Electoral reform needs better PR (Part I)

by on


What better way is there to spend the aftermath of an election than to wring one’s hands about the difference between the parties’ seat count and the popular vote they received? It is, after all, as pure a form of ineffectual handwringing as one could wish: any form of proportional representation (PR) seems as far away as ever.1If any electoral reform takes place, it likely to be to institute ranked ballots with the current electoral boundaries (sometimes called Alternate Vote (AV)). This reform is preferred by Liberal Leader Justin Trudeau and seems likely to benefit his party.

In the next three posts, I’m going to zoom past the despair and get to even more ambitious reforms in the pollyannaish hope that a bigger pot of gold will entice more people to chase the PR rainbow. This first Part will introduce an electoral system that I call “Locally Proportional Representation” (LPR)2There may be another, more popular name, that I don’t know. When I first concocted this scheme c. 2005, I didn’t see another name, so I’m sticking with my original moniker. The closest thing on Wikipedia is Interactive Representation.. Part II will focus on riding boundaries outside the geographic plane, as an alternative or extended law reform plan. Finally, Part III will jump out of the castle in the sky and discuss whether courts have a role to play.

If you’re reading this, I’m going to assume you know about the existing debate on proportional representation (i.e., why many people think the current First Past the Post (FPTP) system isn’t fair). I’m also going to assume you know about the prominent alternatives to FPTP (and why they’re not embraced — beyond the basic reason that proportional representation is likely to lead to minority governments3I think minority governments are great, but many people disagree). Specifically:

If this is the first time you’re seeing these terms, then this post might be a bit abstruse. I’ve linked the Wikipedia summaries above and Fair Vote Canada has some primers. But all you really need to know is that fewer people voted for Liberal candidates (5.5M Canadians) in the election last week than Conservative candidates (5.7M), but the Liberals got more seats in Parliament (159) than the Conservatives (119). And the same goes for the Bloc (1.3M votes, netting 33 seats) and the NDP (3M votes, netting 25 seats). And that kind of mismatch isn’t weird in Canada. If that seems wrong to you, then you might be interested in LPR.


The key feature of LPR is that elected members do not have equal votes in the legislature. Rather, each elected legislator votes in the legislature according to the number of votes they received in their riding in the last election.4Of course, this will require vote-tabulating software in the legislature, rather than simply a “show of hands”. Such an impediment should be trivial to overcome. LPR requires multiple members to be elected in every riding in order to work. It works best with a ranked ballot and redistribution of votes like in AV, and in Canada it would seem appropriate to have three members per riding.5Three members per riding seems right to me, since there are three major parties in most regions of the country, and more members might lead to unduly large ridings or unduly many members. I hold this view relatively weakly, and there may be regions of the country where different numbers of representatives are appropriate. I call this “locally proportional representation” because the strength of each legislator’s vote in the legislature is proportional to the number of voters who voted for them locally in the previous election.

The key feature of LPR is that each member “carries” with them the votes of the people who voted for them when they vote in the legislature. LPR thus creates proportionality in the strength of votes cast in the legislature, rather than in the number of members (as in other proportionality systems). A legislator who received 100,000 votes in one riding would have more power in the House than two legislators who each received only 40,000 votes.

Let’s try an example. There are two ridings in our fictional country, Schol-Mance and Brockton Bay. Schol-Mance has 150,000 voters and Brockton Bay has 80,000 voters. Each has 4 candidates running, for parties L, C, D, and P, and each riding elects 3 members.

In Schol-Mance, the first preference votes are as below. Because Pack came in last, he’s eliminated and his votes are redistributed. In this example, 15,000 people who ranked Pack first ranked Chloe second, so she gets an extra 15,000 votes, while Lorion and Del pick up 2,000 and 3,000 extra, respectively. The result is that Lorion, Chloe, and Del are elected. Lorion’s votes are worth 72,000 points in the legislative assembly; Chloe’s are worth 48,000, and Del’s are worth 30,000.

Candidate (Party)First Preference VotesRedistributed votes
Lorion Lake (L)70,00072,000 (+2,000)
Chloe Crasmussen (C)33,00048,000 (+15,000)
Del Diggins (D)27,00030,000 (+3,000)
Pack Pesting (P)20,000 (eliminated)

In Brockton Bay, first preference votes are as below. Dessica, as the candidate with the fewest first-preference votes, is eliminated, and her votes are redistributed. In this example, 4,000 people who had Dessica as their first choice had Lory as their second preference, and 1,000 had Cody instead. The result is that Lory, Cody, and Prichard are elected. Lory’s votes are worth 19,000 points in the house; Cody’s are worth 33,000, and Prichard’s are worth 28,000.

Candidate (Party)First Preference VotesRedistributed votes
Lory Listner (L)15,00019,000 (+4,000)
Cody Calvert (C)32,00033,000 (+1,000)
Dessica Damada (D)5,000 (eliminated)
Prichard Panders (P)28,00028,000 (+ 0)
Brockton Bay

The overall result of the election, with these two ridings, would be that the L party gets two candidates elected (Lorion and Lory), who together have 91,000 points; the C party gets two candidates elected (Chloe and Cody), who together have 81,000 points; the D party gets one candidate elected (Del), who has 30,000 points; and the P party gets one candidate elected (Prichard), who has 28,000 points. The overall distribution of power in the legislature suggests an L minority government.

LLorion, Lory72,000+19,000=91,00040%
CChloe, Cody48,000+33,000= 81,00035%
Total6 candidates230,000 votes100%
Party-level results

The pitch

The attraction of LPR is that it means every voter’s vote counts, and counts as equally as possible, regardless of the district boundaries. Under FPTP, almost every vote will either be for a candidate who loses and isn’t elected or for a candidate who wins and is elected by a margin of more than one vote. Either way, the vote had no effect on the “outcome” of the election. Other systems like AV, MMP, or STV, similarly compress the “signal” of many votes into a much more coarse signal: the identity of one representative. Under LPR, the signal is kept as high definition as possible. It’s the HDTV of electoral systems.6Or, more accurately, the TIFF or PNG to FPTP’s 3-bit GIF.

Making every vote count changes the dynamics of electoral campaigns. If every vote is equally valuable, what are today “safe seats” can’t be taken for granted, and nor can they be abandoned as hopeless. A voter who is wavering between the Conservatives and Liberals is as worth courting if they live in Red Deer-Mountain View (36,596-vote difference between those parties) as if they live in Fredericton (502-vote difference). So long as a candidate can reach the magic threshold of being one of the top 3 candidates, a vote for them is valuable.

LPR similarly changes how representation works in the legislature. Each voter in a riding would have multiple possible representatives whom they could approach to get recourse from the state,7The so-called “Ombudsman role”: see Reference re Prov Electoral Boundaries (Sask), [1991] 2 SCR 158 all of whom are competing for her vote. It also means that almost all voters will have someone in the legislature that they voted for personally — someone whom they feel they can and ought to hold to account. And, as with STV and other multi-member systems, this representation will change the narratives we tell about places and communities in Canada. There are conservatives in downtown Toronto; there are liberals in rural Alberta. Having representatives of multiple perspectives in the legislature will help combat such reductive stereotypes.

The final important feature of LPR is that it separates the three primary functions of members. One function of members is to vote in the legislature and so directly affect legislation. A second function is to speak in the legislature, and so indirectly affect legislation through discussion. The third function is the ombudsman function: providing an avenue for seeking recourse from the state. Under the existing systems, these three functions are combined in individual members.

Combining all three functions creates certain difficulties. In terms of vote in the legislature, every elector is supposed to be equal, which points toward equal-size ridings. But, as McLachlin J recognized for the majority in Reference re Prov Electoral Boundaries (Sask), [1991] 2 SCR 158, the other functions may push toward unequal riding sizes. As she writes, “rural ridings are harder to serve because of difficulty in transport and communications [and] rural voters make greater demands on their elected representatives”; and further deviations from equal vote may be required to provide “minority representation … [and] ensure that our legislative assemblies effectively represent the diversity of our social mosaic”. LPR resolves the tensions between these functions by allowing votes to be equal even if representatives serve ridings of different sizes. I’ll discuss the further implications of this in Part II and Part III.

The caveats

I think LPR is great — but (as with any electoral system) adopting it may have some non-obvious consequences. Some of these consequences affect candidates; others affect parties; others affect turnout.

The impact on candidates

Under LPR, incumbent candidates who perform poorly in the next election are unlikely to be “voted out of office”. Rather, they may be re-elected with a lower vote count (and so have less power to influence votes in the house). That may or may not be a good thing. Candidates in “safe” ridings are likely to be more responsive than under FPTP, because they have something to lose; candidates in contested ridings may be less responsive than under FPTP, because they are not afraid of losing their job, only a fraction of their vote in the legislature. Overall, this tradeoff seems appropriate to me: elections can be arbitrary, and making someone’s employment dependent on minor vicissitudes possibly prevents good candidates from coming forward.

But there is also a second effect: today, losing an election is an obvious signal for a person to leave a political career and thereby free up a spot to run for that party. LPR (and most PR systems) would replace the unambiguous binary signal of “win” or “loss” with a more complex signal of “gained some votes” or “lost some votes”. There may not be a time when everyone agrees that an incumbent candidate has lost their touch and should move on. If parties are insufficiently cut-throat with their own candidates, candidates may persist and prevent better candidates from coming forward.

LPR’s solution to this is to allow candidates to demonstrate their bona fides by running as Independents or for another party. Ranked ballots significantly reduce the risk of vote-splitting, so that ought not deter such candidacies. Moreover, with the threshold for becoming an MP so much lower, an arriviste might have a better chance. Finally, there are direct incentives for a party to run a better candidate: more power in the legislature. Today, someone who is good enough to keep winning is good enough for the party; under LPR, a party would prefer someone who wins by more.

The impact on parties

The same effect has consequences at the party level: even when a party does poorly, it is likely to have the same number of MPs. A party in opposition will not shrink to a few survivors; a party in government will not have a sudden surplus of ambitious MPs to keep happy. They will simply have a different weight to each of their votes. I don’t feel equipped to speculate about what the further consequences of this would be, or if they’d be desirable or unfortunate. If you’ve some insight in this, please get in touch!

Another impact is that LPR, as described here, will continue to underrepresent minor parties whose support is so widely dispersed across ridings such that they never meet the threshold of being one of the top 3 parties for getting a seat. LPR requires parties to do well in individual ridings to get a seat. Other forms of PR (such as list-based proportionality or MMP), reward parties who do not do well in any particular riding. The Greens would do better under those systems than under LPR (but probably better under LPR than forms of FPTP).8There is a further option: a hybrid of MMP and LPR. For this hybrid, LPR would use unranked ballots. The top candidates in each riding would be elected as before, but their power would be weighted solely according to voters’ first preferences. All votes that are not for an elected local candidate would be pooled and assigned to one national (or provincial) member for each party. The result would be a legislature with mostly local representative, but five or so “national” members, each of whom carries the vote for all “unelected” members of that party. This hybrid option is interesting, and would ensure representation of minor parties, but it would probably significantly strengthen parties’ control over their members versus plain LPR.

But issues embraced by the Green Party might be better represented: other parties would have an incentive to be the second preference of Green voters, even in ridings where that party was poised to “win” or “lose” the riding by a significant margin under FPTP. When every vote matters equally, there’s reason for major parties to care about minor party voters.

The impact on turnout

LPR seems likely to have a complicated, but overall positive effect on voter turnout. LPR seems likely to give renewed meaning to voting and to foster a greater sense of control and ownership by voters. This should drive voting up. Candidates, also, are each individually advantaged by higher turnout in their riding: they should be seeking to maximize turnout because it maximizes their influence in the House. Parties, however, have conflicting incentives. Parties who are strong in a riding would prefer high turnout; parties who are weak in that riding, low turnout.


Locally Proportional Representation (LPR) is a proposed electoral system where each legislator’s vote in the legislature is weighted according to the number of votes the legislator received in the election. It works best with ridings that elect multiple members using ranked ballots. LPR promises to strengthen the connection between constituents and their members and ensure that every vote is seen to count, and count equally. In so doing, it allows ridings to be of unequal size, which could be very helpful for achieving other objectives. I’ll address that topic further in Part II and Part III.