Electoral Reform needs better PR (Part II)

by on

[This is the second of a three part post. In Part I, I proposed an alternative electoral system to provide locally proportional representation (LPR). LPR works by weighting legislators’ votes in the legislature according to the number of votes they received in the election. In Part III, I’ll discuss the courts’ role in all this. In this post, I’m going to focus on riding boundary-drawing.]


In Canada, “electoral reform” mostly means reforms to how votes are cast (e.g., with a ranked ballot) or tabulated (for proportional systems). In the United State, this discursive space seems instead to be occupied by discussion of electoral districting — that is, drawing the boundaries of ridings.1I’m going to stick with the American term “districting” because the alternative is “ridinging” and writing that once was enough. The topics attract a similar flavour of wonkery, and there are deep parallels between the two discussions.

Take an election (any election). If you travelled back in time with a blank map, a fine-tipped pen, and no other restrictions, you could carve up (that is, gerrymander) the districts such that the same voters casting the same votes could result in a very different set of elected representatives. Specifically, travel back in time to last week. If the riding map were different, could the overall seat count have tracked the national party vote? Of course — you might do it with ridings of unequal size (just worse than there are today), or ridings that spanned provinces, but it’s definitely possible. Canadians tend to focus on First Past the Post (FPTP) as the source of mismatches between votes for parties and the seats those parties hold in the legislature, but we could as well point to the geographic concentration of voters.

My point here is not to say we ought to start bickering over riding boundaries. It’s rather to suggest that there are two levers that affect the same system and pushing on one can free up the other for other, better uses. Right now, Canadian districting is constrained by the need for close to equality in population size, because that prevents the right to vote from being unjustifiably infringed. Implementing certain forms of Proportional Representation (especially LPR, discussed in the previous Part), would relax some of the existing constraints on Canadian districting.

That raises the question that’s at the heart of this Part: if we could do anything with ridings, what should we do? I’m going to start with some of the theory on what individual ridings and districting plans as a whole are for. Then I’m going to suggest that whatever it is we’re trying to do, there’s no good reason to define ridings solely according to longitude and latitude. Third, I’m going to imagine what ridings might look like if we followed some plausible desiderata in a PR environment. I end by suggesting that a good riding map would involve combining many existing ridings (according to whether they are rural, urban, or suburban), but (where practical) subdividing them according to the age of the voter, so that Millennials and Boomers don’t vote in the same riding, even if they live in the same house.

A good riding


The desiderata of a riding (or districting plan) can be divided into four categories: practical concerns; political concerns; community concerns; and formal concerns.

I’m relatively confident I’ve hit most of the important points in this section, but if you think I’ve missed something, please be in touch. I’m less confident that I’m right in all of my analyses, and many of opinions I express below are at most weakly held.

Practical concerns

Practical concerns include:

  • Clarity of district. Voters should readily know which riding they are in.
  • Ease of tabulation. The state should readily be able to count votes. Traditionally, this might mean that a polling station handles only one riding.
  • Ease of voting. Voters need to be able to access the polling station.
  • Resistance to strategic riding selection. It should not be practical for voters to change riding simply for the sake of affecting the election.

These concerns all were addressed by traditional, geographically-tight riding boundaries.

Political concerns

Political concerns are much more controversial. In listing them here, I simply note that they are attributes of ridings that may be desired by some people. They include:

  • Maximizing (or minimizing) party seat-count. Some people prefer ridings to be shaped so that a particular party will be more (or less) likely to be victorious. This is generally called “partisan gerrymandering”.
  • Incumbent protection (or vulnerability). Some people prefer ridings to be shaped so that incumbents do not have to fight very hard to keep their seat. They see it as desirable that elected representatives can expect to serve for longer and do not need to spend as much time and resources on campaigning. They also see legislatures being more efficient when more elected representatives have experience with the legislative process. Others prefer ridings to be shaped to make it easer to vote out incumbents, and for new representatives to be elected.
  • Minimizing (or maximizing) competitive ridings. Some people (especially those who work for parties) prefer there to be very few competitive riding so that resources can be efficiently allocated to the “battlegrounds” and others can be ignored. Others prefer to maximize the number of competitive ridings so that more electors have a meaningful choice.

Each of these is possible to accomplish through traditional, geographic-based districting.

Community concerns

What I call “community” concerns are those relating to how well a riding boundary tracks that of an existing community.2To the extent that politics also shapes the boundaries of communities, the divide between “political” concerns and “community” ones is muddy. Canadian courts have long recognized reflecting an existing “community” as a basis for setting proper riding boundaries3See, e.g., Reference re Prov Electoral Boundaries (Sask), [1991] 2 SCR 158, (“rivers and municipal boundaries form natural community dividing lines and hence natural electoral boundaries”; “Factors like geography, community history, community interests and minority representation may need to be taken into account to ensure that our legislative assemblies effectively represent the diversity of our social mosaic.”; “The ‘practical living fact’, to borrow Frankfurter J.’s phrase, is that effective representation and good government in this country compel those charged with setting electoral boundaries sometimes to take into account factors other than voter parity, such as geography and community interests.”); Frank v Canada (Attorney General), 2019 SCC 1 per Rowe J (concurring) at para 103 (“Asked differently, are long-term non-residents likely to be less connected to a constituency or to Canada? Especially at the constituency level, I find it difficult to escape the conclusion that long-term non-residents are likely to be less connected to their community.”); Figueroa v Canada (Attorney General), 2003 SCC 37 per Lebel J (concurring) at para 101 (“While I acknowledge the central importance of individual participation, s. 3 is also inherently concerned with the representation of communities, both the various communities that make up Canadian society and the broader community of all Canadians.”). But what do they mean by “community”?

The field of “community studies” has thought deeply about these terms.4To be clear, I don’t claim any membership in this field. One of the classic concepts there is community as “Gemeinschaft”: “The Gemeinschaft community is modeled on the historic village or small town which has inclusive social ties among members based on holistic views of individuals and families, sentiment, traditionalism, and stable or persistent social rankings that have developed over a long time.”5Ted K. Bradshaw (2008) The Post-Place Community: Contributions to the Debate about the Definition of Community, Community Development, 39:1, 5, 6 DOI: 10.1080/15575330809489738, citing Tönnies, F. (2001). Community and civil society (J. Harris & M. Hollis, Trans.). Cambridge: Cambridge University Press. In this definition of community, rural areas are said to have community, while urban areas are thought to lack it.6This may relate to the seminal definition of a “sense of community” as “a feeling that members have of belonging, a feeling that members matter to one another and to the group, and a shared faith that members’ needs will be met through their commitment to be together”: McMillan, D.W. and Chavis, D.M. (1986), Sense of community: A definition and theory. J. Community Psychol., 14: 6-23. 7 Ted K. Bradshaw (2008) The Post-Place Community: Contributions to the Debate about the Definition of Community, Community Development, 39:1, 5, 9-11 DOI: 10.1080/15575330809489738.

Perhaps the better approach is not to define community, but rather to understand why courts (and legislatures) think “community” is an important thing to respect in riding boundaries. One aspect may be that Gemeinschaft communities are inherently valued and reflecting them in riding boundaries can help strengthen those communities — but this is a justification for how elections can help communities, not how communities can help elections. For the latter, one hint may be in the “solidarity” definition of community, implying that members of the community feel they are more relevantly alike to each other than they are relevantly different. On this understanding, whoever is elected as the representative of the community-as-riding can be seen as representing the whole community/riding, rather than just those that voted for them. This aspect is particularly important in single-member districts.

The difficulty with this approach is that solidarity groups seem today to be at different scales than ridings. Small, rural towns are not large enough to feasibly be individual ridings (given the overall number of MPs in Parliament), and so they must be aggregated. But the aggregation of two towns that each have solidarity would not itself have solidarity. There may also be some level of solidarity at higher scales than ridings (e.g., some people feel more solidarity with Canadians than non-Canadians; or with Newfoundlanders than Quebeckers), but again, solidarity is not divisible. A person who feels solidarity with a large group will not necessarily feel greater solidarity with a subgroup of that large group.

Despite these many caveats, I see at least four relevant attributes of community that may be useful as guides to ridings: informational; coherence of economic interests; coherence of social interests; and coherence of social values. These all seem to have political utility for elections.

  • Informational. One aspect of “communities” is that they are sites of dense informational linkages: that is, information passes more extensively and quickly through them than between them. They are social networks, in the classic sense of those words, not the internet-platform sense (although they can overlap). Because information passes quickly, members of the community can develop more nuanced positions about issues of importance to the community than they can about issues outside the community. This is especially useful in a system of representative democracy, for three reasons. First, elected representatives who are part of communities will be able to understand the needs and issues of that community. Second, the density of information networks will allow local elected representatives to channel information back to constituents about their own decisions in Parliament.8Representatives are entrusted with the responsibility to make judgment calls on behalf of their constituents as part of the process of deliberation in Parliament. See, e.g., Edmund Burke, Speech to the Electors of Bristol 3 Nov. 1774. Third, the personal qualities of representatives (whether they, individually, are good at their job) is the type of high-density information that is easy to spread within a community but difficult outside of it.
  • Coherence of economic interests. This is solidarity, but understood economically, not socially. For many issues, I can see alignment of economic interests even between people who otherwise have no particular feelings of mutual support and respect. Such alignment allows electors to choose someone whom they think will support their interests on key distributional issues.
  • Coherence of social interests. Although many questions in society can be turned into economic questions, others concern choices over the division of non-market goods. For example, some people feel mask mandates are a deep affront to personal freedom. Others see them as an obvious public health measure.
  • Coherence of values. The previous two attributes addressed questions where legislatures make choices that have distributional consequences: negative on some people, positive on others. In those cases, one can see the importance of coherence within a community as ensuring that interests are heard when that distribution is made. Coherence of values is a slightly different matter. It is relevant to non-distributive decisions, or decisions where much of the country are onlookers to the distribution. For example, when the gay marriage debate was ongoing, there was probably significant coherence of social interests within the gay community, but outside that community there were probably few common interests but plenty of people who shared values on how that question ought to be resolved. Electors would probably have preferred their representative to share their values, and shaping ridings to meet electors’ preferences can strengthen that.

All of these attributes of community also appear to provide the basis of another oft-stated goal of districting: ensuring minority representation. One goal of districting is sometimes to make it easier for minorities to elect representatives of their choice (for example, through “minority-majority” ridings where the majority of the population is a member of the minority group).9See Guinier, Lani. “No Two Seats: The Elusive Quest for Political Equality” (1991) 77:8 Virginia Law Review 1413, 1430 for the classic exposition. These attributes suggest ways of identifying which minorities are relevant for voting: they are ones whose interests or values will otherwise be most underrepresented.10By underrepresentation, I don’t just mean “within the present legislature” but “including past legislatures”; past disenfranchisement can warrant present over-enfranchisement.

There are also risks associated with communities-as-ridings. One risk is that a competition to become the representative of a community will tend to elect someone who is the most that community. When communities are small towns where being the most a part of the community means showing up at all the parades and volunteering at the foodbank, the community benefits from such representatives. When communities are instead ethnic or religious groups (as in some consociational systems), being the most a part of the community may mean being the most negative about other groups.11See, e.g., Sisk, Power Sharing and International Mediation in Ethnic Conflicts (United States Institute of Peace, 1996) One risk of making ridings that reflect pre-existing divisions in society is that they may exacerbate those divisions and may elect representatives who promise to promote the interests of their riding to the exclusion of the national interest.

This risk seems likely to be higher in an FPTP electoral system than proportional systems, particularly proportional systems like LPR. FPTP allows a plurality of extremists to elect a representative. It prevents extreme views when extreme views are not concentrated in one riding, but allows them to be over-expressed when they are.12Consider the difference between the PPC and Bloc in the last election; two parties with similar overall vote counts, but the Bloc received many more seats because their support was concentrated in Quebec LPR (and other proportional systems, to varying degrees) allows extreme views to be represented proportionally to the extent voters hold them, and no more.

Formal concerns

The final category is formal concerns: that is, concerns about the literal ‘form’ of a riding. These might include13For some treatment of this, see the dissents of Souter, Stevens, and Breyer JJ in Vieth et al. v. Jubelirer, 541 US 267 (2004):

  • Compactness
  • Respect for political subdivisions
  • Contiguity
  • Conformity with geographic features

These concerns appear to be, at best, proxies for the previous concerns; their only intrinsic value lies in their aesthetic effect on the voting map. Although an oddly-shaped district may suggest it was created intentionally, if one agrees with the intention, then that odd shape may well be good!14See, e.g., Lowenstein, Daniel H and Jonathan Steinberg. “Quest for Legislative Districting in the Public Interest: Elusive or Illusory, The Symposium: Gerrymandering and the Courts” (1985–1986) 33 UCLA L Rev 1, 22. Circular districts can be political gerrymanders as much as complex curves — circles are a very good way of “packing” urban voters.

Finally, an implicit, obvious, and paramount concern is for parity or fairness as between electors. Such parity is of “prime importance”, but can be deviated from for other values, including the aforementioned practical and community concerns.15Reference re Prov Electoral Boundaries (Sask), [1991] 2 SCR 158, 184 Overall, the objective of districting seems to be to achieve the community concerns as much as is permitted by practical concerns and the desire for parity — in Canada, expressly political districting appears to have no real support.

The traditional riding, today

As alluded to in some of the discussion above, the traditional riding exhibited many of the positive characteristics and few of the negative characteristics. Geographically defined populations were both social and economic communities, that shared interests and could be sites of some salient contestation as well as some salient agreement. It was also immensely practical to tabulate votes locally. Taking that to today, however, reveals that geographic ridings are not so necessary or desirable as once they were.

First, consider clarity. With modern communication, it’s possible for individuals to learn about their candidates and the policies these espouse even if they are physically far apart. And, the advantage of geography has reduced: people (especially in cities) travel through multiple ridings in the course of a day so it may not be clear to many voters who their options are.

Second, consider ease of voting and voting tabulation. Although polling stations need to be close to voters, there is no imperative that polling stations be close to each other, or for polling stations to handle only one riding. Any logistical challenges seem easy enough to solve.

But that simply shows that solely-geographic definitions are not necessary. Are they still, however, desirable, or more precisely, at which scale are they desirable?

Existing geographic riding boundaries probably reflect social boundaries less well than ever before.16Unless you’re using this geography! As discussed above, while traditional communities have declined, non-place and post-place communities have risen. These communities do not have all the same properties as traditional communities, but they are the closest modern analogue.

Another point against the existing geographic ridings is that they have led to the underrepresentation of certain groups. Various reports, including a 2016 House of Commons Report17Francis Scarpaleggia, Strengthening Democracy in Canada: Principles, Process and Public Engagement for Electoral Reform, (Dec 2016) and a 2019 iPolitics article[src], document this. Here’s a table showing some of the findings.

Group% Canada% Parliament% Difference
Women50%26%18House of Commons Report, 117-48%
Visible Minorities23%[src] 15%[src]-35%
– South Asian5.6%[i] 7.1%[i] (+27%)
– Arab & West Asian2.3%[i] 3%[i] (+30%)
– Black3.5%[i] 1.5%[i] -57%
– East Asian5.4%[i] 2.7%[i] -50%
– Latin American1.3%[i] 0.6%[i] -54%
– Southeast Asian & Filipino3.2%[i] 0%[i] -100%
Age(% voting pop[src])(among known)
– 20-2917%3%[i] -84%
– 30-3918%14%[i] -23%
– 40-4916%30%[i] (+90%)
– 50-5917%32%[i] (+87%)
– 60-6916%15%[i] -7%
– 70+16%6%[i] -61%
Openly LGBTQ3.3%[src]1.18%[i] -64%
[i] indicates datapoints from the iPolitics article

The data here points to some simple conclusions: women are underrepresented; so are many (but not all!) visible minorities and Indigenous persons; young and old people are underrepresented, as are openly LGBTQ people. This is of course only suggestive data: it could be that a riding of all women would elect a male representative (indeed, it is not as though women are minorities in ridings now). For now, suffice to note that geographic ridings, as currently constructed, do not appear particularly good at ensuring representation for these minority groups.

A better riding for tomorrow

Imagining a better riding for tomorrow requires going beyond geographic criteria. Rather than say “every elector living within [these metes and bounds] votes in [this riding]”, one could define ridings according to non-geographic dimensions in addition to geographic ridings. The list of groups underrepresented in Parliament gives a good starting point: one could say that “every elector living within [these metes and bounds] and who has the assigned sex of ‘Female’ votes in [this riding]”. To this list, I’d add two other dimensions that are possibly of interest but not revealed in the iPolitics data: personal or household income and source of citizenship (birth vs naturalization). This list isn’t meant to be exhaustive,19One can imagine other ways of stratifying society. Technolibertarians might like a riding for “those who use cryptocurrency”, which you can vote in only if you make a $1 Bitcoin payment. Anti-natalists might want a riding that doesn’t include any parents. Educational elitists might think that voters with post-secondary degrees should be separated from those without; and et cetera and so on. I’m not going to address those options. and, as I discuss further below, nor do i think all of these attributes are appropriate for use as district boundaries. In defining a multi-dimensional riding, there are a few key concerns. I’ll address those first-off, then return to which dimensions should be adopted.

Multi-dimensional ridings

I should be clear what I mean by “multi-dimensional” riding. I don’t mean ridings that are in different timelines20Those would obviously be pruned by the TVA or ridings that pass through an interdimensional portal to Oz or Pylea. Rather, I mean a riding where the voters in it are defined by multiple criteria: in data science, these are often called dimensions. Ridings defined by residence in a particular area are two-dimensional. Adding further criteria on top of residence makes a riding multi-dimensional.

A few traits are generally desirable for any multidimensional districting plan: that they are comprehensive, non-overlapping, and unambiguous. These properties are implicit in geographic districting.

For a districting plan to be comprehensive, every elector must be affiliated with at least one riding. So, if there is a riding that is limited to “those assigned the sex ‘Female'”, there must be another riding that allows “those not assigned the sex ‘Female'”. Comprehensiveness is not only a desirable traits, it is a necessary one to protect electors’ right to vote.

For a districting plan to be non-overlapping, every elector must be a member of no more than one riding. The non-overlapping trait is generally desirable, but is not necessary — the alternative is for electors to choose which riding to vote in. For example, today many post-secondary students can choose whether to vote in the riding of their term-time residence or in the riding they lived in prior to relocating for their studies.

Nonetheless, overlapping ridings are problematic. They attenuate the accountability of the member(s) with the electors, may confuse electors, and create the possibility of electors making strategic choices of riding to gain an advantage (i.e. rather than voting in the riding that the electors feel most connected to, an elector may vote in the riding where they have the greater chance of affecting the result). Overlapping ridings also make determining how many voters there are in a riding difficult and so make it difficult to ensure voter parity. As a general rule, overlapping ridings should be avoided.

For a districting plan to be unambiguous, both an elector and the state should know which riding(s) the elector is a part of. As with the non-overlapping trait, the purpose of ridings is better accomplished when elected representatives know who their constituents are, and when electors know who their elected representatives are. Administratively, unambiguous traits make it much easier for poll station workers to provide the right ballot.

The continuing role of geography

Looking at “community” in terms of the four aspects described above (informational; shared economic interests; shared social interests; and shared social values) suggests that geography may continue to matter — but how it matters is at scales above and below the traditional riding. Scales below the traditional riding are not useful for present purposes, because smaller ridings would mean more MPs. Two geographic dimensions that seem particularly relevant are provincial boundaries and the urban / suburban / rural divide.

Provincial boundaries, especially for provinces other than Ontario, appear relevant on all four aspects. Provinces have provincial media and provincial politics that pass information in the classic way; provinces also have shared economic industries (from oil in Alberta to tourism in PEI) or distributional economic concerns (what will come of Newfoundland & Labrador’s parlous financial state?). Moreover, provinces appear to share certain social interests and values, for example on questions like “how risky should we be with new technology?” or “how important are official language requirements?”. Having independent legislatures has fostered (maintained?) independent political cultures and associated social values.

The urban/suburban/rural split also appears to track these aspects of “community”. The informational aspect of community may be less obvious (individual urban neighbourhoods or rural small towns may have networks, but a province-spanning rural network does not seem to exist), but economic interests appear to. The industries present in rural areas versus in cities are very different21Consider, for example, the professional service firms that colocate in what Saskia Sassen famously called “Global Cities”; and, for at least the property-holding classes, the change in land value of one property tends to have externalities for other landowners.

But perhaps the more significant difference is in terms of social values and social interests. Rural, suburban, and urban communities tend to differ in their fundamental way of life, from how communitarian they are to how much they see diversity of others’ experiences as positive or negative. Although I’m certainly undertheorizing the distinction, both “community” scholars22See Ted K. Bradshaw (2008) The Post-Place Community: Contributions to the Debate about the Definition of Community, Community Development, 39:1, 5, passim DOI: 10.1080/15575330809489738 and courts have accepted23In Reference re Prov Electoral Boundaries (Sask), [1991] 2 SCR 158, the majority found “it is not self-evident that [dormitory] communities should be joined with the communities where the residents worked.  Their interests may differ from those of the community in the urban riding, and their inclusion might sweep in truly rural residents.” there being a fundamental difference between these areas that is right to reflect in ridings.

The advantages of LPR

Creating equal-population non-overlapping, multidimensional ridings is likely to lead to extremely arbitrary divisions that may not map onto any of the desiderata for a good riding (e.g., for age, “those born between before January 21st, 1978 and April 11th, 1994”). I introduced a solution to this in the previous Part of this post. Under my proposed locally proportional representation (LPR) system, each legislator’s votes in the legislature are weighted according to the number of votes they received in the previous election.

LPR is particularly useful with multidimensional ridings, because it frees these ridings from the constraints of “equal population” and allows creating the riding that best reflects existing communities without infringing others’ voting rights. LPR also makes overlapping riding definitions less harmful, since electors’ choice about which riding to vote in will be reflected in the ultimate strength of their representative’s vote. Keeping riding size equal before the vote is less important.

Multidimensional ridings may be more prone to create tokenization, unless LPR is implemented. Without LPR, multidimensional ridings increase the risk of tokenization because the social salience of a multidimensional riding has the potential to be much higher than with existing geographic ridings — that is, a multidimensional riding may more directly be the riding of a defined group. Under the present First Past the Post (FPTP) system, all the complexity and disagreement of a community is collapsed to a single value, like a lossy JPEG compression algorithm. FPTP is inherently reductive — it reduces an election to “which MP gets a plurality of the vote”.

LPR helps solve this problem, because it elects multiple members in every riding. As I explained in the previous post, LPR would prevent Torontonians (Albertans) from all being painted as latte-sipping liberals (rodeo-going anti-vaxxers).

Going forward, I’m going to assume the electoral system is LPR, not FPTP.

Which non-geographic features

A useful starting place for thinking about non-geographic features is to look to which groups are underrepresented in Parliament: women; visible minorities; openly LGBTQ people; Indigenous people; young or old people; and also naturalized citizens or previously low-income earners. A naive assumption could be that these groups are underrepresented in Parliament because the current riding boundary system disadvantages them.

The same desiderata for defining a “good” districting plan are useful for non-geographic dimensions as for geographic ones, especially the practical and social concerns. This section considers each of these possible dimensions in light of those concerns. It concludes that stratifying by age and Indigeneity may be the most relevant to consider. I’m not at all confident in these conclusions: they’re where my analysis led this time, but I expect there are at least some things I haven’t properly considered.

One practical concern is resistance to games-playing. This concern excludes any purely self-reported (i.e. self-subjective) characteristic, including LGBTQ identification and likely also membership in a visible minority group. One could readily imagine griefers flooding an “LGBTQ” riding by falsely self-identifying as LGBTQ. Neither LGBTQ status nor visible minority status can readily be made “objective” — there is no “test” for “being” LGBTQ other than self-identification. And a state-imposed “objective” test for visible minority membership seems likely to lead to controversy about who “really” counts. They may be important dimensions, but they are not dimensions that can readily be used by the state to stratify the elector population into ridings.

Indigeneity is more complex, since the notion of who “counts” as an Indigenous person is contested. Self-identification could run into problematic situations24e.g., Michelle Latimer, as well as griefers. Relying on registration (status) under the Indian Act would be a possible solution, but may not be ideal. Status is a politically controversial topic, and would exclude many Indigenous persons: about half the self-reported Indigenous persons appear not to be registered under the Indian Act.25“According to the Census of Canada, there were 820,120 Registered Indians in Canada in 2016, comprising 49% of the Indigenous population.”: Indigenous Services Canada, Annual Report to Parliament 2020. But to say that at present there is no objectively discernible definition of Indigeneity is not to say there could never be. One approach is suggested by Rowe J’s majority opinion in R v Desautel, 2021 SCC 17: “it is for Aboriginal peoples … to define themselves” (at para 86). It would be an ambitious project to build a system by which Indigenous persons could be identified by Aboriginal peoples, and for that identification to be accepted by the Canadian state; but it would be a worthy project as well.

Age, gender, income, and source of citizenship, by contrast, do not raise any practical concerns. Age, gender, income, and source of citizenship are already recorded by various parts of the state. Although state-assigned gender can be changed, it is sufficiently painful to change that it seems unlikely to be a target for griefers. Moreover, these dimensions all cut across existing geographies and so are unlikely to be represented directly as it stands.

Considering the aforementioned aspects of community winnows this list further. As described above, some of the aspects of community that are relevant for districting include informational aspects, coherence of economic interests; coherence of social interests; and coherence of social values. And, in the negative aspect, using an attribute as a districting dimension seems likely to increase the political salience of that dimension and foster it as an identity among those constituents. One of the saving graces of present riding geography is that the riding boundaries are so deeply arbitrary that there is no risk of candidates attempting to outdo each other by being the most “from Nepean”, or of them trying to make Nepean distinct from “Ottawa West-Nepean”. These risks are more acute with other dimensions.

In particular, source of citizenship appears undesirable. Although Canada famously has a “mosaic” approach to ethnicities, it does so in part by asserting that naturalized immigrants are as Canadian as born citizens. An “immigrant” riding seems likely to be seen as othering rather than embracing.

Gender would seem a salient dividing line for many of the politically-relevant aspects of community: there absolutely are ways in which information may flow on gendered lines, and experiences of life differ by gender, and economic interests too are affected by gender26Through the gender wage gap, among other phenomena. But increasing the political salience of gender (either as a binary, or with non-binary people being unhelpfully lumped into one or the other group) also does not seem particularly useful in present day. In an earlier era of feminism, a distinct “women’s vote” may have been appropriate, especially when the rest of the vote was a de facto men’s vote. Today, however, the male Prime Minister asserts his own feminism. And, whatever the cause of the underrepresentation of women in Parliament27Increased toxicity directed at female candidates? Expectations for gender roles that prevent a spouse (most elected representatives are married) from playing a support role?, “women are a minority group” cannot be it.

Indigeneity, income, and age all, however, have greater attractions.

Heightening the salience of Indigenous identity seems likely to be valued in Canadian society as a self-conscious act of non-assimilation. This is the one time where I would see merit in a nation-wide riding, as a way of respecting Indigenous’ persons pre-existing occupation of what is now Canadian territory. There appears no good reason for superimposing the colonial boundaries of provinces upon Indigenous representation in Parliament. Providing a nationwide riding each for the First Nations, Métis, and Inuit was suggested in the aforementioned House of Commons report28(at 122-123). Since these groups have significantly different populations, fairness would only be preserved if a scheme like LPR is implemented. Allowing Indigenous persons to choose whether they wish to vote in a local riding or a national Indigenous riding seems a reasonable cause to deviate from the non-overlapping rule.

Heightening the salience of reported individual or household income has its attractions, but appears likely to be politically unpalatable. Class consciousness is lacking in Canadian society today. Much of the political discourse falsely insists that everyone is middle class. From an economic perspective, the interests of members of the same income tranche will often align, and be opposed to those of another income tranche (especially on issues such as taxes, or landlord-tenant rules, or urban development restrictions). But it is precisely that opposition to entrenched interest that is why defining ridings in a way that increases class consciousness is likely a non-starter.

By contrast, heightening the salience of age appears both unlikely to be problematic, and immediately helpful at reducing underrepresentation in Parliament. Youth alienation from politics may in part be due to the predominance of the middle-aged in current legislatures and the sense that no one there represents them. A voice for youth may ensure that underrepresented values are heard.

Focusing on age as a distinguishing feature also plays well for the other social values. Social networks are often much stronger within age groups than between them: there are clear generational divides even on which tools they use to communicate (e.g., Facebook is now for the olds; TikTok is for the tweens). Distinct social values appear in each generation. And, now perhaps even more than ever, there are distinct coherent economic interests in each generation. One of the largest distributional questions facing modern states is the extent to which current citizens should pay to mitigate or prevent problems that will be born by the next generation, whether that problem is climate change, the national debt, or infrastructure in need of repair. These are questions that deserve proper debate in the legislatures, in a way that is only possible if there are representatives of the clashing views.

For all this discussion, I reach a fairly simple conclusion: create three national ridings for Indigenous groups, and stratify geographic ridings by age.

The imagined electoral map

Where does this all get us, practically speaking? Here’s a back-of-the-napkin attempt. First, using three-member ridings under LPR, and keeping the total number of MPs constant at 338 (let’s say 336 just to make it divisible by 3), 112 ridings could exist. With three national Indigenous ridings (one each for First Nations, Métis, and Inuit), that leaves 109. Of those 109 ridings, the remainder need to be assigned to a geographic area and an age group.

In general, this can be done by dividing each province into its most salient regions. In BC, this might be Vancouver, the Vancouver suburbs, Vancouver Island, the BC interior, and Northern BC. Then, for each of these regions, decide how many age tranches you can afford in the region. How would the age cohorts be chosen? Tracking major generational divides seems sensible, where it is possible: if there are 5 cohorts cut the population into Gen Z, Millennials, Gen X, Boomers, and their parents; if there are only 2, split “Boomers and older” from “Gen X and younger”. Are there the appropriate number of ridings, given the overall population of the province? If there are too many, adjust ridings by collapsing them together, by reducing age cohorts within a geographic region or combining age cohorts across regions (e.g., create a riding of “seniors in Northern BC and the BC interior”). If there are too few, do the opposite.

I’m not going to suggest what the ultimate electoral map might look like; I tried, and it was so thoroughly arbitrary that it did not seem useful even as a strawman. So that exercise is left to the reader.


Although Canadian riding boundaries don’t often occasion much debate, they’re important and significantly affect election results. It’s worth thinking about why we have them, and what makes one set of riding boundaries “good” and another one “bad”; in the Canadian context, probably the answer to this is “good riding maps are practical and don’t feel politically-motivated and do reflect existing communities”. That’s a fine answer, but purely geographic boundaries for maps may not be the best way of reflecting existing communities. People now have communities that are not so trammelled by place, but which are stratified on other dimensions.

This then raised the question: what are these other dimensions, and can they be used to construct better districts? I present the weakly-held view that Indigenous identity and age may be relevant dimensions that would better serve the purposes of riding districting than geographic boundaries alone. They are especially useful if a system like LPR is implemented — one that does not require equal-population ridings to reach a fair result.

Even if you like this idea, you may think it is so much faffery, given the lack of appetite in Parliament for electoral reform. Part III is my answer to that: in an inversion of my argument about s 33, here I think it’s not Parliament we should turn to, but the Courts.