STV for Ontario Submission accompanying a presentation by Eileen Wennekers December 6, 2006 This voting system presentation has been funded in part by an academic fellowship from Fair Vote Ontario. It is one of four such projects funded by Fair Vote Ontario to assist the deliberations of the Ontario Citizens’ Assembly. However the work is solely the product of its author and does not carry the endorsement of Fair Vote Ontario. Contents 1 Why choose STV? 3 1.1 Introduction ................................ 3 1.2 STV is simple ............................... 4 1.3 STV is proportional ............................ 5 1.4 STV is widespread ............................. 6 1.5 STV and women and minorities ..................... 7 1.6 The Richard commission ......................... 10 1.7 STV and the 9 criteria .......................... 11 1.8 STV and Ontario ............................. 14 2 What might an STV map look like? 17 2.1 Geographic representation ........................ 17 2.2 Rightsizing the legislature ......................... 18 2.3 STV for a larger legislature ........................ 20 2.4 Constructing the 106-seat map ...................... 21 2.5 Constructing the 135-seat map ...................... 23 3 Maps and tables 27 3.1 The 106-seat map ............................. 27 3.2 The 135-seat map ............................. 30 1
Submission accompanying a presentation by Eileen Wennekers
December 6, 2006
This voting system presentation has been funded in part by an academic fellowshipfrom Fair Vote Ontario. It is one of four such projects funded by Fair Vote Ontario toassist the deliberations of the Ontario Citizens’ Assembly. However the work is solelythe product of its author and does not carry the endorsement of Fair Vote Ontario.
The British Columbia Citizens Assembly on Electoral Reform spent a year evaluatingelectoral systems and concluded by recommending the adoption of STV. The OntarioCitizen’s Assembly should make the same recommendation.
STV is the oldest, best-established, simplest, and, in the English-speaking world,the most commonly used system of proportional representation. It is also probablythe most popular voting system, in the sense that “those who like it, like it a lot”.
And within Canada, it is the most misunderstood. It has been called arcane, whenin fact it has been used more frequently than rival systems such as the German system,and until 10 years ago was significantly more common. It has been called complexand difficult to understand, whereas in fact it is simpler than other forms of PR,especially in terms of understanding the effect of one’s vote. Some critics have goneso far as to say that STV is not a system of proportional representation, whereas it ishighly proportional, and in fact the terms “proportional representation” and “singletransferable vote” were long synonymous in English. In the IDEA Handbook, onlySTV and list systems are considered to be forms or proportional representation. Themixed-member system used in Germany (hereafter, MMP) is listed under “mixed”systems.
David Farrell provides an excellent summary of STV’s qualities in the introductionto chapter six of his book Comparing Electoral Systems (Prentice Hall: 1997):
From the very beginning, STV has had its supporters. Indeed, if onewere to carry out a head count of the scholars writing about electoralsystems today, it is likely that many would rate STV very highly. On theface of it, there appear to be good reasons for this, for here is a systemwhich is both proportional and which facilitates constituency politicians.In contrast to the list systems of PR, under the STV system the electorsvote for candidates, not parties; electors stand a better chance of seeingtheir preferred candidate elected; and, unlike first past the post (FPTP)and the majoritarian systems, the voters have a choice between a numberof constituency politicians. (p.110-1)
1.2 STV is simple
STV’s reputation as a difficult system is completely undeserved. The core functioningof STV, as readily explained to a class of middle-school students in a school gym-nasium, is as follows. A number of candidates step forward and supporters grouparound them. Each candidate attempts to attract a sufficiently large group of sup-porters (called a quota) to ensure election. Supporters of weaker candidates naturallyabandon them and gravitate to other, stronger, candidates who still represent theirviews. Supporters of very strong candidates who are beyond quota also move along,to help secure the election of aligned candidates.
That is all that needs to be explained. One could go on into more detail, buttry it and you will find that you are simply going over old ground which has alreadybeen covered. Perceived complexities come, not from the principles of the underlyingsystem, but from two other sources: the replacement of random transfers by fractionaltransfers, and a need to couple the system with a secret ballot – which requires thatpreferences are expressed in advance and followed out by counters who carry out thevoter’s expressed wishes as though the voter were still present.
This self-division of the electorate into equal groups, each electing one representa-tive, is the very essence of proportionality. Those who claim not to have understoodthis explanation of STV (and who usually blame STV for a supposed "incomprehen-sibility") do not really understand proportional representation, or why people believeit is important.
We submit, indeed, that in its basic workings STV is simpler than a typicalparty-list system, which requires understanding a potentially complicated countingmethod, and certainly simpler than MMP with its two methods of being elected, andpotentially candidates who are simultaneously attempting both. And a party-listsystem also requires some kind of reasonable method for the drawing up of the lists;if one isn’t keen on having party insiders dictate the compositin of the lists, one has tocome up with a reasonable method for filling a list, i.e. selecting multiple candidatesfrom a large field. In this case, one is naturally driven to consider something likeSTV!
Some in British Columbia argued that the confusion expressed by some commen-tators over STV would have been avoided entirely had MMP been put to the votersinstead, because MMP, they argue, is straightforward and easy to understand. Thatthis is far from being the case is well exemplified by a survey of Scottish voters, whichasked six yes-no questions about MMP, the system being used to elect the ScottishParliament, and found that the percentage of correct answers was lower than the 50%which could be expected from random guessing. 1.
The simplicity (and elegance) of STV is even more apparent when one tries tofollow the effect of one’s vote. Pursuing the playing-field analogy, when the votingstops, you only need look to see what group you are standing in to know which can-didate your vote helped elect. In a party-list context, when your favourite candidateis number 3 on some party’s list, and you also like candidates 1, 4 and 6, but dislike
candidates 2 and 5, and then you cast a vote which helps your party capture a fifthseat (thus helping to elect candidate 5, whom you don’t like) are you glad you votedfor that party, or not? What was your best strategy before the polls closed: to votefor the party, hoping you would be pushing number 4 or number 6 over the top, orto refrain?
STV creates no such moral quandaries. You vote for a candidate if you like her,and you don’t if you don’t.
1.3 STV is proportional
STV is also sometimes thought of as not fully proportional, in that one doesn’t simplytot up a party’s vote and count out a matching number of representatives. However,it is not necessary to do this in order to match a party’s seat percentage to its votepercentage, as even a cursory review of STV results from the Republic of Ireland willattest.
STV can most easily be thought of as a party list sytem in which a voter’s abilityto express preferences within her party’s list has been freed to a radical degree. Notonly can she decide which of her party’s candidates she likes first, second, third,etc., she can even abandon the party after a while if the complete set of candidatespresented is not to her liking. She can even take her vote away from her first-choiceparty in favour of other candidates, and then bring it back again later.
This freedom allows a voter who is torn among several parties or candidates toexpress her complex set of preferences and thus to more accurately position herself inthe concept-space of possible preferences. Therefore the intentions of the electorateare more accurately captured than in a straight party-list system where each voteris forced to “round off” her preferences to a single party, effectively forcing all thepoints in the concept-space to be repositioned as projections into a single dimension.Retaining the more complete multivariate data permits STV to be more proportionalto the true intentions of the electorate.
This greater inherent proportionality allows STV to function with a smaller dis-trict magnitude than more closed list-PR systems, increasing STV’s ability to pro-vide for local geographic representation. The local-representation aspect is furtherenhanced by the direct linkage which STV provides between the voter and the candi-date, ultimately allowing STV to dispense entirely with the single-member districtswhich MMP grafts on to list-PR.
But the real success of STV as a proportional system is that it takes proportion-ality to the level of individual candidates. Each MPP who is elected has collected thebacking of the same number of voters, called a quota. (In practice, there would bean exemption where the quotas would be smaller for northern MPPs, as under thesingle-member system we use today.) Therefore, the electorate has been divided into106 separate little piles, each of the same size, and each with the same amount ofrepresentation: one MPP2. What could be more proportional than that? By contrast,
2These piles will include the vast majority of voters, as there will be few whose ballot
a party-focused system would divide the electorate into just three or four (at most,perhaps, five) piles, and associate a collection of MPPs with each pile. How can thatbe more proportional?
1.4 STV is widespread
Supporters of MMP often imply that their preferred system is widely used aroundthe world, while STV is an obscure system used only, because of historical quirks, intwo small island nations, Ireland and Malta.
However, people who claim that MMP is used in dozens of countries are countingthe more widely used list systems as being MMP, simply because MMP uses lists. Thisis nonsense. It would be just as correct to say that Single Member Plurality ("first-past-the-post") is a form of MMP, because MMP uses single-member constituencies.
Less careless commentators will note, not that dozens of countries use MMP(blatantly false) but that dozens of countries use "proportional representation", bywhich they mean list-PR and MMP taken together. Well, fair enough, but STV isalso a form of proportional representation, so any category broad enough to coverboth list-PR and MMP countries should include STV countries as well. This "gowith the flow" argument is, therefore, no more an argument in favour of MMP thanit is one in favour of STV.
In fact, from 1949 (when the Federal Republic of Germany adopted its constitu-tion) until the mid-1990s, MMP was used in only one country, Germany, although itwas used at both state and federal levels. It has subsequently been adopted by NewZealand in 1993, Bolivia in 19963, and by Lesotho and Mexico4. It is also used toelect the Scottish Parliament and the Welsh Assembly5, which are arguably examplesof use at a national level, although these legislatures do not have the full powers ofsovereign national parliaments. Examples at a sub-national level (analogous to On-tario) seem to be very rare, confined to the German states and the London Assembly(created in 2000).
STV has been used in Ireland since 1921 (the year prior to the formation of theIrish Free State), in Malta, and in Northern Ireland. It is used in Australia at thenational level for Senate elections, and, since its recent adoption in Victoria, in allstates except for Queensland. Although some of these Australian examples are not“pure” STV because they use group voting tickets, those of Tasmania (where it hasbeen used continuously since 1896) and the Australian Capital Territory are. Atthe subnational level, STV has recently been adopted for Scottish local elections, forsome New Zealand municipalities, and for New Zealand health boards. It was at onetime used in 22 US cities including New York City, Cincinnati and Cleveland, and isstill used in Cambridge, Massachusetts. Finally, in Canada, STV was once used to
preferences include no successful candidate.3http://www.aceproject.org/ace-en/topics/es/esy/esy_bo4http://en.wikipedia.org/wiki/Mixed_member_proportional_representation5Where it is known as AMS, the Additional Members System
elect city governments in Winnipeg and Vancouver, and to elect the urban MLAs inManitoba and Alberta from the 1920s to the mid-1950s.
This quick survey should immediately convince one that MMP is not a morewidespread system than STV. In fact, as recently as 1996, STV was much morewidely used than MMP. Only one country, Germany, has an established history withMMP. New Zealand is establishing one, having now conducted four MMP elections.
In contrast, Tasmania has used STV for over one hundred years, and Tasmaniansare reportedly very proud of their system. And, importantly, STV is the only formof proportional representation with an established history in Canada. Indeed, it willagain be present in Canada if the British Columbia referendum passes in 2009.
1.5 STV and women and minorities
Many argue that one goal of electoral reform should be to increase the number ofwomen in the legislature, and some argue that MMP is the best way to do this, becauseit uses lists drawn up by the parties. This argument relies on the concept of "zippered"lists, i.e. those in which the candidates alternate [male/]female/male/female . . . , likethe interlocking teeth of the two halves of a zip fastener. (Similar arguments could bepresented for other underrepresented groups, such as aboriginals, but it is less clearhow a list-quota for a small group would be established. As the numbers of men andwomen happen to be roughly equal, a simple alternation suffices.)
Effectively, under MMP the lists are constructed by the parties in advance, whileunder STV the candidates are ranked by the voters on election day. Why shouldwe expect the parties, rather than the voters, to do a better job of treating menand women equally? This will happen, say MMP advocates, either if the partiesare compelled to do so via legislation, as has been done in France (i.e. if the modelput forward by the Citizens’ Assembly is not merely MMP, but MMP plus zipperedlists), or if the parties fear the loss of voter support if they do not zipper their lists.Experience in France suggests that the second mechanism, voter pressure, worksbetter on some parties than others; some parties, indeed, opted to pay fines ratherthan nominate equal numbers of men and women, demonstrating that they felt nopressure at all to comply.
But, supposing that zippered lists have been presented, the argument runs asfollows. Suppose a party has won enough votes to elect 9 people. As these will bethe group of 9 people on top of its (closed, zippered) list, it will necessarily containeither four or five women, as close to parity as that party can achieve. If the numberelected is even, then exact parity can be achieved. Of course, MMP does not simplyuse lists; there are constituencies as well. Suppose that a disproportionate number ofmen have been elected in the constituencies. These men will be crossed off the list(hopefully somewhere near the top) leaving the list top-heavy with women, so thatagain the party’s caucus achieves parity.
This simple argument has two devastating flaws. First, it requires that the men onthe lists are the same men as the ones who have been elected in all those constituencies.In practice, this is likely to be the case if dual candidacies are permitted. Some MMP
models suggest that ’dual candidacies’, i.e. standing both in a constituency or ona list, should be banned, in order to prevent the all-too-common scenario in whicha candidate defeated by the voters in a constituency is nevertheless returned on thelists. If this is the case, then equal numbers of men and women will be returned fromthe list, together with all the men from the constituencies. The same problem willoccur if dual candidacies are permitted (and used) but if the party has opted to putthose candidates it thinks have safe constituencies at the bottom, rather than thetop, of its list. Again, men won’t be removed from the top of the list, so it will notbecome correspondingly woman-heavy. One can, however, realistically hope that thisscenario will not materialize.
The second flaw is more fundamental. While the simple argument is correct, thatzippered lists achieve balanced party caucuses, they can only do so for those partiesin which the lists provide a significant number of MPPs. The lists in an MMP systemare often referred to as ’compensatory lists’, for the reason that they are there tocompensate parties who did not get their fair share of seats in the constituencies.The lists are not used at all for parties who do not require compensation: mostcommonly, this means the party which won the election and is forming the core ofthe government!
The experience in Wales is extremely instructive, although it relates to ethnicminorities rather than to women. (The Welsh Assembly is dominated by left-wingparties who have no trouble nominating and electing women, and there have beennumerous high-profile women elected by Labour in Welsh Westminster constituencies,as well as to the Assembly.)
Wales is a small, homogeneous country with low immigration and correspondinglysmall minority populations. The 2001 census puts the population at 96% BritishWhites (nearly all Welsh or English), about 2% other Europeans (such as Irish), andabout 2% non-white (Asian, Black or mixed). Even at their most concentrated, inparts of Cardiff and Newport, non-whites do not form the majority of the populationanywhere.
Hence it has been difficult for non-whites to make electoral gains in the same waythat Blacks and Muslims have done in parts of England. The Welsh Labour Partydecided to use the vehicle of MMP to change that, by promoting the candidacy ofone minority candidate on each of the five regional lists.
Unfortunately for the champions of minority representation, Labour easily wonthe election, winning 30 of the constituency seats. (There are 40 constituency seatsand 20 list seats in the Assembly, for a total of 60 Assembly Members.) As Labourhad thus won exactly 50% of the seats in the Assembly outright, and done so on amere 38% of the popular vote, Labour therefore needed no compensatory seats, andreceived none. The 30 Labour Assembly Members elected were therefore the 30 whohad been elected in their constituencies.
The problem was not that Labour had not placed the candidates high enough uptheir lists. One does not need to be at the top of a list if all the higher candidateshave won in constituencies. In fact, three of the candidates (ranked 1st, 3rd and 4th)were next in line had Labour picked up even a single list seat in their region, and theother two (both ranked 2nd) needed Labour to pick up two list seats. But Labour
won zero list seats; there was no position which would have been high enough.Of course, a great many of the other candidates on the lists were elected, namely
through constituencies. In fact, of White candidates on lists, 28 out of 39 were elected(all in constituencies). In South Wales, where most non-white Welsh live, 20 of 22White candidates were elected for Labour, and 0 of 3 non-whites.
The only way Labour could have secured the election of its minority candidateswould have been to ignore the lists, and parachute them into nominations in winnableconstituencies. In other words, exactly the same avenue which would be open to themunder a FPTP system, and which doesn’t work because of a desire to leave localmembers in control of local nominations.
What failed for minorities in Wales could equally fail for women under MMPelsewhere6. The compensatory lists would kick in to elect women in all parties exceptthe one which matters, the largest one.
The real-world evidence for this from New Zealand is necessarily limited becauseonly four MMP elections have been conducted there. Nevertheless, there is someevidence that what our theoretical considerations lead us fear is in fact borne out inpractice. The proportion of women in the New Zealand Labour Party caucus has beenrelatively stable across the four MMP elections (35%, 37%, 35%, 38%), regardless ofthe party’s overall fortunes. But in the National Party, the proportion of women inthe party’s caucus was only 16% following National’s victory in the 1996 election,but increased to 21%, 22% and 23% in the three subsequent elections, each of whichNational lost to Labour.
At least in the National Party, therefore, the number of women elected appears toincrease in opposition, and decrease in government. The overall proportion of womenin the parliament might get somewhere into the neighbourhood of 50%, but on thegovernment side it could be much lower. And it is exactly these government MPsfrom whom the prime minister selects the cabinet.
STV does not suffer from this government/opposition problem because it treatsall parties, and indeed all candidates, symmetrically and equally.
Indeed, the symmetry of STV (all candidates are presented side-by-side andequally; each successful candidate has received the same number of vores, one quota)will be a further benefit to elected women and minorities. Because each elected repre-sentative has gone through the same process, there is no question that any MMP owesher or his position to tokenism or special consideration. This perception of equalityamong all MMPs can only help elected women and minority representives in theirsearch for genuine positions of power and responsibility within the legislature.
Considerable research tells us that the voters are not as a body prejudiced; oncewomen have secured candidacies, voters will select them equally with men. In a STVscenario, the onus is placed upon the party to develop women into strong candidates,rather that coercing voters into a statistical but not pragmatical gender-equitabledistribution of power through the imposition of zippered lists.
6As Karen Etheridge noted upon reading the Richard Commission report in 2004 (per-sonal communication).
In short, and despite the circulation of some facile arguments to the contrary,there is no solid evidence that STV will perform less well than any other system ofproportional representation in terms of securing equality of representation for womenand minorities. Indeed, it should perform better than MMP because MMP suffersfrom an asymmetry between government and opposition, and in another respect it isbetter than any ordered-list system because it is not vulnerable to charges of tokenism.
To be fair to MMP, the government/opposition effect can be somewhat mitigatedby the use of regional lists (instead of one long list for the whole province): becausedifferent parties will predominate in different regions, each party will presumably haverecourse to some list, somewhere. However, the problem in Wales arose specificallyusing regional lists, and it is not hard to see this scenario repeating in, for exampleToronto, where it easy to imagine the Liberal Party sweeping the constituencies andtherefore failing to return any list-only candidate it had preferred there.
1.6 The Richard commission
Following the Welsh Assembly election of 2003, an independent commission, chairedby Lord Richard, a former EEC commissioner and UK ambassador to the UnitedNations, was established to report on the powers and electoral arrangements for theWelsh Assembly.
Those who appointed the commission did not expect them to recommend a radicaloverhaul of the MMP system being used in Wales. Rather, they were expected tosettle such matters as appropriate terms for enlarging the Assembly in preparationfor the devolution of more powers from Westminster, and the question of whetherdual candidacies should be continued. This latter issue is sometimes known in Walesas the "Clwyd West" question, after a constituency in which five candidates stoodfor election in 2003, and four of them ended up elected to the Assembly. Indeed, 18of the 20 list members were defeated constituency candidates.
Nevertheless, the Richard Commission recommended abandoning MMP alto-gether, in favour of STV. To quote from their rationale:
Thus, on proportionality grounds alone, the choice between AMS7
and STV is not clear-cut. In our view, the most important advantagesof STV over AMS are three-fold:
* all Members would have equal status and share the same relation-ship with constituents;
* the majority of votes would count and there would be no such thingas a wholly safe seat – giving all the parties an incentive to campaign inevery constituency;
* there are opportunities for greater representation of minority inter-ests;
7As MMP is known in Wales and Scotland (our footnote).
Although the STV recommendation from an independent commission has subse-quently been rejected by the politicians, this should not necessarily be considered amark against STV.
Indeed, politicians have a history of rejecting STV, a system remarkable for hand-ing power over to ordinary voters.
Farrell concludes his comparative study of electoral systems with these words:
. . . the STV system perhaps comes closest to an ideal electoral sys-tem. It combines the virtues of proportionality with those of preferentialvoting. It is a system which politicians, given a choice, would proba-bly least like to see introduced, but which voters, given a choice, shouldchoose. (Farrell, p. 168)
1.7 STV and the 9 criteria
The Citizens’ Assembly has been asked to consider eight criteria in its evaluationof proposed electoral systems; in a later document a ninth criterion, simplicity andpracticality, was added8. STV scores very highly according to these measures.
Some of the criteria identify properties which are held in common by all formsof proportional representation, including STV, most notably (2b) proportionality.All forms of proportional representation have also been linked to (7) higher voterturnout. Since they encourage coalitions, they can facilitate (5) stable and effec-tive government, and since they ensure that opposition parties have caucuses of ahealthy size, they can help create an (6) effective parliament, one with a strongopposition as well as a strong government.
We believe that the real advantage of STV over other forms of proportional rep-resentation is the degree of (3) voter choice; indeed STV can best be thought ofas proportional representation together with voter choice, the ability to prefer notmerely a party, but a candidate or candidates within that party. For many, othercriteria will be seen as following on from this: (1) legitimacy and (8) account-ability can be viewed as consequent to this direct link between the voters and theirelected representatives. In particular, any system using a party list can be critiquedon grounds of accountability (not to mention voter choice): if there is one particularpolitician on a list whom the voters wish to remove, they have in general no mecha-nism to express this wish. They could abandon the party in question altogether, butthis would have the effect only of preventing the election of candidates at the bottomof the list, not necessarily the one of concern.
We have heard some cogent critiques of STV9 on the grounds that it might un-dermine (4) effective parties, essentially by giving the voters too much choice, andthe parties not enough. However, we feel that this concern is misplaced. If STV facil-itates the rise of local stars, who may appeal to voters of their district even if they arenot the favourites of the party insiders, this can only strengthen the party caucuses
8Citizens Talking to Citizens, Ontario Citizens’ Assembly, 20069Notably by Ed Broadbent, in personal communications.
in the long run. It is often said, moreover, that in STV jurisdictions such as Ireland,candidates must run against other candidates from their own party in order to beelected. This is, of course, the consequence of giving the voters more choice. But itneed not lead to fractioning of the parties; indeed, competitions among candidatesfrom the same party is exactly what goes on in nomination contests and leadershipconventions today, and what would go on in any party-list system as candidates viefor positions on their party lists. In spite of some notable recent examples, this doesnot usually tear parties asunder.
Moreover, STV requires a candidate to appeal to the supporters of other can-didates for their second and further preferences. This could tend to create a bettersense of collegiality and cooperation among candidates, rather than the reverse. Thiscould be a cohesive force, both for parties and for larger coalitions, meaning thatSTV could in fact be a force in favour of effective parties, and indeed in favour ofeffective government and an effective parliament as well.
As we have argued at length above, STV may very well outscore even other formsof proportional representation on the criterion of (2a) demographic representa-tion. Indeed the Richard Commission in Wales recommended switching away fromanother form of proportional representation to STV, on the specific ground that itwould facilitate the election of visible minorities, which the existing form of PR wasfailing to do. As we have argued, STV will also be good for women (an essentialcomponent of demographic representation) as it will promote the election of womento both the government and opposition benches. List-PR would also do this, butMMP most likely would not.
Two interrelated concepts are combined in the final criterion, (9) simplicity andpracticality. As we discussed in section 1.2, we believe that STV is the simplestform of proportional representation available. This requires a bit of explanation inthat critics of STV have often railed against it for its supposed complexity.
Under STV, a party needs to hold a nomination meeting in a district and selecta slate of candidates. It does not need to order this slate. This is much simpler thanthe corresponding task under MMP, which requires not simply selecting lists but alsoordering them, a task with a considerable degree of computational complexity, andone which in the real world is likely to be divisive and acrimonious (or, alternatively,centralized and autocratic).
In many ways, the voter’s task is simpler under STV as well. She is presentedwith an array of candidates, some presented by parties in slates, some standing asindependents, and she simply has to find one she likes, and vote for her. If she wishes,she may also make a second selection and a third, etc. By contrast, under any systeminvolving ordered lists, she must study the whole list from each party, and weigh thelists against one another. On the whole, she prefers the candidates of Party A, buther very favourite candidate is on the list of Party B. What should she do? Does itmake a difference whether her favourite candidate is ranked first, third or twentiethon Party B’s list?
From the purity and symmetry of an STV ballot (all candidates are equal andthere is the same opportunity to vote for each one) flows its simplicity. By contrast,MMP ballots have a considerable degree of structure (there are both constituencies
and lists; and they are ordered lists, so not all candidates are equal) and this (unnec-essary!) structure introduces unnecessary complexity.
Likewise, STV is very practical because it provides parties with a task whichthey can carry out without too much complexity or acrimony: nominate a slateof candidates in each district. MMP provides a much more complicated task whichmany parties would not find practical or pragmatic in the least: take all the candidatesacross the province (or, with regional lists, across a large swath of the province). Now,rank them all. Which one is the very most important one? Which one is second?Which third? and so on, all the way down. Although this is commonly done in otherjurisdictions, notably Germany, it is done under central party control in a way whichwould not be acceptable to many accustomed to Canadian political culture.
The results of STV are also simpler to publish an interpret: each district hasa small group of representatives. This is simpler for the voter than reading theresults of an MMP election; a casual reader might only look at the name of the localrepresentative in her own district, and never be aware that her party-list vote hadalso contributed to the election of a long list of candidates printed elsewhere.
Somewhat oddly, the terms of reference of the Citizens’ Assembly do not includegeographic representation, by which we mean a strong link between communitiesand their elected representatives, among their eight criteria. There is (2c) represen-tation by population, but this is not in any way the same thing. Indeed, one vastSTV district electing all 106 MPPs would achieve precise equality of representationby population, but would provide for no local representation at all.
We tend to think that the omission of geographic representation from the list isbecause the committee took its desirability for granted, rather than that they thoughtit was unimportant.
STV is the best system of proportional representation on the criterion of localrepresentation, because it manages with the smallest determinative regions. STV dis-tricts in practice return between 3 and 7 members10, smaller than those typically usedwith list-PR. STV gets away with smaller regions because it uses more informationabout voter preferences; the fact that votes may transfer from party A to party B tomake up a quota in one district, but from party B to party A to make up a quotaelsewhere, means that a a party’s vote-share will match its seat-share better than ifthe votes of aligned parties were hermetically sealed off from one another. This trans-lates to more proportionality for a given district magnitude, or alternatively (and aswe prefer) smaller districts for the same amount of proportionality.
MMP, of course, uses single-member constituencies, just like those we have today.However, these are not the determinative regions, which can be very large indeed.Some models of MMP (such as the one used in New Zealand) aggregate all the votesfrom all voters across the whole country. Even if an MMP Ontario did not consistof one great 106-seat region, one could reasonably expect the regions to be of size 20seats and up. This means that votes from Windsor will be aggregated together withvotes from London, and quite probably with votes from Kitchener or even Hamiltonas well. In parliamentary terms, this means that list MPPs will arrive at Queen’s
10Our 135-seat map includes one with 9.
Park with a responsibility to represent the whole of southwestern Ontario (or thewhole of Metro Toronto), and hardly accountable to anyone.
In STV this simply does not happen. London’s votes are not aggregated withWindsor’s. (In our 135-seat model, the one we prefer, they are not aggregated withany votes at all outside the regional municipality of London.) Every MPP is respon-sible to a clear set of voters: voters who marked her name (not her party’s) on theirballot parties.
1.8 STV and Ontario
A frequent criticism of STV in British Columbia was that STV has not been tried ina jurisdiction as geographically large as B.C. (944,000 km2), and that it is unsuitedto such a vast province.
It is true that that the enormous geographic variances in a province like BritishColumbia, or the equally large Ontario (1,075,000 km2) do put a strain on the electoralsystem. But it is important to realize that this same strain is present regardless of thespecific choice of system. It is present under ’first-past-the-post’: it is difficult havingsome members who can walk to Queen’s Park from ridings the size of St. Paul’s,and others from ridings the size of Kenora, who require a day’s travel to get home totheir constituents. One naturally feels that members from the northern ridings haveto shoulder a greater burden to serve their constituencies.
Such disparities will be present as well under STV, as they would under MMPor any other system. But STV provides a special opportunity to mitigate thesegeographic disparities, by varying the district magnitudes from region to region.
Actually, there are two quantities which we have varied in our model. The firstquantity is the district magnitude, the number of MPPs per district. We would allowthis to range from as low as two in the most remote parts of Northern Ontario to ashigh as seven or nine in the most densely populated areas. The second quantity is thenumber of citizens per MPP, which we have allowed to vary in a generally acceptedrange of +/-25%. Thus, assuming a provincial quota of 100,000 citizens per MPP, anorthern district with 2 MPPs could have a population as small as (2 x 75,000) =150,000, while an urban district with 7 MPPs could be as large as (7 x 125,000) =875,000. In practice, we have not used the whole of this range, nor do we believe aboundaries commission should do so.
There is no existing precedent for an STV system with as great a range of dis-trict magnitudes as 2-7. In Ireland, for example, the range is only from 3 to 5.The precedent comes from British Columbia, where the same range was adopted bythe Citizens’ Assembly as suitable for that province’s urban/rural divide. However,such variability is common in party-list systems; for instance in Sweden, the smallestdistrict has two representatives (the second-smallest has 5), and the largest has 35.There is no reason why a range of 2-7 would not work well under STV.
Incidentally, it is not true that STV has never been used for a jurisdiction as largeas British Columbia or Ontario. STV is used to elect the 34 members of WesternAustralia’s legislative council (senate). Western Australia has an area of 2,645,000
km2, and is divided into just six districts11. Here it has not been thought necessaryto use a wide range of district magnitudes: each district elects either 5 or 7 senators.The largest district has an area of 2,223,000 km2 and a population of only 68,000(under 14,000 per senator); the most populous has an area of under 1000 km2 anda population of 374,000 (over 53,000 per senator). However, the requirement toprovide constituency representation in a unicameral legislature is much greater thanin an upper house, as is the requirement to balance populations per member, so wewould not recommend anything like the Western Australia map for Ontario.
An MMP map in which r of the seats come from lists will require the single-memberconstituencies to be 1/(1-r) times as large as under a pure single-member system.Thus if 1/2 the seats come from lists, constituencies are twice as large; if 1/3 comefrom lists, constituencies are 1.5 times as large; etc.
Faced with this criticism, MMP advocates have suggested in reply that one mightsimply enlarge the legislature until the constituencies have shrunk back to whateversize is desired.
This reply, however, is missing half of the point. Criticism of MMP’s failings inthe dimension of geographic representation are based only partly on the enlarged sizeof the single-member districts. An equally significant problem is that MMP mustagglomerate votes across vast regions in order to allocate the compensatory list seats;this is where its entire claim to proportionality is seated. An MMP system usingdeterminative regions with 12 seats (larger than the whole of Northern Ontario) isusually considered borderline for proportionality; 25 or 50 seats would be more typical.
In other jurisdictions where MMP is used, appropriate regional divisions are avail-able which are relatively homogeneous or political and culturally meaningful. Forinstance, the Federal Republic of Germany allocates list seats to individual states.
In Ontario there are no corresponding subprovincial regions available. There isno clearly defined “Northern Ontario” or “Eastern Ontario”. An MMP system usingregional lists would therefore require defining such entities. Moreover, the entities sodefined would be far from homogeneous. What sense does it make to agglomerateall the votes between Kenora and North Bay, and to use this total to assign listrepresentatives to represent this whole region? Sudbury is a very different city fromThunder Bay, with completely different demographics, culture and economy.
STV does not suffer from this problem; the largest determinative region is a multi-member constituency with somewhere between 3 and 7 MMPs. Thus we have beenable to put Thunder Bay, Sault Ste. Marie and North Bay into separate regions.In Eastern Ontario, we have been able to create a district dominated by Franco-Ontarians, rather than aggregating their votes into a majority Anglophone region, asMMP would do.
We believe that this ability to provide better representation for geographicallydistinct regions is one of the key features which drove the BC Citizens’ Assemblyto choose STV over MMP. A BC-STV map will be able to separate the Cariboo,the Robson Valley, the Peace River district and the Stikine. An MMP map would
necessarily group all these very different regions together into an undifferentiated"North". Precisely the same kind of regional differentiation exists in Ontario.
Moreover, enlarging an MMP legislature in order to reduce the average popu-lation of the constituencies accomplishes exactly this and no more. As we can seeby comparing the two maps in our proposal (the first with 106 seats, and the secondwith 135), STV is more responsive to a marginal increase in the size of the legislature,allowing us to present a 135-seat map substantially different in character from the106-seat map.
2.2 Rightsizing the legislature
Regardless of what voting system is adopted in Ontario, we submit that the legislatureis presently too small for effective governance.
It is tempting to think that any marginal decrease in the size of the legislaturewill save money by decreasing the number of elected politicians and their staff. Andthis is without a doubt true for large legislatures: a Queen’s Park with 1000 MPPswould be smaller, more efficient and cheaper than one with 2000.
However, a moment’s reflection will convince one that such a cost savings cannotcontinue forever. If the legislature had been reduced to 10 MPPs, could the provincesave money by further reducing it to five? Obviously not. Constituencies, alreadyhaving over a million residents each, would double in size, requiring MPPs to shuttleamong more and more local offices, racking up enormous travel budgets and takingon even more staff. Meanwhile, back at Queen’s Park, legislative committees, eachmade up out of the same five MPPs, would be unable to meet simultaneously, andwould be full of overworked, underbriefed politicians.
This simple thought-experiment quickly convinces one of what must be true:there is an optimal size to the legislature, a comfortable middle ground from whicheither increasing or decreasing the size of the legislature would degrade efficiency andtherefore increase costs.
What size is this? The American political scientists Rein Taagapera and MathewShugart1 provide a theoretical analysis which says that the best balance betweenlegislature size and constituency size is achieved by taking the cube-root of the pop-ulation, and then multiplying by a constant, which depends on the nature of thejurisdiction. They provide a number of real-world examples which demonstrate thatthis theoretical model is a surprisingly good model of the real situation which hasevolved organically over time.
In the Canadian context, the cube-root of the population size is an amazinglygood predictor of the size of the legislature, as can be seen in the accompanyingchart. Eight of the ten provinces, and all three territories, lie close to a straight linewhen these two variables are graphed. The exceptions are Ontario, which was muchcloser to the line before the recent reduction in the size of the legislature, and British
1Taagapera, Rein and Mathew S. Shugart (1989), Seats and Votes: The Effects andDeterminants of Electoral Systems. New Haven: Yale University Press
Columbia, which historically was close to the line for decades, but fell off it during aperiod in the mid-20th Century when its population doubled but its legislature didnot increase in size at all.
We can therefore use the data from the other provinces to determine the appropri-ate constant for Canadian provincial legislatures, and thereby "predict" the "correct"size for the Ontario legislature.
There are several choices to be made. First, we might decide to include the datapoint from British Columbia, or we might decide to remove it. In the first instance,we would be using all nine provinces to establish a concordant size for Ontario; in thesecond instance, using eight provinces to establish best-fit sizes for both Ontario andB.C. A second choice is whether the line of best fit should be forced to pass throughthe origin; in other words whether our model should necessarily predict a legislatureof zero seats when a province has zero population.
In the accompanying table, we try each of these options, finding the regressionline as an equation relating the size of the legislature (L) to the cube-root of thepopulation (P). We also quote the coefficients of determination, r2; the fact thatthese are close to 1.00 shows that the linear models are good fits. Under our variousassumptions, we find that the Ontario legislature “should” have variously 130.6, 134.8,134.5 or 142.3 seats. We have adopted 135 as a number representative of the middleof this range.
Another option, which we have not considered, would be to include the threeterritorial legislatures. As can be seen from the chart, they also fit the cube-rootmodel closely, so their inclusion would not greatly alter the predictions.
2.3 STV for a larger legislature
The 2001 census population of Ontario was 11,410,046. In a 106-seat legislature, thisimplies an average population per seat (quota) of 107,642. In a 135-seat legislature,the quota drops to 84,519.
A relatively-high population census division like Middlesex County (population403,185) is not large enough to support a multi-member district on its own in a106-seat house (having well under 4 quotas), and so we have no choice but to includesome of the surrounding area to build a larger region. This situation characterizes themap we have drawn for a 106-seat legislature. Each of the regional centres (Windsor,London, Waterloo, Hamilton, etc.) is the anchor of an enlarged district which includesboth urban and rural portions.
While there is nothing wrong with this, it is something which enlarging the leg-islature allows us to avoid. With Middlesex County now almost at five quotas, it isable to sustain a five-seat district on its own. This means that we can now breakWindsor, London and Waterloo off into purely urban districts, while the rural areassurrounding them have been regrouped into their own three-seat districts.
The creation of these southern three-seat districts means that overall districtmagnitude has decreased. Whereas our 106 seats were grouped into 19 districts, foran average district magnitude of 5.58, the new 135-seat maps has 27 districts, for anaverage magnitude of exactly 5.0.
This is not mere caprice on our part. Rather, we claim that a larger legislaturecan deliver the same proportionality using smaller district magnitudes.
A thought experiment will help understand this point. Imagine a city electingfive members in a single STV district, and in which an upstart party has 12% of thevote. This is unlikely to be enough to capture one of of the five seats. Now supposethe number of seats in the city is doubled to 10, five to be elected on the west side,and five on the east. This party will most likely be stronger on one side than theother; suppose its 12% is made up of 16% on one side of the city and 8% on the other.This 16% will most likely be enough to capture one of the five seats where the partyis stronger.
In this way, although the average district magnitude being employed has notchanged, the system has become more proportional. An increase in the size of thelegislature will enhance proportionality, even if average district magnitude is fixed.
So, if the legislature is enlarged and district magnitude is fixed, the result isgreater proportionality. Alternatively, one can imagine fixing the desired level ofproportionality, and so decreasing the average district magnitude.
Thus the populations of our districts in the 135-seat model have been decreased intwo ways: both the quotas (population/MPP) and the district magnitudes (MMPs/district)are smaller. The population per district is the product of these two ratios.
MMP does not have the same flexibility. An MMP map with 135 seats will lookmuch like one with 106 seats; the single-member districts will just be about 20%smaller.
By contrast, an enlargement from 106 seats to 135 seats enables a fundamentalchange in character in our STV map. Whereas mixed urban/rural districts predomi-nated in the former, the latter separates urban from rural areas, and provides distinctrepresentation for each. Accordingly, the second map makes better use of the flexi-bility of STV to use different district magnitudes side by side. In our first map, thevast majority of voters lived in 6- or 7-seat districts. In the new map, half (67) of theMPPs come from 3-, 4- and 5-seat districts.
The downside to using smaller district magnitudes is that it is harder to balancepopulations; a deviation from an even multiple of quotas forms a bigger discrepancyper MPP if the district magnitude is smaller. This greater variability can be seen inour tables. We take a philosophical stance on this: as there are more representativesto go around, districts which are slightly over quota are not, we feel, unduly disadvan-taged. We have taken a similar position with regard to the balance between NorthernOntario and Southern Ontario: in the 135-seat model we have allowed northern dis-tricts to dip even further below quota, adding a full three northern MPPs, a 33%increase. This is slightly disproportionate; on the other hand 26 MPPs are beingadded in southern Ontario so no region should feel particularly underrepresented.
2.4 Constructing the 106-seat map
In the projected British Columbia referendum of 2009, voters will have the unprece-dented luxury of selecting between two electoral systems, SMP and STV, with accom-panying maps. These maps will have been constructed by a boundaries commissionafter extensive public consultations and input.
Such is manifestly not the case in our sample maps, and readers should notmake the mistake of thinking our examples are in any way definitive. They arefor illustrative purposes only; a proper boundaries commission, with access to inputfrom local residents and political actors, could surely do much better. Nevertheless,we think that they usefully illustrate several important points.
Any districting, whether into single-member or multi-member districts, is neces-sarily a balancing act involving tradeoffs among conflicting criteria. The two mostimportant ones, and also the two in greatest conflict, are (1) keeping adjacent districtsrelatively equal in size, and (2) respecting natural boundaries.
Here we say "adjacent districts" because it is often considered that different stan-dards should apply for urban districts, rural districts, and remote (usually northern)
districts, and indeed we have adopted that view. But districts clustered together inone part of the province should, to a reasonable extent, each have the same population(or, in the present case of multi-member districts, the same population per MPP).
On the other hand, we would also like the dividing line between two districts tofall, not arbitrarily between the houses of the two families in the exact middle of thepopulation, but at some natural boundary, which might be a geographic boundarysuch as a river or a freeway, or might be a jurisdictional boundary, such as the linedividing two wards in a city, or two cities, or two counties.
In the United States, constitutionally, the population criterion dominates, leadingto districting maps with census populations all aligned to within one percent, andthe most arbitrary placement of boundaries. In Canada, it has historically beenconsidered that community of interest is also an important criterion, and we havetended to this latter view. (Conveniently, this also makes our districting task mucheasier, as we have been able to build our districts up out of chunks such as counties,municipalities, or occasionally federal electoral districts, and the aggregate censuspopulations of all of these entities are readily available. But this is a lucky coincidence,and not our primary motivator.)
In our calculations, we have chosen to work with the 2001 census data throughout.In many regions, these figures are already out of date, but at least it gives us acomplete set of data at both the census division and the municipality levels. We havenot attempted to calculate the 2006 populations for any of our proposed districts.
Northern Ontario There is not generally a lot of room to manoeuvre in North-ern Ontario. Clearly, in such a sparsely-populated area, one wants to use districtmagnitudes as small as possible; as three is generally considered to be the smallestdesirable district2 , we have used this minimum throughout the region. Moreover,the existing ridings are all built to sit just around the tolerable limit for deviationbelow quota. We have therefore let the commission which established these ridingsbe our guide as to what is acceptable, and built districts largely by aggregatingridings. Our Lakehead-Kenora corresponds to Kenora plus the two Thunder Bay rid-ings; our Algoma-James Bay to Sault Ste. Marie, Algoma-Manitoulin-Kapuskasing,and Timmins-James Bay; our Nickel Belt to Sudbury, Nickel Belt and Nipissing;and our Algonquin to Parry Sound-Muskoka, Haliburton- Kawartha Lakes-Brockand Renfrew-Nipissing-Pembroke. However, we have made adjustments so that ourdistricts follow the jurisdictional boundaries rather than the riding boundaries; forinstance, the whole of Kenora is in Lakehead-Kenora; the easternmost portion doesnot follow Timmins-James Bay into Algoma-James Bay.
Eastern Ontario Starting at the eastern border of the province, one quicklysees that even a three-seat district could not be assembled without cutting into thecity limits of Ottawa. Ottawa will therefore have to be divided into two districts; wehave chosen to do this so that the wards with the heaviest concentrations of Fran-cophones are grouped with the two counties having large Francophone populations.
2The British Columbia assembly recommended a limit of two, to be used in extremecases. It will be interesting to see whether the British Columbia boundaries commissionmakes use of this.
We kept the district magnitude here down to five, in order to assure the largest pos-sible Francophone concentration. The western part of Ottawa becomes the core of aseven-seat district; as this extends from Parliament Hill in the east to Kingston in thewest we have named it "National Capitals" by way of an impromptu history lesson3.The remainder of eastern Ontario becomes a somewhat ungainly district stretchingfrom Oshawa to Peterborough; a boundaries commission might well want to opt forsome smaller districts here, but this is a judgment call. Oshawa-Peterborough is alsosomewhat over population quota; it could be rebalanced with Algonquin at a cost ofviolating jurisdictional boundaries.
Southwestern Ontario In this region, one quickly discovers that Ontario geog-raphy is dominated by peninsulas; almost everywhere one is forced to "start in thecorner" and work outward until a reasonable district has been established. Beginningat the Niagara River, one has not reached four quotas before encroaching on Hamil-ton; as we have to take some of it, we opted to take most of it, and constructed anurban district with a large DM. We have done much the same throughout the region:Kitchener/Waterloo, London, Windsor, Guelph and Barrie each becomes the anchormunicipality for a large district adjoining rural areas to the city.
Toronto CMA Heavily urbanized regions should have district magnitudes aslarge as possible, to maximize proportionality. Conveniently, the former Toronto andEast York make a nice tidy seven-seat district, just 5% above provincial quota, whileYork and North York make a district just 0.7% above quota. On the other hand,Scarborough has to pick up a bit of population outside the city in order to reach7 seats, and Etobicoke has to be joined with quite a large part of Brampton. Thisleaves Mississauga to join with the bulk of Halton to form a district we have namedLakeshore; thus Lake Ontario joins the other great lakes Superior (“Lakehead”), Huronand Erie in becoming the identifying feature of a district. The district contained in theRegional District of York has been left at six MPPs; we assume this would increase toseven after the 2006 census, hopefully with no need to adjust the district boundaries.
It will be noted that the decision to underpopulate the twelve seats in NorthernOntario (especially the six most remote ones) means that southern seats will be onaverage slightly above the provincial quota. Indeed, it is probably a good idea torecalculate the quota: removing 1,018,951 people and 12 seats for Northern Ontarioleaves 10,391,095 people and 94 seats for Southern Ontario, and a quota of 110,544 inthis region. Our proposed districts range from 5.6% below this quota (Huron-Guelph)to 5.6% above it (Oshawa- Peterborough); most are within 3% of it.
2.5 Constructing the 135-seat map
As discussed above, we believe that 135 seats is an appropriate size for the Ontariolegislature. We also discussed how a larger legislature would permit slight smaller
3We had some trouble calculating the exact populations in the two halves of Ottawa,because the commission which recently redrew their ward boundaries did so on the basisof 2004 population estimates, rather than the 2001 census data. We have run their growthestimates backwards to arrive at approximations of the 2001 figures.
district magnitudes, so we have used more 3-seat districts in rural areas.As we warned above, this has led to a greater variability in population per MPP.
A Southern Ontario population of 10,360,377 divided among 119 seats gives a regionalquota of 87,062. Elgin Lambton is 8.4% below, and London 7.4% below, this quota;on the other hand Oxford-Brant is 6.7% above it. As these are all adjacent, theycould be rebalanced against one another.
The largest deviation from the regional quota is Brampton-Vaughan at 6.8%.Moreover, York, beside it, is at plus 4.8%. Presumably this situation would best bealleviated by the addition of extra seats after the next census.
However, thirteen of the 22 southern districts still lie within 5% either side of theadjusted southern quota.
Northern Ontario We considered, following the British Columbia recommen-dation, of going all the way down to a two-seat district. However, Kenora and RainyRiver alone have a combined population of only 84,000, whereas about 120,000 wouldbe necessary to sustain two seats. But adding Thunder Bay would take us far beyondthree seats, whereas splitting Thunder Bay seems undesirable, so instead we tried tocreate an entirely rural district stretching across the whole northern boundary of theprovince. Many would opt for a different approach, but we thought this might at leastbe considered. The rest of the north is fairly straightforward: Thunder Bay is intact,but Sudbury has been split. Laking detailed census tracts for Sudbury, we could onlyguess at an appropriate division; in any case, so many options are available in urbanareas that this might best be left to the reader. We have brought Peterborough intothe north in order to avoid splitting it; thus at the cost of a slightly expanded def-inition of "Northern Ontario", its seat allocation has been increased from twelve tofully sixteen.
Eastern Ontario We have again created a district with a significant Francophonepopulation. This time, the Ottawa city limits are sufficient to construct the nextdistrict over. As it is the western part of Ottawa where rapid growth is concentrated,we have left it somewhat below quota4. Lanark-Renfrew is our first example of athree-seat district in rural southern Ontario, separated from and lying between theOttawa districts and the one containing Kingston.
Southwestern Ontario As discussed previously, our ideal distribution here isto have all cities in urban five-seat districts, and all small towns in rural three-seatdistricts. In practice, we have come close to this ideal. It is awkward to build anappropriate district around Windsor5. Ours has gone over quota in order to avoidsplitting Chatham; of course this also crosses a jurisdictional boundary and puts somerural communities into an urban riding. Another problem is with Haldimand-Norfolk,now a single-tier municipality. We have wound up with Haldimand in Niagara district,while Norfolk votes with Oxford-Brant. But the rest is quite satisfactory.
4As before, the Ottawa population counts are estimates, found by working backwardsfrom the 2004 ward population estimates.
5Actually, Windsor plus Sarnia would do well, except for being split into two non-contiguous parts.
Toronto CMA The amalgamated Toronto again split quite nicely into separatedistricts based on the six former municipalities, this time also respecting the outerMetro boundaries. However, this does come at the cost of creating a single 9-seat dis-trict; these were formerly used in the Irish Free State (1922-1937), but have not beenused in the Irish Republic (founded 1937). It does not seem unmanageable for down-town Toronto, which besides being a media centre has a history of electing high-profilelegislators, so we might expect its voters to be able to select from among candidateson a slightly larger ballot. The districts around Toronto, Halton-Guelph, Mississauga,Brampton-Vaughan, York, and Northumberland-Durham, all seem relatively naturalas well.
Maps and tables
3.1 The 106-seat map
We here present maps and population counts for an STV model with 106 MPPs.There are 19 districts. The district magnitudes range from 3 to 7, with an averageDM of 5.58.
In the table, we list the total population for the district, followed by the numberof MPPs (district magnitude), the average population per MPP, and the amount bywhich this varies from the provincial quota, which is 107,642.
3.2 The 135-seat map
We here present maps and population counts for an STV model with 135 MPPs.There are 27 districts. The district magnitudes range from 3 to 9, with an averageDM of 5.00.
In the table, we list the total population for the district, followed by the numberof MPPs (district magnitude), the average population per MPP, and the amount bywhich this varies from the provincial quota, which is 84,519.