Tom Flanagan: The Alternative Vote
A 2001 article on how AV would improve Canadian politics and the benefits of a preferential ballot.
Department of Political Science
The University of Calgary
Votes and Seats:
Opportunities and challenges for Electoral Reform in Canada
Institute for Research on Public Policy
May 2-3, 2001
I have to express a degree of agnosticism about comparing the merits of electoral systems. Fifty years of theoretical research deriving from the Arrow General Impossibility Theorem has demonstrated that there is no perfect method of aggregating votes into a collective choice. The criterion of perfection is the “Condorcet winner,” that is, the option that would be preferred to all other options in a series of pairwise comparisons. For example, in a three-way choice among options a, b, and c, there might not be a majority for any of the three. But a would be the Condorcet winner if a majority preferred it to b and to c. As such, a is the rational choice for the voting group, for a majority of its members prefer no option over a.
All voting methods that reach a determinate outcome are capable, under some conditions, of picking an option other than the Condorcet winner, that is, reaching an outcome that, to a majority of the group, is less preferred than one of the other available options. Proportional representation is an exception, but only because it does not pick a winner at all, at least in the sense of selecting a government or a prime minister. Proportional representation (PR) only composes a legislature, which then provides a backdrop against which party leaders negotiate the composition of a government; and a coalition put together in this way could well fail to be a Condorcet winner if presented to the voters. Would the voters of Austria have preferred the coalition of the People’s Party and the Freedom Party, which party leaders gave them after the election of 1999, to a coalition of the People’s Party and the Social Democrats? We will never know because the people never got to vote on that choice, but there is no reason to assume that the coalition that took power was actually a Condorcet winner.
If theoretical analysis shows that no electoral method is unambiguously superior to all others, common-sense reflection leads to the same conclusion. Great Britain uses first-past-the-post (FPP) or single-member-plurality voting; France uses a two-stage run-off; Australia uses the alternative vote (AV) for its House of Representatives; Ireland uses the singe transferable vote; the Netherlands uses the list form of proportional representation; Germany uses the so-called mixed-member-proportional system; and Japan used the single non-transferable vote until the 1996 election, when it switched to a parllel combination of PR and first-past-the-post which does not give as proportional an outcome as the German model. Although each of these seven countries has a different electoral system, all have been highly successful for the last fifty years. They have had their ups and downs, to be sure; but by world standards they all enjoy political stability, democratic government, and a productive capitalist economy underwriting a high standard of living. I conclude that the success of liberal democracy depends on other factors that all these countries have in common: respect for the constitution; adherence to the rule of law; extension of the franchise to all adults, including racial, linguistic, and religious minorities; and protection of property rights. In this tableau, the electoral system is clearly a secondary factor.
Secondary factors, however, are not necessarily unimportant factors. Although any country may succeed or fail with any electoral system, the choice of electoral system does have important consequences for the political system. It affects the number of political parties and their bases of support, the way in which the parties compete or cooperate with one another during election campaigns, and the way in which they behave when they are in government or in opposition. Although no electoral system is demonstrably the best as a general proposition, it is quite possible that a certain electoral system will be better suited than others to the needs of a particular country at a particular stage in its history.
That is the context for pondering electoral reform in Canada. Although Canada is a highly successful country, it has recurrent problems of political fragmentation, which many scholars, from Alan Cairns onward, have plausibly traced to reliance on first-past-the-post voting. Would another system work better, if not in general, at least for Canada at this point in our history?
To seek an appropriate remedy, we must first be clear about the nature of the malady. Like Richard Johnston, I see Canada’s enduring political problem as a tendency for political parties to fragment along linguistic and regional lines, thus exacerbating regional and linguistic tensions and even encouraging separatist movements. This tendency was apparent as early as 1885, with the formation of Honoré Mercier’s Parti National as a reaction to the hanging of Louis Riel, and again in 1911, with the entry of Henri Bourassa’s Nationalist League into electoral politics. Then came the era of Western-based parties, with the entry into federal politics of the Progressives in 1921 and Social Credit and the Co-operative Commonwealth Federation in 1935. Finally we have the events of recent history, with the creation of the Reform Party of Canada in 1987 and the Bloc Québécois in 1990. Each time a new party emerged, one or both of the historic parties, the Conservatives and Liberals, lost a large bloc of regionally or linguistically concentrated voters. The Conservatives have been more battered than the Liberals, but the latter have also suffered huge losses at times, such as in 1921, when they lost the western farm vote to the Progressives, or 1984, when the francophone voters of Quebec switched en masse from the Liberals to the Conservatives, as a dress rehearsal for creation of the Bloc Québécois.
A century of splitting old parties along regional and linguistic lines and creating new ones has given us the contemporary five-party federal system in which only the Liberals have the breadth of support necessary to win an election. Although they command only about 40% of the vote in a five-party contest, they are undoubtedly a Condorcet winner. Polling data on voters’ second choices make it clear that none of the other four parties could come close to beating the Liberals in a two-party contest. In that sense, the fact that first-past-the-post voting gives the Liberals over half the seats with only 40% of the popular vote is not really unfair; rather, it reflects their underlying strength as a Condorcet winner.
Yet all is not well. Even if the Liberals are a Condorcet winner in the current five-party configuration, it is unhealthy for there to be only one potential winner. A party that can always count on winning is likely to become unresponsive and even corrupt. With the passage of ever more unbroken time in power, the “usual operation” of Jean Chretién’s Liberals could morph into the unrestrained cronyism of the PRI in Mexico and the Christian Democrats in Italy in their heyday.
The underlying cause of this tendency towards political fragmentation is the imposition of a parliamentary form of government with disciplined parties upon a sprawling, diverse country. In contrast, the American executive does not require the support of the legislature to remain in power, and thus political parties do not need to be as disciplined as Canadian parties. Republicans and Democrats coming from various parts of the United States and holding divergent views on many issues can cheerfully coexist within their respective parties, because there is no need for party members to vote the same way all the time. Canada could deal with the fragmentation of political parties by adopting an American-style system based on the separation of powers; but after two centuries of parliamentary government, most Canadians would regard such action as extreme and even treasonous. Tinkering with the electoral system is a less drastic remedy.
The problem, to repeat, is the tendency of Canada’s parties to split along linguistic and regional lines. A century of such fragmentation, culminating in today’s five-party configuration, has demonstrated the power of that tendency. Because it levies heavy penalties upon small parties, the first-past-the-post electoral system tends to discourage fragmentation in general; but it has one exception that actually encourages fragmentation in Canada’s case. New parties with regionally concentrated support can thrive for long periods of time by playing to the voters in their territorial base. In this respect FPP may not be the best electoral system for Canada.
Proportional representation, as proposed by Nick Loenen, Henry Milner, Louis Massicotte, and many others, would recognize the tendency towards fragmentation and carve it in electoral stone. New parties would crowd in to join the five already represented in Parliament. The Greens would almost certainly surmount any reasonable threshold for representation, as they have in many other PR countries. The Canadian Alliance might well split along lines of cleavage between fiscal and social conservatives, or western and eastern supporters. The Liberals might also split if—to take only one example—pro-life Catholic voters concluded they might have more influence as a separate party than as a largely submerged voting bloc within the Liberal Party. If Canada adopts PR, all future governments will be coalitions of two or more parties—perhaps even minority coalitions dependent on other parties for their survival, as has been the case in New Zealand since adoption of MMP.
The style of politics engendered by PR would be novel for Canada. I am sure we could adapt to it and perhaps even prosper with it, but I am not sure that Canadians should make such a radical leap of faith when less sweeping alternatives are available—alternative voting as practiced in Australia and the two-stage run-off as practiced in France. Because these two so-called “majoritarian” systems would have much the same practical effect, I will concentrate here on AV, which is in effect a temporally condensed run-off.
Under AV, most things are the same as in the FPP system to which Canadians are accustomed. The ridings will retain the same boundaries and still be represented by a single person, whom residents of the riding can regard as their own Member of Parliament. Voters will still vote for candidates representing parties rather than for parties as such; there will be no party members appointed by and representing only the party apparatus. Moreover, Canadians are already familiar to some degree with the alternative vote. A few older citizens may remember it from the 1950s, when it was still in use in provincial elections in Manitoba, Alberta, and British Columbia. Others have encountered it in recent years in the process of nominating candidates for election campaigns or choosing a party leader. Even those who have not encountered AV will be aware of the run-off method, which has been widely used for the same purposes and which is, as noted above, both logically and practically similar to AV.
AV differs from FPP only in one respect as far as the voter is concerned. Instead of putting an X before the name of the favoured candidate, the voter is required to rank all the candidates from first to last, marking them 1, 2, 3, etc. (Some versions of AV allow the voter to mark only the first choice, or to stop ranking anywhere before the bottom, but these refinements need not be discussed here.) When the ballots are counted the first time, only the first choices are tabulated. A candidate who receives 50% + 1 or more of the first count is declared elected at that point. If no one has a majority, the candidate with the lowest number of first preferences is removed and all those ballots are transferred to other candidates based on the indication of second preference. The process of recounting, elimination, and vote transfer is repeated until someone emerges with a majority, at which point the victor is declared and the election ceases. With paper ballots, the process can take several hours, or even a couple of days; but if Canada ever modernizes its voting technology, computers could do the recounting almost instantaneously.
If one party is overwhelmingly dominant, or if two big parties compete with each other, AV will give much the same results as FPP; but in multi-party races, AV opens up a new world of possibilities for cooperation. Political parties can retain their separate identities, organizations, and programs while forging electoral alliances with each other by “exchanging preferences,” as the Australians call it. Ever since the end of World War I, when AV was introduced, the Liberal Party and the National (previously Country) Party have been encouraging their supporters to vote first for their own candidate but to rank the other party’s candidate second. In this way, a Liberal can vote Liberal, and a National can vote National, without guaranteeing a victory for the Labor candidate. If the Labor candidate trails after the first count, then the Liberal or National candidate will win on the second count. If the trailing candidate is Liberal or National, most of the transferred ballots will go to the other; and the Labor candidate will win only if he gets a majority on the second count. The net effect is to allow the Liberal and National Parties to run an effective electoral alliance without splitting the vote so that Labor wins a majority of seats.
Over time, the Liberals and Nationals have taken their cooperation quite far, so that they usually sit in a joint caucus, both in government and in opposition, and negotiate a joint platform at election time. The cooperation is sometimes looser, and it has occasionally broken down altogether, as in 1973, when the two parties fell out with each other during a spell in opposition. But the system also allows cooperation and exchange of preferences to be re-established, which happened in time for the 1974 election.
Would AV allow the Canadian Alliance and Progressive Conservatives, with or without the Bloc Québécois, to make an electoral alliance that could defeat the Liberals? Simulations based on data about voters’ second preferences suggest that the present answer is No. Because the Liberals are the leading second choice of Alliance, PC, and BQ voters, the Liberals would actually win more seats under AV than they do under FPP. However, it is wrong to take a purely static view of the consequences of introducing AV. The new electoral system would create a new set of incentives for parties to cooperate with each other in order to beat the Liberals. The Canadian Allliance and Progressive Conservatives would have an opportunity to reconsider their positions so as to make the other party willing to exchange preferences. This should be easier to achieve than a full-scale merger, which Canadian history suggests would be short-lived, even if it could somehow be accomplished. The Liberals and NDP might also decide to exchange preferences in some future election if a coalition of the right proved successful. Given the current situation in Canada, with parties spread out across a wide ideological spectrum, we might end up like France with its run-off system, where multiple parties continue to coexist, but elections are usually fought between coalitions of left and right.
As with PR, the long-term result of bringing AV to Canada would probably be a preponderance of coalition governments. Under PR, however, governing coalitions are put together by party leaders after elections have taken place, whereas under AV the coalitions are formed before elections and the voters get a chance to choose between them. Coalitions tend to be long lasting so the voters become familiar with them. In these respects, AV gives voters a higher degree of control over politicians than PR does.
The objection that AV does not yield proportionality between percentage of popular votes and percentage of seats is certainly true. Its operation in this respect is closer to FPP than to PR. Those to whom proportionality is the highest political value should choose one of the many varieties of PR rather than AV.
In my view, however, proportionality is a purely abstract value of little importance in the real world of politics. I believe what is important is to have governments capable of taking decisive action when needed, and to give the voters an effective choice between governing coalitions. For Canada, what is particularly important is to create a political vehicle that can alternate with the Liberals in government. If a single party can do that, well and good; but the historical record of political fragmentation in this country suggests that such a party will be difficult to create and probably short-lived if it does emerge.
By allowing separate parties to retain their distinct identities while forming effective electoral coalitions, AV offers Canada a way out of its historic difficulties. The theoretical properties of AV, its long record of encouraging electoral coalitions in Australia, and the similar effect of the closely related run-off system in the French Fifth Republic—all suggest that the alternative vote could play midwife to the birth of a new coalition broad enough to win an election. In the first instance, the leaders of the Canadian Alliance, PCs, and perhaps the BQ would get an opportunity to put something together. If they were unable or unwilling to try, they would deserve to fail, and the system would provide positive incentives for other, more cooperative parties to try their hand.
AV is based on an understanding of practical problems, modest in scope, only incrementally different from past practice, and easily rescinded if it does not live up to expectations. It is the sort of reform that a cautious people like Canadians could agree to undertake.