The Best of All Worlds:
Partisan advocates of "ward" and "at-large" should compromise and choose neither

Julian West                                                                        Paul Tennant
Department of Mathematics                                             Department of Political Science
Malaspina University-College                                          University of British Columbia

CONTENTS

Introduction
An examination of the literature
How proportional is SMP?
Administrative costs of SMP
It is time to widen the debate
Proportional voting systems
1. Cumulative voting
2. Alternative voting
3. Single-transferable vote
4. Open party lists
5. MMP with open party lists
STV and the individual voter
Possible voting systems for the proposed GVA and CRA
PR and “Direct Democracy”
Recommendations
 

Introduction

In a widely-reported study for the Ministry of Municipal Affairs and Housing, Patrick Smith and Kennedy Stewart recommend that the ideal system for civic elections in large municipalities would be single-member plurality (SMP) in neighbourhood wards, also known as "winner-take-all or "first-past-the-post" (FPTP). This system is nothing other than the system by which Canadian voters elect the House of Commons and provincial legislatures, so we know it well. Familiarity, however, may be its only real advantage. Smith and Stewart base their recommendation on a direct comparison of SMP to the existing "at-large" (AL) system, finding SMP to be the more proportional of the two. In so doing, they have ignored whole families of possible systems, many of which combine the best features of SMP (neighbourhood representation, accountability) with those of at-large, and so may be more likely to win over advocates of the status quo. 

These more sophisticated systems are also much more proportional than either SMP or at-large, both of which are notoriously non-proportional systems. Smith and Stewart's assertion that SMP is acceptably proportional rests on two foundations: an appeal to authority, and a mathematical analysis. Both of these are highly suspect, as we will show. We conclude by recommending several proportional possibilities, and indicate which are broadly suitable for small/rural, middle-sized/suburban and large/urban communities respectively. Our favourite among these is single-transferable vote (STV), which we examine at more length.

An examination of the literature

A typical academic study of possible voting systems at the municipal level is Leon Weaver's "Semi-Proportional and Proportional Systems in the United States". It appears in the book "Choosing an Electoral System: Issues and Alternatives", which, as its name suggests, should be on the required reading list for anyone charged with designing and implementing a new electoral system. Weaver's paper examines the following options: at-large (AL), single-member districts (SMD or SMP), limited voting, limited nominations, combined LV/LN, fixed ratio, single transferable vote (STV), cumulative voting (CV), proportional representation with party lists. This range of options is typical for an academic study. Smith and Stewart, however, confine themselves to the first two.

In this context, it is useful to examine some of the source literature referenced by Smith and Stewart. Their most extensive, and most prominent, quotation is from Patrick Dunleavy, whom they hail, with some justification, as the "architect of the electoral system for the new Greater London Authority". They do this without troubling to mention that Dunleavy explicitly recommended against SMP, and that the three schemes he forwarded to the UK government for consideration all involved proportional representation. Smith and Stewart (page 5) quote at length from a paper by Dunleavy and Helen Margetts, listing a number of "criteria from democratic theory" which can be used to assess electoral systems. These criteria are grouped under four headings: political equality, representation of viewpoints, accountability, and the importance of elections.

As they have cited these criteria, one would expect Smith and Stewart would apply them to see how their favourite system measures up, but in fact the criteria are scarcely referred to again. One mention is on page 15, where Smith and Stewart state that at-large produces "lopsided" councils and therefore fails to meet Dunleavy's criterion of "adequate representation of viewpoints.” In fact, Dunleavy takes this criterion to include "minorities can win seats; the legislature is socially representative". An examination of Dunleavy and Margetts' paper reveals that they also believe SMP to fail to meet this criterion, and that it is only properly met by PR.

Dunleavy and Margetts, indeed, introduced this list of criteria for no other reason than to see how different systems fare. They found that SMP was strong on "accountability, ability to punish parties" and that governments would likely not fall between elections. However, SMP clearly did not fare well on the criteria of "proportionality and wasted votes". While, conversely, some PR systems (list types) were weak on accountability, other PR systems had no negatives at all, and counted among their positives "proportionality, no wasted votes, minority access and range of choice". In short, by these criteria SMP fails to provide "minority access and range of choice" and strongly fails to provide "proportionality". Almost the only argument which Smith and Stewart make in favour of SMP is its supposed "proportionality" which is here refuted.

Elsewhere (e.g., page 9) they claim to be in search of a system which will provide access for minorities, including political minorities as well as under represented social minorities. If so, Dunleavy and Margetts would steer them away from SMP and toward some form of PR. On page 32, Smith and Stewart state that there is "very little in the way of legislation that can be offered to encourage the selection of women and minorities as candidates for election."

Not so: numerous theoretical and empirical studies have shown that the access of women and minorities to party nominations is significantly enhanced by a conversion to PR. One typical source states "There are many ways to enhance the representation of women, minorities and communal groups. These include multi-member PR or semi-PR election systems which use reasonably large district magnitudes of seven or more members, thus encouraging parties to nominate women and minorities to increase their electoral chances." And again, "PR systems in general ... encourage parties to present inclusive and socially diverse lists of candidates ... to maximize your national vote, regardless of where those votes might come from [and they] make it more likely that the representatives of minority cultures/groups are elected."

When New Zealand changed to a mixed-member form of PR in 1996, the number of women in parliament increased significantly, and the number of Maoris dramatically. Moreover, Asians and Pacific Islanders entered parliament for the first time. In numerical terms, "The Inter-Parliamentary Union's annual study of Women in Parliament" in 1995 found that on average women made up 11% of the parliamentarians in established democracies using FPTP, but the figure almost doubled to 20% in those countries using some form of Proportional Representation. This pattern has been mirrored in new democracies, especially in Africa."

Smith and Stewart next turn to the issue of voter turnout. They state that a voter in a single-member district of 10,000 voters will necessarily have a more effective vote than one in a multi-member district of 100,000 voters. However, this is not necessarily the case. For instance, a city might be divided into ten wards, of which the five on the east side are usually won by party A with 7000 votes, ahead of party B's 3000 votes. Meanwhile, the five on the west side might be won by party B by identical margins. Clearly, there is little incentive to vote in any of these ten wards. However, under at-large voting, each party would have about 50,000 votes total and an individual voter might then have a great influence on the outcome. The degree of voter influence under both at-large and SMP is therefore a function of the configuration of the parties contesting the election, and not simply of the size of the constituencies, as Smith and Stewart naively assert.

The dilemma is neatly sidestepped under proportional representation systems, in which a voter may have considerable influence over the fortune of her party of choice, regardless of whether this is a larger or a smaller party. One would therefore expect there to be a greater incentive to vote under PR systems in general. Evidence to this effect is cited by Arend Lijphart in the very paper to which Smith and Stewart refer on page 14. Lijphart not only points to the problem of low turnout and its disparate effects on low-income neighbourhoods, he also (as Smith and Stewart seem pointedly to ignore) proposes a number of "institutional mechanisms" to solve it: "voter-friendly registration rules, proportional representation, infrequent elections, weekend voting” and holding several elections concurrently (emphasis added). Lijphart states that "recent comparative studies have estimated that the turnout boost from PR is somewhere between 9 and 12%", quoting Andre Blais and Ken Carty as well as a number of other sources. It is fair to say that while "the at-large system"s effect of depressing voter turnout has been widely recognized" (Smith and Stewart, 13), exactly the same criticism may be leveled at SMP.

Finally, Smith and Stewart quote Lijphart: "Single-member districts do not make majoritarian systems into proportional ones, but limit the degree of disproportionality." It is important to realize that this sentence only means that single-member systems have a tendency to be more proportional than at-large systems. The idea that they actually attain a standard of proportionality, however weak, is specifically refuted. Furthermore, the reference is only to long-term average behaviour; SMP in single-member districts does not include any juridic limits on disproportionality. Thus, for instance, the Chretien Liberals were able to capture 101 of 103 seats in Ontario in 1997, despite not even having a majority of the vote. In 1987, the McKenna Liberals in New Brunswick won 58 out of 58 seats. It is not possible to be more disproportional than this; in fact, these examples are worse than recent experience in Vancouver under at-large, where the NPA has won "only" 51 of the last 52 contested board seats. There is every reason to suspect that SMP in a B.C. municipality would be just as drastically disproportional, as we shall see in the next section.

The very next sentence in Lijphart's book is even more revealing: "The exact degree of disproportionality and of the discouragement of multipartism that remains will be analyzed in chapters 4 and 5." Chapters 4 and 5 of Lijphart's book are extremely technical empirical studies and are not especially relevant here, especially as SMP is not a majoritarian system and so is not directly addressed by Lijphart's statement. However, SMP does share with majoritarian systems both a strong degree of disproportionality and a very strong discouragement of multipartism. We see this as a significant peril in British Columbia, which has a strong multiparty tradition. It is especially worrying at the municipal level, where party systems are just emerging: to force a city into a two-party system before the parties have even emerged is a significant curtailment of voter choice.

In summary, Smith and Stewart do not make use of the standard literature on the subject of electoral systems, including even the papers which they themselves cite. Their comments about PR are limited to such statements as "while systems based on proportional representation are better at giving voters what they ask for, Canada's single member, first-past-the-post system stands well in comparison." (page 8) In other words, their support for SMP is supposedly conditioned on an understanding that it is reasonably proportional in comparison to AL, which they describe (page 15) as "the winner-take all at-large system". Most sources likewise describe SMP as a "winner-take-all" system, and none describe it as proportional. In the next section we refute the claims that SMP is reasonably proportional. According to their own statements, this should lead Smith and Stewart to reject SMP and join the search for a system which is "better at giving voters what they ask for".

How proportional is SMP?

Smith and Stewart analyze the single-member plurality system using a well-known measure of disproportionality, the least-squares index introduced in 1991 by Michael Gallagher, and championed by Arend Lijphart. This is a well-accepted and useful measure of disproportionality, though not the only one in use. For instance, Dunleavy and Margetts generally use the older and better-known Loosemore-Hanby index. One interpretation of these indices, favoured by many authors including Dunleavy and Margetts, is that they are indications of the percentage of voters whose votes are ignored by the system. High percentages are thus obviously not to be taken lightly. We will return to this subject below.

First, here are two examples to show how both indices work. The first presents the results of the 34 B.C. ridings in the federal election of 1997.

 
 
votes %, vi
seats %, si
di = si - v
di2
Reform
0.43
0.74
0.31
0.093
Liberal
0.29
0.18
- 0.11
0.013
NDP
0.18
0.09
- 0.09
0.008
PC
0.06
0.00
- 0.06
0.004
Green
0.02
0.00
- 0.02
0.000
other
0.02
0.00
- 0.02
0.000
 

 

 

 

 

 

 

The Loosemore-Hanby index is one-half of the sum of the absolute values in the third column of numbers, and in this case is equal to 31%. This represents the total amount by which the over represented parties (in this case, only Reform) are over represented, which is necessarily the same as the total amount by which the underrepresented parties are underrepresented.

The Gallagher least-squares index is the square-root of half the sum of the numbers in the rightmost column, and in this case is equal to 24%. This represents roughly the same thing as the Loosemore-Hanby index, except that the degree of misrepresentation of each party has been weighted by the size of the party itself before averaging. This makes the index less sensitive to the presence or absence of smaller parties.

The second example derives from municipal politics. It is necessarily hypothetical as SMP wards have not been used in B.C. municipalities for some time. Following the 1996 Vancouver civic election, the Vancouver Sun estimated that, had the election been contested under a ward system, the NPA would have won 9 council seats and COPE would have won 1. All other parties would have won no seats. We can take these as our hypothetical measures of si (seats). Measuring vi (votes) is harder since under at-large each voter had 10 votes and some parties nominated fewer than 10 candidates. It is therefore harder to predict how voters would have behaved given only a single vote. In the following table, we have settled on reasonable measures of vi which will serve for illustrative purposes. The Loosemore-Hanby and Gallagher indices will be fairly resistant to reasonable upwards or downwards revisions in these numbers.

 
 
votes %, vi
seats %, si
di = si - v
di2
NPA
0.47
0.90
0.43
0.185
COPE
0.30
0.10
- 0.20
0.040
VOICE
0.10
0.00
- 0.10
0.010
Green
0.09
0.00
- 0.09
0.008
LWP
0.02
0.00
- 0.02
0.000
others
0.02
0.00
- 0.02
0.000
 

 

 

 

 

 

In this instance, the Loosemore-Hanby index is 43% and the least-squares index is 35%.

The obvious feature of the examples is that both are massively disproportional. The number of seats awarded has seemingly nothing to do with the number of votes obtained, except that (1) there is, at least, no party which received fewer votes yet more seats than another and (2) the largest party has been awarded a massive overrepresentation in seats at the expense of all other parties. This is absolutely par for the course with SMP. Indeed, as recent experience in British Columbia has shown, the party with the largest vote share does not necessarily win the largest number of seats. In 1996, the second largest party, with 39% of the vote, not only received more seats than the party with 42% of the vote, but actually formed a majority government. (Intriguingly in the context of municipal government, the same situation prevailed in the City of Vancouver: the NDP received fewer votes than the Liberals in the city, but won 6 of the 10 "wards".)

These two examples, about as disproportional and arbitrary as it is possible to imagine an electoral system as being, give a gauge for how large we can expect the least-squares index to get. It appears that 24% is a bad score and 35% is an extremely bad one. About the worst possible score can be estimated by recalculating the second example with all 10 seats awarded to the NPA (as they indeed were under at-large). This gives a Loosemore-Hanby index of 53% and a least-squares index of 44%, slightly but by no means spectacularly worse than the 35% which could be expected under SMP.

Our gauge for interpreting deviation scores is confirmed by none other than Patrick Dunleavy and Helen Margetts, who state:

To avoid misinterpreting the DV [i.e. Loosemore-Hanby] statistic it is very important to note that in a liberal democracy the index does not even theoretically have a maximum of 100 per cent. A 100 per cent DV score would mean that all the seats in parliament were allocated to one or more parties who got no votes at all, while no seats were allocated to any party that did win votes! A political system which achieved such a result could not conceivably be called a liberal democracy. So what then is a practical and relevant view of the maximum disproportionality possible in a liberal democracy? Our suggestion is to envisage a situation where the party with most votes automatically got all the seats in Parliament, whatever its vote share. A political system run on these lines could just be viewed as democratic -- after all, this exaggeration of the largest party's vote share only accomplishes in a somewhat more extreme fashion the normal effects of "first-past-the-post" elections. [emphasis added] Now on our criterion the maximum DV score attainable in a liberal democracy would always be 100 per cent minus the largest party's share of the vote. ... Thus in our City of Vancouver example, the largest value of the DV statistic would be 53%, as we have calculated. For the last BC election it would have been 58% -- that is, 100 per cent minus the 42% of the vote secured by the Liberals. We have already seen that the least-squares statistic generally comes in about 10 percent lower than the DV statistic -- in these two examples it would be 44% and 50% respectively. Recognizing that it is especially unlikely that a party with only 42% of the vote would sweep all the seats against a second-placed party with 39%, we regard this 50% as an especially unlikely value of the least-squares statistic; let us therefore set the maximum possible value of the least-squares statistic to be in the ball-park of 45% for a typical election. By this standard it is even clearer that 35% should be regarded as a very bad score indeed.

This is not, however, the conclusion of Smith and Stewart, who report that 35% is an acceptable score, and indeed that just this number should be considered as an international standard for acceptable disproportionality. They claim to have adapted this notion from Dunleavy himself, who is said to have reported that a score over 35% would be "unacceptable in a liberal democracy". The quotation from Dunleavy and his co-authors, above, makes clear that this is not intended to imply that all scores under 35% should be considered acceptable, much less that they should be considered proportional. Dunleavy's own guidelines for proportionality are much lower: “the 4 to 8 per cent range of disproportionality ... is generally held to be acceptable for a working PR system.”

Dunleavy also provides a useful interpretation of what the Loosemore-Hanby index means in absolute terms. "The index can be thought of as showing what proportion of M.P.s in Parliament occupy seats that are not justified in terms of their party's share of the national vote" or, even more strongly "on average 28 per cent of voters would have found that the electoral system in 1992 took no account of their vote in allocating seats, more than one in four." When one has this understanding of the Loosemore-Hanby index -- that it represents the percentage of voters whose votes are being entirely disregarded -- it is immediately clear that 35% is not going to be considered acceptable.

Stewart and Smith calculate the least-squares disproportionality for the past 19 BC and 19 federal elections, and find it to be bouncing erratically between about 2% and 28%, with the middle quartiles for Canadian elections close to 10%-20%, and those for B.C. elections somewhat higher. These numbers are clearly outside Dunleavy's range of acceptable proportionality.

In fact, only twice in the 38 elections studies was a score below 8% obtained. Even here, the apparent proportionality on the large scale usually masks more significant disproportionality on the smaller scale. For instance, in the 1997 federal election, Smith and Stewart find a disproportionality score of about 13%. In our first example above, we found that for the British Columbia ridings the score was twice this, 24%. In fact, very high disproportionality in each region has been masked by averaging-out effects over the country as a whole. For instance, the Gallagher index for Alberta was 31%, for Ontario 40%, for Quebec 18%, and for Nova Scotia 29%, giving a national average of about 30%. In terms of the individual voter, approximately 30% of the votes can be said to have been disregarded in composing the parliament, even though the overall disproportionality is only 13%. This effect can be very extreme. In 1994 in Malawi, using SMP, the disproportionality was by pure chance only 2% but the "wasted vote" total was still nearly 25%!

This sort of averaging-out mitigation of the distorting effects of SMP is more likely to happen in large jurisdictions with many constituencies. Since Canada, with about 300, has about 4 times as many constituencies as B.C., we might expect the disproportionality scores for Canada to be typically lower than those of B.C., which is indeed what Smith and Stewart found. An extrapolation should be made: since Vancouver has only 10 council seats, and other B.C. municipalities even fewer, one should not expect the empirical evidence from Canada (n=301) or B.C. (n=75) to have much application to the municipal context. The SMP ward system will presumably be even more disproportional than the up-to-28% which Smith and Stewart warn us to expect. This tallies with our second example, based on the Vancouver Sun study, which shows that the expected disproportionality for the last Vancouver election if replayed under SMP was 35%, worse than anything measured at the federal or provincial level.

It must be emphasized that Smith and Stewart insist on calling such scores low, when we have already seen that they are in fact close to as bad as things can get (45% being almost unimaginably bad). They further insist on describing scores which range from 2 to 28 as "remarkably consistent":

The striking conclusion indicated by these sets of results that occur in a constituency-based system with political parties is that not only is proportionality relatively high, but also remarkably consistent. (page 19) How do they manage to reach such a conclusion? By the simple expedient of taking all the disproportionality scores and subtracting them from 100. While the numbers 2 and 28 may not seem very close together, somehow 98 and 72 seem much more "consistent". This numerical impression is reinforced visually by a set of bar charts with nice tall bars reaching up from 0 all the way to those "consistent, high" scores in the 70s and 80s. However, as we have seen, a score of 0 is impossible, and a score of 55% (that is, 100% minus 45%) is pretty well the floor. Smith and Stewart are therefore committing what is known in statistical terms as a "baseline error".

We suggest the following experiment: take a piece of paper and lay it across their charts at the 55% level. Suddenly their results no longer look either "relatively high" or "remarkably consistent". Furthermore, their threshold of 65% makes it look as though someone has been setting the bar rather low. In the following diagram (Figure 2), we repeat Smith and Stewart's data in this more honest format. We include bars at 96% and 92% to show Dunleavy's strict (4%) and weak (8%) bounds for proportionality, as well as a bar at 65% for the "Smith-Stewart bound". Apart from our formatting, this is exactly the same chart whose data Smith and Stewart describe as "relatively high, but also remarkably consistent."

For comparison (Figure 3), we show the election data for Finland, which has a proportional system, over the same period. We chose Finland because we had the detailed data to hand but it is also a convenient choice as it has a similar population (5,000,000) and geography to British Columbia's. Among geographic similarities are the size (as large as one-third the size of B.C.), northern character, 75% forest cover, and dependency on resource industries. More importantly for electoral geography is a pronounced urban-rural split which leaves many geographically large regions returning few members to parliament. This could be expected to have a negative effect on proportionality, but as can be seen from the chart the Finnish system fares very well.

For a further comparison (Figure 1), we show the recent performance of the at-large system in the City of Vancouver. Because the city clerk's office does not show party affiliation on the official election returns, we could only carry out calculations reliably back to 1988. However, the evidence is clear: although at-large is massively disproportional system, SMP is hardly any better.

 

 
 
 

 

Incidentally, we have claimed that the floor should be seen as 55% because we determined the worst-case score for the 1996 Vancouver election to be 45% (though the 1996 B.C. election scored a little higher). Smith and Stewart found the Gallagher index for the Vancouver election to be 70%. This is impossibly high and simply represents an error. We cannot determine with certainty the nature of their error, nor have they responded to a request for clarification. (One possible explanation is that they multiplied the percentages for each party by the turnout, therefore giving the election result as NPA 15%, COPE 10%, etc. This would produce a number close to 70%. But this is clearly nonsensical.) The gap in Gallagher scores between (BC) SMP and (Vancouver) AL is 28% to 45% and not 28% to 70% as Smith and Stewart claim.

To summarize our discussion of the graphs, we state that Smith and Stewart have exaggerated the difference in proportionality between at-large and SMP in all the following ways:

by choosing the Gallagher index, which tends to produce lower scores than Shapley-Shubik

by reporting "proportionality" rather than the more generally used disproportionality

by setting the baseline of the graphs at a low score which is unattainable even in theory

by calculating the SMP scores for a polity with more seats (n=75) than they use for AL (n=10)

by selecting only the two most recent, anomalously disproportional, results for the AL series

by miscalculating the Gallagher indices for all the AL elections quoted

by inventing 35% as an arbitrary cutoff, without stating that this is the fraction of votes ignored

by not providing a PR system as a basis for comparison

There is one further very significant reason why we can expect even worse performance from SMP in municipalities than we have seen in provincial elections. This is that SMP functions increasingly badly as the effective number of parties increases. Indeed, in a two-party context, SMP does not look all that bad. Dunleavy and his co-authors find that "in the United States, where there is a perfect two-party system in Congressional elections, the deviation from proportionality is very stable at around 7 per cent. So the British system is broadly three times worse at translating seats into votes accurately than the main countries against which we tend to measure our democracy. The major countries which still achieve high deviation scores like Britain's are former British Empire territories which also use plurality rule elections, especially Canada, Australia and India." The same result is found by Taagepera and Shugart, they find the average Loosemore-Hanby index for Canada to be 24.9%, for New Zealand (before PR) 19.0%, for the U.K. 23.4%, for the U.S. 6.7%. Smith and Stewart's statement (page 17) that the Canadian system scores better than the U.S. is thus not reliable.

Saying that SMP works tolerably well in a two-party system, of course, sweeps under the rug the role that SMP plays in creating and maintaining a two-party system. According to Duverger's famous Law, SMP actively discourages the existence of third and smaller parties. Is this what British Columbians want? We submit that it is manifestly not. At the federal level, one-fifth of British Columbians voted for parties other than the largest two; these parties won only 3 seats. At the provincial level, the exact same figures applied, 20% of voters and 3 seats. It can be assumed that even more voters would abandon the main parties if the system were not trying so hard to squeeze all smaller parties out of existence (Duverger's "psychological" reason why smaller parties tend to disappear under SMP).

At the municipal level, the vote is if anything more fractionated. We examine the case of Vancouver, which publishes the most detailed data, although much the same can be said in other municipalities. There is no single dominant opposition party in Vancouver. While COPE was the leading opposition party east of Cambie Street, the Vancouver Greens were the leading party west of Cambie (with the exception of Kitsilano, where the Greens were strong but COPE was stronger). At the level of individual candidates, in most neighbourhoods the top opposition vote-getter was either Mel Lehan or Tim Louis of COPE, but in Dunbar-Southlands it was Frederic Bass of the Vancouver Greens (followed by his running-mate Valerie Jerome), in Shaughnessy, Kerrisdale and Kensington it was Connie Fogal of VOICE (followed by Jerome or Bass), and in Oakridge it was Raymond Leung of VOICE (followed by independent Stephen Chong). If Vancouver were to convert to an SMP system with about 20 wards, each the size of one neighbourhood, we could therefore expect the front-runners to include candidates from COPE, VOICE and the Greens in different parts of the city, in addition to the NPA in all parts of the city. With this many credible party slates we can expect SMP to fare very badly on the standard indices of (dis)proportionality.

It seems particularly unwarranted to impose two-party systems on municipalities where party structures have not yet fully emerged. Yet this is exactly what SMP does. It would be far better to use a system which neither encourages nor discourages party formation, so that voters themselves could decide whether to organize into two parties, or several, or to eschew party slates altogether.

In short, SMP in a community the size of Vancouver or smaller is likely to be even more disproportional than the 20% or so reported for B.C., both because the number of seats is smaller and because the number of credible parties is likely to be larger. In our view, shared by most commentators including Dunleavy, 20% is already an unacceptable score.

Once the level of disproportionality which can be expected from SMP is acknowledged, one is led quickly to reject SMP for exactly those reasons which Smith and Stewart themselves have enumerated:

"outdated electoral systems need to be significantly rethought and modernized" (page 1)

"past ideas ... of "one size fits all" are clearly inadequate" (page 1)

“more incentive to vote" is needed (page 11)

"the primary purpose of electoral systems is to convert votes into seats" (page 17)

"voters do not get what they vote for" under the current at-large format (page 23)

Administrative costs of SMP

In addition to the fact that SMP is simply not a good electoral system, there is another factor which Smith and Stewart do not address: the cost.

Under their proposal, SMP would be introduced in all municipalities of over 25,000 people. At present this amounts to 26 municipalities with a total population (1997 estimate) of 2,680,000, over two-thirds of the total population of B.C. (Additionally, SMP would presumably be extended to Courtenay, Langford, Port Moody and Langley City, all of which could well be over 25,000 by the next census.) These 26 communities have, in total, 180 city councillors.

Therefore, even if the city councils were not enlarged in the conversion to SMP, a boundaries commission would have to be charged with drawing 180 constituencies. This is over twice as many as the present boundaries commission for the province of British Columbia, which is drawing up between 75 and 81 constituencies. Furthermore, the median size of these municipal wards would be just 10,000 people (with most between 7000 and 15,000), which is just one-fifth of the average district size for provincial constituencies. To keep the wards relatively equal in size within each city would require much more intricate information about population distributions at the city-block level, meaning that each ward will require more effort than a typical provincial constituency. It seems fair to say that each redistricting would cost at least twice as much as a provincial redistricting.

Also, while the provincial boundaries commission has to perform its task every second election cycle (typically, once every ten years), municipal boundaries might well have to be redrawn for each election. This is because B.C.'s urban population is growing rapidly and, while some areas of the province are growing faster than others, the disparity increases at finer scales. This is easy to verify using the census data from 1991 and 1996. The 1991 census data would only have been available barely in time to use for the 1993 election, while the 1996 census data reflects the real situation by the time of the next election three years later. Some examples:

1. If wards had been drawn for Coquitlam using 1991 data, one natural ward might have been census tracts 282 and 283 in the southwest corner of the city. With a 1991 population of 12,118, this ward would be just 13% below the city average. A second natural ward might be 287.01/02/03, in the northeast, with 1991 population 15,968 or 14% above average. By 1996, the first ward had grown by 6% and the second by 72%, to respective populations of 12,964 (24% below average) and 27,539 (62% above average). The second ward is now more than twice the size of the first one, just 3 years after the election.

2. If wards were used in Kelowna, a 1991 ward would have a population of about 9,000. Census tract 19, with a population of 5,979, would have been almost two-thirds of a ward. By 1996 its population had almost exactly doubled, while the city as a whole had grown by only 18%.

3. Tract 151 in Richmond with a 1991 population of 9,705, would have been two-thirds of a ward. By 1996 its population had grown 70% to 16,523, fuelling most of overall city growth of 18%. Other tracts in Richmond grew by less than 7% in population and some actually declined.

4. Tracts 187 and 188 in Surrey together would have made an exactly average-sized ward of 30,768 people in 1991 but at 48,278 would have been 18% above average in 1996. Thus, even very large wards are not immune to the problem of uneven population growth.

5. In Abbotsford the fastest-growing census tract increased 159% in the five-year period 1991-96, while several other tracts declined in population over the same period. The city as a whole grew by 21%.

Thus, it is not an exaggeration to say that a municipal boundaries commission would have to repeat its task every three years, again tripling the cost. This could make the average annual cost significantly higher for municipal redistricting than for provincial redistricting.

Furthermore, to whom would a municipal ward boundaries commission be accountable? Obviously it is critical that the commission be demonstrably non-partisan. An obvious possibility, certainly fitting in with any provincial imposition of wards, would be to have a single provincial authority (along the lines of the Assessment Authority) in charge of drawing boundaries within each municipality. The existence and actions of such a body would add a further element of non-accountable decision-making for the local level.

At the provincial level, this the redistribution process includes accepting input from half-a-dozen political parties (as well as the public, whose concerns are mostly about community issues, not gerrymandering), each of which will scrutinize the work of the commission to make sure that there is no partisan bias. But as each municipality has a different set of active political parties, there would effectively have to be a separate commission process for each. And sometimes municipal political parties spring into existence just months before the election, after the boundaries have been set (1996 examples include VOICE in Vancouver and the Delta Youth Coalition).

Finally, there would be an obvious political need to keep wards both compact and finely balanced in size. The latitude permitted by the courts for provincial redistricting is a function of the special circumstances around northern and remote representation, which clearly will not apply in municipalities. Under these constraints, wards would not usually accord with recognized, established neighbourhoods. Some neighbourhoods would be split, rather than united, by ward boundaries, while others will be overwhelmed by amalgamation with a more populous neighbour. Nevertheless, like many ward advocates in Vancouver, Smith and Stewart ignore these factors and equate wards with neighbourhood representation.

It is time to widen the debate.

In the previous two sectionswe have revealed some of the major flaws of the SMP-ward system: it is not proportional, and it would be subject to frequent, expensive and contentious redistricting. However, Smith and Stewart do make raise several concerns that might seem to favour local constituencies: the need to ensure that residents all sections of the municipality will have representation on city council, the need for accountability to the voters, and the need for voters to get to know the candidates for whom they are voting. Smith and Stewart observe that "to argue that elections in large polities be based on a constituency system is to confirm an acknowledged fact for most Canadians". The key thing to realize here is that having a constituency system need not imply using SMP. Therefore, meeting these needs does not necessitate the use of single-member wards.

There are numerous neighbourhood-based systems which are reliably proportional and combine local representation and accountability with "giving the voters what they want". These are systems we should be considering (although we should keep in mind that city-wide proportional electoral systems also produce area representation, and do not require enlarging councils). We describe a range of neighbourhood-based systems in the following section. We are recommending them all for consideration. We do so in order to allow the examination of all major alternative possibilities, and to provide, at no cost to the tax payer, the sort of information that we would have expected from Smith and Stewart.

On page 22, Smith and Stewart quote the results of an Angus Reid survey, in which voters were asked "If you were given the choice of voting for city councillors who would represent all of Vancouver, regardless of where they lived in the city, or voting for a city councillor who would represent your neighbourhood, which would you prefer?" (The "neighbourhood" solution was supported by a hardly overwhelming 52% of respondents.) However, the question is based on a false dichotomy, one that assumes the choice can be made in ignorance of single-transferable vote and mixed-member proportional reveals . On this issue, voters can have their cake and eat it too, as these other systems provide for city-wide voting and election of councillors having their support bases in particular areas.

Smith and Stewart go on to note that the SMP ward issue "has been raised in Surrey, Burnaby and other large municipalities" (in addition to Vancouver). However, they fail to note that "the PR issue" has also been raised by municipal candidates and in fact was placed on the last Vancouver ballot. One wonders what counts as "raising" an issue. Indeed, the very fact that the raising of an issue by some voices is considered worthy of inclusion in their paper, while other equally passionate voices are not, confirms an established and regrettable fact of partisan two-party politics: not all voices are heard.

Given our extensive discussion on the subject of disproportionality in the preceding sections, it should come as no surprise that we share Smith and Stewart's rejection of FPTP at-large, a notoriously non-proportional system. However, it should be noted that there is considerable public sympathy in favour of the at-large system, mostly for heartfelt, sentimental reasons of "civic unity" from observers who have not put much thought into the actual functioning of the system. The good news is that their aspirations for a system with city-wide voting can be satisfied by a system which also provides neighbourhood representation and proportionality. In other words, people with all three views of what a municipal electoral system can accomplish can be satisfied with a single system. And furthermore, there is a wide variety of such systems, providing considerable freedom of choice. While some of these systems have small neighbourhood constituencies (and thus would be associated with larger councils and expensive and contentious process of boundary adjustment), our preferred system, STV, can be used with larger wards or without wards at all — in the latter case, however, still achieving area representation.

Typical of the deeply felt support for the at-large system is the following commentary from the editorial in the Nanaimo Daily News for August 12, 1998: "if Nanaimo had a ward system as it did prior to 1984, voters in north Nanaimo would be denied the opportunity to vote for someone like Coun. Loyd Sherry." It is a good point, but voters in a democratic system need more than the "opportunity to vote" for someone. If there are enough voters with a like mind, a democratic system must provide them with an "opportunity to elect". At-large satisfies the first yearning, but not the second. As the 1996 Vancouver election showed, up to 40% or more of the electorate can support the same slate of candidates without managing to elect a single one. And, as we have seen, the same situation can also hold under SMP.

The simple system of cumulative voting, described below, satisfies the desires of the Nanaimo Daily News to allow all voters a choice among all candidates, while also providing minority representation. Another way to accomplish the same thing is single-transferable vote. Although STV is generally used in constituencies of between 5 and 7 members, it would be perfectly feasible to elect the 8 Nanaimo city councillors in a single city-wide STV election.

Proponents of the ward system point to the need for neighbourhood representation. It should be noted that either CV or STV allows for neighbourhood representation, because a candidate can be elected by assembling a neighbourhood power base. She will then be recognized as accountable to that neighbourhood where she got most of her votes. However, candidates can also assemble their support among people who have something in common besides place of residence -- ethnic background, for instance, or occupation. Voters are free to vote for candidates according to whatever criteria they deem important — in so doing, at the ballot box, the voters decide what issues and candidates they prefer.

SMP straight jackets them into voting for someone who happens to live in their neighbourhood. For instance, a city might contain 10,000 members of an ethnic minority, mostly concentrated into one neighbourhood; it would be quite likely that a ward containing this neighbourhood would return a councillor from the locally dominant ethnic group. The same city might also contain 10,000 -- or even 20,000 -- members of another ethnic group, evenly dispersed across the city. SMP would not give them a chance to elect a member of their own community. CV and STV, however, would give them exactly the same opportunity to concentrate their vote as the geographically-concentrated group enjoys.

Despite the lack of attention which has been paid to electoral systems by British Columbia politicians and media, these sorts of considerations are not academic niceties. Issues of minority representation, fairness and accountability are fundamental questions of democracy. Proper consideration must be given to them if municipalities are to become accountable democracies with expanded powers. And proper consideration will lead to the rejection of both at-large and SMP.

proportional voting systems

In this section we outline a number of systems. With one exception, these are either explicitly or potentially forms of proportional representation. The exception is the alternative vote (AV), which we cover briefly because it is the best solution for single-winner elections, such as those for mayor.

Proportional-representation systems come in many forms. They meet the following criteria, all of which have been clearly championed in B.C. municipalities, and therefore probably have widespread popular support:

1. They permit representation of large individual neighbourhoods or of several adjacent smaller neighbourhoods, and so should appeal to advocates of wards.

2. They permit voters to cast ballots across the city as a whole, and so should appeal to supporters of at-large.

3. They are proportional, and therefore inclusive, providing representation for all major voting groups, and so should appeal to all those who reject the minority monopoly perpetrated by FPTP systems.

3. They provide a high degree of voter choice, and therefore allow elected officials to be held accountable.

Of course, some systems will meet each of these criteria better than others. This is why we believe municipalities should be given the flexibility to design a system to meet local needs, and why we have presented more than one system. We believe our preferred system, single-transferable vote (STV), meets all four criteria admirably well. In fact, we believe it meets the second criterion just as well as does at-large, and the first criterion better than do SMP-wards. (Neither SMP at-large nor SMP-wards comes anywhere close to meeting criterion 3 or 4.)

The city of Vancouver is already divided into 10 "ward"for provincial electoral purposes. In general, if a candidate can amass about 10,000 votes in one “ward,” she or he is successful in getting elected. Now, assuming the provincial ridings were real wards for council elections, consider a popular local grass-roots politician who has assembled a coalition of voters determined to put her onto city council -- perhaps because they have a strong community concern. Perhaps they all live in the West End, and want to see the vehicular traffic restricted to the downtown core. If all 10,000 happen to live in the same ward, she is elected. On the other hand, if they are still tightly concentrated geographically but happen to be divided by a ward boundary, she will not be elected. Why should the second case be different from the first? Shouldn't 10,000 like-minded voters still be able to control a place on council?

Now suppose the 10,000 supporters are spread geographically over a wide area, and are united by a less neighbourhood-oriented concern. Perhaps they are all opposed to gambling expansion, and come from all parts of the city. Shouldn't they still be able to put their champion onto council? This is exactly what proportional representation systems allow. (We do not wish our example to imply that PR provides a haven for single-issue politicians. Quite the opposite. But in cases where voters in a particular part of the city feel strongly about an individual candidate, their support translates into representation. Reality is generally not this clear cut -- there are overlapping groupings of voters, whose existence, and whose preferences, become evident through their individual ballots on voting day.)

We describe several systems of PR -- cumulative voting (CV), limited voting (LV), single-transferable vote (STV, our favourite), open party lists, and mixed-member plurality (MMP). Each of these systems would allow our block of voters to put its favourite candidate on council.

Fundamentally then, proportional representation is about empowering the voter. It means that one vote is worth the same as another -- whether that vote is cast for the largest party, the second-largest party, a small party, or for a determined individual who feels strongly about a particular set of local issues.

Under SMP, one vote is not worth the same as another. Votes for smaller parties are completely ignored; even though there might be 10,000 of them spread through the city, there will never be enough in a single ward to elect a councillor. Voters for large parties count more than voters for small parties. This is so well understood that many or most supporters of smaller parties either stay home, or cast their vote for a party they don't really care for, in a desperate attempt to have their vote count for something. Sometimes, parties can have 20% of the vote or more but still be "small" enough to be squeezed entirely out of office.

This is how bad the situation is provincially under SMP (and locally under the current at-large). We always know, even before election day, that all parties except the largest two will be under represented in terms of seats. For instance, in the last provincial election the Reform Party captured 9% of the vote, and only 2.7% of the seats; the PDA captured 6% of the vote and only 1.3% of the seats. But we never know, even after election day, how much these parties are under represented by, because we cannot know how many voters actually preferred the Reform Party, but voted for the Liberal Party instead. This is how bad the situation is: the day after an election, we don't even know how many voters actually preferred each party. So we can't even calculate how disproportional the system is; we can't say how what percentage of voter preferences were ignored.

It is an almost universal feature of proportional representation systems that this so-called "strategic voting" completely disappears. There is no reason for a voter not to express her true preference. For instance, under STV a voter in Vancouver whose first choice is the Labour Welfare Party can vote for the LWP even if it seems unlikely that any candidates from this party will be elected. If none are elected, her vote is not lost -- she can transfer it to another party (or independent) later. In practice, this might free up so many voters to "vote their conscience' in support of the LWP that someone from that party actually might be elected. The freedom to vote their conscience without worrying about somehow losing their vote is another way in which voters are empowered under proportional representation.

We began this paper by citing Leon Weaver's classic paper "Semi-Proportional and Proportional Systems in the United States". We now quote at length from his conclusions, as we believe them to be highly pertinent.

“How should we choose or design an electoral system? ... the answer to that question may be sought not in a quest for an ideal or optimum system which is best for any situation regardless of circumstances, but in research-validated techniques which enable us to custom-tailor a system composed of features which, in a particular set of conditions, conduce toward results in line with our values.

“Following this line of thinking results in an attitude of acceptance of a fairly broad range of electoral systems provided they fit "the situation"; and a realization that choices between systems involve compromises and trade-offs between various values. A brief explication of this position is best served by examples:

1. The home town. Population: four thousand. Homogenous population. Little or no political competition Nonpartisan ballot and political culture. A community where local government serves primarily as a "caretaker" because there are virtually no issues and little or no social conflict. An AL system is acceptable, and probably preferable. Why not? Since people in one part of town are much like people in the other parts, districting would seem to be superfluous, and so would a PR or SPR [semi-proportional] system.

2. The dormitory suburb. Population: forty thousand. Considerable social stratification by neighbourhoods, but the range in SES [socio-economic status] is narrow and there is no substantial racial or ethnic conflict because practically everybody is a WASP. Rapid growth. Consequently, politics is dominated by the politics of development (planning, zoning, etc.) The present partisan ballot is being replaced in a new Home Rule charter by a nonpartisan ballot. Under the present AL system the minority party, which normally polls forty to forty-five percent of the vote, virtually never elects a council member. It is estimated that with a nonpartisan ballot, party involvement will be low, but probably not lowest on the scale of Adrian's (1959) typology. A proposed SMD system has districts that conform well to existing neighbourhoods, that are reasonably compact, and that contain social mixes which should result in a council considerably more representative of geographical areas, social groups, and the minority party than the present one. Therefore, why not accept the SMD system?

3. Minimetropolis. Population: One hundred thousand. Wide variations in socioeconomic status. Considerable conflict; cleavages concern racial, ethnic, religious, social and residential-development issues. Dynamic population changes. Changing neighbourhoods. Economic decline for industry, but increase for think-tanks and high-tech enterprises. An AL system combined with primaries would result in a council dominated by the largest ethnic group, one committed to traditionalist and conservative instincts, and one which would probably carry out policies which would exacerbate social conflict. These effects might be somewhat mitigated by a district system, but because of the geographic distribution of group strengths the reform political organization and its minority-group supporters would be seriously underrepresented. In such a situation the reasonable alternatives seem to dwindle down to some kind of PR or SPR system, perhaps in multi-seat districts.

4. Maximetropolis. Population: 1 million and up. Conditions somewhat similar to Minimetropolis, but more varied and more complex. Same conclusion.”

Weaver recommends AL for small communities and SMP for larger ones. His break is somewhere between 4000 and 40,000, and his recommendation is therefore in line with Smith and Stewart's suggestion of a breakpoint at 25,000. However, Weaver makes emphatically clear that a second transition, from SMP to PR, is needed between 40,000 and 100,000. Clearly, communities like Vancouver and Surrey have far more in common with Weaver's hypothetical "mini-metropolis" than with the "dormitory suburb". In British Columbia, there is a significant break between the 15 largest communities, each having over 75,000 people and an established or emerging urban core, and those below 45,000. Only 4 communities -- Chilliwack, Maple Ridge, New Westminster and Port Coquitlam -- fall in the gap. In the British Columbia context, we can therefore recommend that some form of proportional representation, ideally one which provides neighbourhood representation as well, should be used in communities of over 75,000 people.

Weaver specifically recommends against designing an electoral system without a careful examination of the situation to hand. In recommending a single electoral system to cover all BC communities with population over 25,000 -- that is, to cover both North Cowichan and Vancouver as well as everything in between -- Smith and Stewart have in fact opted for the "one-size-fits-all" formula which they profess to disparage. At best, they have come up with "two sizes" and assigned one of the two to each community. We would, instead, recommend that each community reject at-large, but find a replacement which fits local needs. We do not recommend AL or SMP for any community. Instead, our recommendations are as follows.

1. Cumulative voting

This system represents perhaps the simplest possible change to AL. In a city with 10 councillors, a voter is still given 10 votes, but is allowed to dispose of them as she sees fit, without the artificial restriction that she give them to 10 different candidates. This solves the dilemma of the voter who cannot find as many as 10 candidates to support, or who feels particularly strongly about one or two candidates. This simple change instantly turns AL into a system which is potentially proportional. If a party expects to have the support of 30% of the electorate, they need only put up 3 candidates, each of whom will collect 10% of the total vote. These candidates are now unbeatable by any strategy used by the other candidates and voters to deploy the remaining 70% of the vote.

We feel that this system would be more popular than the essentially similar limited voting, in which each voter is given only one vote, because of the perception that the voter's 10 votes are being used to determine the makeup of the entire council. Limited voting mimics a CV election in which every voter gives all 10 of her votes in a block to a single candidate.

CV would be a useful system in any municipality, but we particularly recommend it in those smaller "home town" municipalities where Weaver recommends AL. Examples might be Metchosin, Duncan, Hope, Kimberley.

CV might, however, begin to break down in a larger community dominated by parties. This is because it becomes increasingly difficult for a party to ensure that all of its candidates receive approximately the same number of votes. (The difficulties are even greater under limited voting.) In such a situation, the natural thing is to use STV (see below). However, a CV election is very inexpensive to conduct and represents the smallest possible change from the familiar at-large system. It is therefore probably the best choice for smaller communities.

2. Alternative voting

Uses single-member constituencies like SMP. The voter makes a numbered list of preferred candidates. The first-preference votes are counted and, if no candidate has a majority (50%) of the vote, the last-placed candidate is eliminated and her votes are redistributed according to second-preferences. This process is continued until one candidate has majority support and is elected. This voting system essentially mirrors the way in which political parties choose their leaders, but avoids the need for several "rounds" of voting. Unlike SMP, AV allows voters to "vote their conscience" instead of having to choose one of the two front running candidates. Thus (again unlike SMP) it is at least possible to determine after the election what the voters actually wanted, even if they didn't necessarily get it.

There are two major problems with AV. Both are serious, and we cannot recommend this system with much enthusiasm. However, those observers for whom single-member, neighbourhood constituencies are paramount should at least prefer AV to SMP.

First, AV requires single-member districts and is therefore incurs all the same costs of redistricting as does SMP. Second, AV is not a proportional system and should not be used to elect councillors in a community which has, or expects to develop, a mature multi-party system. However, it might be useful in middle-sized communities which have distinct neighbourhoods but where politics is dominated by independent candidates or by two parties or emergent parties. It might be a good solution for a place like North Cowichan, population 26,000, which is divided into many geographically distinct communities.

The main reason we mention alternative voting here is that it is probably the best system to use when there is only one position to be filled. AV should be used in preference to SMP for electing mayors.

3. Single-transferable vote

This is basically AV in multi-member constituencies. Each voter has only one vote, but marks the ballot to show how his or her vote is to be transferred, if necessary. Last-placed candidates are repeatedly eliminated and their votes transferred to higher-numbered preferences. In addition, surplus votes for candidates declared elected are also transferred. This allows parties/slates to balance automatically "their" votes so as to give their most popular candidates (i.e. those who received the largest number of low-numbered preferences from the voters) the best possible chance of election.

This elegant system has many advantages. In particular, it is strongly proportional. It also keeps the emphasis squarely on individual candidates rather than on party slates. It is therefore an excellent system for polities in which many independent candidates compete alongside party candidates. For that matter, it also allows voters to "ticket split" among more than one party; this practice is seemingly popular in B.C. municipal voting.

In a smaller municipality with 6 councillors, SMP could be used in a single "at-large" district. This avoids having to involve a boundaries commission entirely. In a larger municipality, like Surrey, there might be two districts, each returning 4 members. While we are hesitant to recommend enlarging councils, given the costs involved, and also the resulting weakening of council itself (in relation to the executive body), it is the case that enlarging councils is the surest way to optimize direct representation of recognized neighbourhoods.

Thus Vancouver might have five 5-member districts. The cost of dividing Vancouver into 5 parts would be much smaller than the cost of dividing it evenly into ten (or 25) wards. Furthermore, redistricting would be necessary only occasionally -- if one of the five districts became significantly larger in population than the others, an extra councillor could be added for that district, rather than redistricting the entire city. There is no reason that all districts would have to return the same number of councillors. If the number of councillors to be elected for one district became too great (more than 7 or 8, say), then the district could be split in two. This means that in Surrey, where there are already 6 officially recognized neighbourhoods, these could simply be taken to be the fixed boundaries for the constituencies; each one could be assigned 3, 4 or 5 members depending on population. This would make the difficult and expensive work of a boundaries commission entirely unnecessary.

Thus, in this circumstance, STV would strike a balance by providing neighbourhood representation without overly fragmenting the city. In order to get elected, a councillor must line one-fifth (or, under some variants, one-sixth) of the voters in the ward. She or he might accomplish this by having a strong power base in one geographically concentrated neighbourhood, or in one ethnic community, or in a geographically diffuse community having something else in common, such as a commitment to the environment. This therefore allows even stronger neighbourhood representation than the simple division into 5 wards would suggest -- provided only that this is what the voters desire.

Note that this means that STV provides a degree of neighbourhood representation even when there are no wards and STV election is "at-large". In Ireland, where STV is used for parliamentary elections, the members for a given constituency often informally divide it up into their various "home" areas when they open constituency offices.

STV also increases accountability even over SMP, because it gives voters the chance to reject an unpopular councillor in favour of another representative from the same party. Under SMP, partisan voters who disapprove of their party's local nominee have no reasonable option, and parties can protect insiders by placing them in "safe" wards. Not so under STV.

In a city like Vancouver with a pronounced east/west split in turnout, STV using districts would also guarantee a fair level of representation for all parts of the city, even if the change to a proportional system does not have the desired effect of equalizing turnout at an elevated level. Even if city council were not enlarged, 5 councillors could be returned from the west "ward" of the city and 5 from the east "ward", thus solving the turnout problem.

The two drawbacks to STV are that it will require a positive effort to educate the voters about the new system, and that the actual mechanics of the count will be somewhat expensive. Either volunteers will have to be trained to conduct the count properly, or an investment will have to be made in suitable automatic equipment. However, this should not be prohibitively expensive for a large community. And if many cities were to adopt it, there would be considerable economies of scale. The expense involved is slight compared to the cost of non-partisan redistricting for a ward system.

We recommend STV as the most generally appropriate electoral system for municipalities. It provides for area representation, but also allows each voter to select among a wide variety of candidates running across the whole municipality. It allows maximum voter choice, as each voter chooses among all candidates and it allows for maximum voter effectiveness.

4. Open party lists

In a city dominated by parties (such as Vancouver and perhaps Victoria) a possible alternative to STV would be to use a party list in each multi-member constituency. Voters cast ballots not for individuals but for parties; these are totalled in each constituency and the members allocated to the parties according to an agreed PR formula (such as D'Hondt, Ste-Lague or the Swedish modified Ste-Lague). Independent candidates are not ruled out by this system: they simply present themselves as one-name lists. However, the emphasis is on parties.

Because of concerns about accountability, "open" lists could be used in which voters control the order in which each party's candidates are declared elected. There are numerous possible mechanisms. We put forward this system as a PR option, though with somewhat less enthusiasm than for STV, and only for cities which already have strong party systems.

One advantage over STV is that it is simpler to use, since voters only need to select a single party list, and thus to mark a single “X”. The corresponding disadvantage, of course, is that this restricts voter's freedom to split their ballot between candidates for more than one party, or for independents. An intermediate solution has been used in Australia -- the counting system is STV, but voters have the option to select a standard ballot suggested by a party, essentially resembling a party list. This provides both simplicity and flexibility.

5. MMP with open party lists

Mixed-member proportional is the well-known PR system used nationally in Germany and in New Zealand. There are two types of members, those elected in single-member constituencies and those elected from party lists, either city-wide or in large regions consisting of several single constituencies combined together. Typically the list members comprise half the total number elected, though recent work by Dunleavy and Margetts ("Making Votes Count 2") shows that proportional results can still be obtained if this is reduced to as little as one-sixth. The overall composition of council is proportional to the overall votes cast, but there is still strong district representation. In Hungary, the constituency elections use AV rather than SMP.

This is the best PR system for those who are convinced of a need for reasonably small wards. The fear that the overall composition of council will still be dictated by a high-turnout "west side" community can be removed, again by dividing the city in half and assigning the additional members by topping up on each side separately. It should be noted, however, that this is once again a system based on single-member constituencies and would require frequent redistricting. However, the expense might be reasonable in a very large city. Furthermore, the need to keep ward populations exactly balanced is somewhat reduced, since all voters are treated equally in determining the overall composition of council; preserving the neighbourhood character of the wards is therefore more important than equalizing their populations.

This system is a PR option only for very large cities such as Vancouver, Surrey, Burnaby. Also, it is appropriate only where, as in these three, a strong a mature local party system has already developed.

STV and the individual voter

We have already indicated that we believe the single transferable vote (STV) to be the ideal solution for B.C. municipalities. With larger municipalities having either 6 or 8 councillors (that is, not counting the mayor), except for Vancouver with 10, we believe that at-large STV would be appropriate everywhere. We emphasize again that this is not a recommendation against neighbourhood representation, since at-large STV provides for the degree of area representation actually desired by voters, However, multi-member STV wards could be used to structure area representation explicitly. We believe that STV would be a good solution for all municipalities, large or small.

At the level of the individual voter, an STV ballot captures voter preferences better than either an SMP or an at-large ballot. The SMP ballot requires the voter to mark only a single cross. Often, this is not an accurate reflection of the voter's preferences. She might like two or three candidates almost equally well. Alternatively, she might not have a preferred candidate at all, but instead wish to single out one which she really doesn't like. By allowing her to rank all the candidates in order, the STV ballot gives her a better chance to express all her opinions about the candidates.

Likewise, the at-large ballot allows the voter to indicate a number preferred council candidates, but not to indicate which are more preferred. But it is very unlikely that he likes all six equally well, and dislikes all other candidates equally. The STV ranked ballot allows him to express which of the candidates is his very favourite.

The same latitude is given by cumulative voting. If a voter wishes to spread her votes among six equally-preferred candidates, she may. She may, however, have strong feelings about a single candidate; CV gives her a chance to express this by giving all six votes to the one candidate. Or she could give her top three candidates 3, 2 and 1 vote in order. Or use any of numerous other options. STV (and CV) ballots are clearly less blunt instruments for expressing voter preferences.

At the same time, we recognize that STV proposals always suffer from a number of criticisms, many of which seem to be very convincing to those rooted in SMP systems.

Criticism 1. The count may take days; voters want a result on the election night.

Vancouver already uses machine ballots which are counted by computer. The mechanics of performing an STV count are easily performed by computers in seconds, once all the ballot information is entered. Cambridge, Massachusetts, which has used STV for decades, has recently field-tested an entirely automated system.

Criticism 2. Voters do not really understand the mechanics of the count.

While SMP advocates always assert that SMP is a simple system, it is not in fact. For instance, how many voters can clearly explain how the NDP were able to form a government while finishing in second-place in the popular vote? A complete answer would separate two phenomena: forming a majority government with a minority of votes (the respondent should note that this happens more often than not), and gaining more seats with fewer votes (this is unusual but not unheard-of). Then it would need to be explained that the more-seats-with-fewer-votes phenomenon was a function of the relative size of the NDP's pluralities compared with the Liberals' pluralities. Even supposedly well-informed journalists often seem to think it has something to do with the Liberals' wins being in higher-population ridings; having unequal-sized ridings is neither a necessary nor a sufficient condition for the phenomenon.

We furthermore believe that most voters do not realize that the at-large system is a system of block voting. If they did, it might have been rejected as undemocratic years ago.

There is a second reason we say that SMP is a complicated system. Thoughtful voters find it almost impossible to act rationally under it. Are they supposed to vote for their most-preferred candidate? Or are they, rather, supposed to make an assessment of how their neighbours will likely vote, and then vote strategically? There is no accepted answer to this question, although most political elites seem to encourage strategic voting. Effective strategic voting is a very complicated task, involving considerable guesswork.

By contrast, the voting task under SMP is simple: voters honestly rank candidates from most- to least-preferred. There is no reason to deviate from this simple plan; it is neither desirable nor possible to vote strategically. Voters do not need to know in detail how their preferences are translated into a result; what they do need is confidence that this is being done in a fair and honest fashion. STV supplies this. As voters observe a few counts in practice, they will come to understand the count in more detail, probably as well as most understand SMP or at-large today.

Criticism 3. The marking of ballots is so complicated that many are spoiled.

There is nothing complicated about marking a ballot "1,2,3...". Although it may be slightly more difficult mechanically than placing a single "x", the procedure of deciding where that "x" should go is much more complicated (because strategic voting is a rational strategy under SMP) than deciding which candidate to rank "1,2,3...".

The reason STV has led to many spoiled ballots in the past is that elections authorities make the voting task more difficult than it needs to be. For instance, they might require that each voter rank every candidate on the ballot. A ballot is then disqualified if fewer than all candidates are ranked. Even if this is not required, a ballot is usually rejected if, for instance, two candidates are both ranked 4th. In this case, the ballot should be accepted, and at least the first 3 preferences should be considered. If this policy is adopted, relatively few ballots will be entirely rejected.

STV ballots are clearly easier to mark than at-large ballots. If 8 crosses are the maximum to be marked under at-large, it is easy to miscount and mark more. In this case the voter has inadvertently thrown away all his votes (in Vancouver in former years, as many as 1 in 20 voters did precisely this). Under STV the voter can mark as many or few preferences as she wishes, so no particular problem occurs. Criticism 4. Many voters do not wish to express more than one preference

They don't have to. An STV ballot marked with a single "x" should be counted as a ballot with a single first preference expressed. So this option is allowed, even if it is not the best strategy for the voter. Conversely, under SMP, many voters do not wish to limit themselves to a single preference. But this is the only option allowed.

Experience in countries, such as Ireland, where STV has been used for many elections shows that approximately 80% of voters see their first-choice candidate elected, while 95% see at least one of their first two choices elected, and 97% one of their first three. Therefore if a voter marks as few as 3 preferences on her STV ballot, the overwhelming likelihood is that his or her preferences will not be exhausted and the vote will be counted. It should be remembered that there will be multiple candidates running for each party. The voter might therefore wish to vote for only 3 candidates from the same party, and the great likelihood is that one of them will be elected. Most voters will have at least this many preferences.

Voters used to "at-large", where as many as 6, 8 or even 10 crosses may be marked, should have an easy time adapting to STV, in which a ballot can have full utility with as few as three ranked preferences.

Possible voting systems for any GVA or CRA

The report by Smith and Stewart also recommends the creation of an elected Greater Vancouver Authority and possibly also a Capital Region Authority. While we here take no position on this recommendation (although Tennant wishes it noted that he is opposed, believing that healthy municipal communities are a sound and legitimate base for regional government), we do have observations as to the voting systems which might be used were the recommendation to be adopted.

In attempting to devise multi-member constituencies for the proposed new Greater Vancouver Authority, one quickly realizes that, while many municipalities have well-established parties, the party configuration is not the same from one municipality to another. Even communities which might be naturally grouped together within one multi-member district often have separate parties. This makes a region-wide party list system next to impossible. The natural solution is to use STV. This furthermore easily accommodates the independent candidates which still flourish in many parts of the GVRD.

There could be a natural clustering of municipalities into multi-member constituencies, each with a definite regional character. Each would be assigned a number of members under STV by dividing the total population by a GVRD-wide quota, and rounding up. This quota might initially be set at 40,000.

 
 
constituency 1997 population whole or partial quotas
North Shore 165,000 5
Vancouver-Burrard 272,000 7
Vancouver-Fraser 272,000 7
Richmond-Delta 257,000 7
Surrey-White Rock 220,000 6
Langley-Cloverdale 220,000 6
Burnaby-New Westminster 241,000 7
Port Moody-Coquitlam 181,000 5
 

This gives a total of 50 GVRD councillors. We split Vancouver hypothetically along 16th Avenue, but it could also be split in a number of other ways. We also had to separate Cloverdale from the rest of Surrey, but there is some sympathy for this locally. Other than this, municipal boundaries are entirely respected. The advantage of using "natural" boundaries for multi-member constituencies is that redistricting is not necessary; this saves on expense and also avoids the potential for gerrymandering which is inherent in single-member constituencies. If the population in a constituency increases, a new member is simply added; if a constituency reaches 10 members then it can be split into two 5-member constituencies. After sufficient region-wide growth, the overall quota might also be adjusted.

A GVA Chair could be selected by council or elected directly. If elected directly, some system akin to AV should be used. Using a plurality election for mayor risks seeing a mayor elected from a crowded field with a derisory percentage of the vote; such a "winner" would lack legitimacy.

 

In the CRA, the initial quota might be 7,000, leading to something like the following.
 
constituency 1997 population whole or partial quotas
Victoria East 38,000 6
Victoria West 38,000 6
Oak Bay 18,000 3
Saanich Southeast 36,000 6
Saanich Southwest 36,000 6
Saanich Central 36,000 6
Central/North Saanich, Sidney 36,000 6
Esquimalt, View Royal, Metchosin 30,000 5
Colwood, Langford 34,000 5
 

This would give a council of 49 members initially. It is usually not a good idea to have an STV constituency with only 3 members. However, Oak Bay is a relatively homogeneous community and it might work well.

PR and "Direct Democracy"

One of the most exciting developments in the last decade has been the emergence in Canada of the idea of "Direct Democracy" -- shifting power away from responsible elected bodies altogether and placing it directly with citizens. This has been the subject of numerous books and articles, and a major survey in The Economist. More than one Reform Party M.P. has lent support to the notion, and several study groups are active in British Columbia. Finally, the subject is mentioned on the web site of the Ministry of Municipal Affairs and Housing.

"Direct Democracy" is a catchy phrase and can seem more exciting and easier to understand than technical arguments about the merits of various mathematical formulae for counting votes. There is thus some danger that people may get so excited about the seemingly limitless future potential of Direct Democracy that the nature of the electoral system may appear to be a side issue. After all, if all decision-making will reside directly with the people, then elected bodies will become irrelevant and it does not much matter how they are selected.

We reject this argument. Direct Democracy is potentially an exciting development and might be a strong complement to a legitimate, representative electoral system. However, Direct Democracy will be much weaker if the representative institutions remain only weakly representative. Direct Democracy must go hand in hand with a properly representative electoral system.

Consider the following vantage point: the idea of the Direct Democracy movement is to set a balance between two competing ideas of democracy -- representative and "direct" democracy. One might very well feel that this balance could and should be adjusted away from representative democracy, but there is nothing wrong with the idea of representative democracy itself.

The idea of the Electoral Reform movement is to scrap non-representative, disproportional systems like AL and SMP and replace them with fairer and more accountable systems. There is a real sense of urgency about this because there is something wrong with AL and SMP. These systems are broken, right now, and need to be fixed. If "Direct Democracy" is about bringing in a good idea, Electoral Reform is about something more urgent -- it is about clearing out the bad. It is because real injustices are being perpetrated right now, and real voters' voices are being ignored, that we believe the reform of the electoral system should receive more attention than "direct democracy" in the current discussions.

Some people regard the ideas of "direct democracy" and "proportional representation" as two nice ideas competing for attention, and find "direct democracy" to be the more seductive because it seems to have the greater long-term promise, or because it involves the more fundamental rethinking of institutions. But the Electoral Reform movement is fundamentally not about promoting "proportional representation" as a nice idea. It is about exposing the existing systems -- both AL and SMP -- as fundamentally wrong and antidemocratic systems, because they rob voters of their rights to express their opinions freely.

Recommendations

At present, the Municipal Act permits municipalities to choose between at-large and SMP-wards, but not to use any other voting system. This is the exact opposite of the situation we incline toward, namely that these two systems should be ruled out and all other more democratic systems permitted! However, it would still be a big step forward the Municipal Act were simply expanded to allow municipalities much more choice in their voting systems.

This is our primary recommendation:

The new Municipal Act should permit a wide range of voting systems, certainly including cumulative voting, limited voting, STV, and open party lists. (It might also permit other systems, such as approval voting, which we have not discussed in this paper.)

 

We further recommend that the Ministry of Municipal Affairs and Housing: Encourage all municipalities to examine their electoral systems in comparative context

Encourage larger municipalities to select fully proportional systems

Work to educate the public about all forms of electoral systems

Ensure upcoming public consultations are not focused on a false dichotomy between at-large and wards (that is, between SMP at-large and SMP-wards)

Create a Municipal Elections Office with experts in comparative electoral systems, available to design electoral systems in consultation with municipalities

We make the following recommendations to the UBCM: Urge the Ministry not to impose SMP ward systems on any municipality

Urge the Ministry to provide guidelines for electoral systems that empower the voter -- and then permit municipalities to design their own systems within these guidelines

 

We make the following recommendations to municipalities: Consider proportional electoral systems for electing councillors and the alternative system for electing mayors

Urge the Ministry to work with all municipalities to develop modern systems, rather than imposing systems from above

Involve the public to the maximum extent.

 

********

September 21, 1998