldn’t necessarily see in strict proportionality. It essentially creates an institutional incentive to listen and compromise.

I suspect (but have not studied and therefore cannot prove) that ugly things like xenophobia find easier expression in proportional systems, where politicians keep their jobs primarily by caring only about their own. The UK experience with UKIP provides an interesting case study in this regard. UKIP has flourished in the proportional European elections, but no such party has a foothold in the non-proportional HoC. Undoubtedly xenophobes are voting at both levels, but domestically, their voice is folded in with (and therefore muted by) the Conservative party, whose leaders have to look beyond narrow confines even in the context of a FPTP election.

Another element of AV—one I think it shares with STV—is the way it forces voters to identify with parties other than their own. Again, no evidence, but I wonder whether and to what extent this practice tends to weaken effects of identity politics and tribalism. If you’re compelled to pick several horses in a given race, you may be less inclined to give all your affection to just one

It’s certainly a topic for another day, but it’s just as certainly worth thinking through how best to weight the various potentially desirable elements of a democratic system, both in general and within the particularities and peculiarities of the Canadian system.

For now though, thanks again for the informative conversation.

MM: My pleasure. Perhaps we can revisit some of these latter questions at a later date.

photo credit: orarewedancer via photopin cc

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s