Different uses for voting need different types of voting.
Most dangers in using Condorcet rules come from the notions held by voters and politicians accustomed to bad voting rules. A prime one comes from voters who might think they can manipulate a Condorcet rule in the usual ways: with decapitation and punishing votes.
For example, Bob, a Bush [Major, Kohl] supporter, may think he can help Bush by ranking Clinton [Blair, Schröder] below Kennedy [Livingstone, Lafontaine]. That has no affect on Bush in his comparisons against Clinton or Kennedy; Bush always wins Bob's vote no matter how he ranks the other candidates. But the change does help Kennedy in his 1 on 1 against Clinton. If many Clinton supporters also rank Kennedy second, he may beat both Bush and Clinton — because of strategic votes that had negative strategic value!
That strategy is called a “punishing vote” (against Clinton) or “burying” the major rival, or “raising turkeys”, giving high ranks to hopeless losers.
Sincere but uninformed voters also may put a little-known candidate in between their favorite and her major rival. In this way a candidate could be elected almost by accident! She might be a political weakling without a mandate to govern, a strong base of support or the respect of a majority on the council.
The CW might be the candidate who offends no one. But this
When voters become familiar with the strength of Condorcet's rule they will not make that mistake.
A quicker solution requires each candidate to win 5 or 10 percent of the first-choice votes to qualify for the Condorcet tally. This threshold excludes very small parties from the tally just as the thresholds for proportional representation do. But it is not likely to work well.
This threshold pushes some voters to decapitate a weak first choice in order to help a lower choice reach the threshold.
Unfortunately a threshold motivates a candidate to focus narrowly on a few voters; that 10% is her first priority. Her second is to position her policies between the big parties — and that might not require a broad appeal. As the swing voter of an ensemble council, she might push a selfish agenda to serve her narrow base of support.
Speaking of uninformed voters, most candidates give speeches full of vague, feel-good symbolism but avoid offering detailed solutions to controversial issues. This lets voters fill in what they want to hear and imagine the speaker is suggesting. Sometimes the candidates as a group don't even raise important but formidable problems — making the campaign an empty ritual. These strategies seem to win votes under all rules but critics claim the central tendency of Condorcet rules might make it worse, more likely to elect candidates who “sit on the fence”.
Several dangers have more to do with the psychology of politicians than of voters. The chief executive is the mayor, governor or president who leads the executive branch. She must have a clear sense of direction and be strong willed to make the bureaucrats pull together. These qualities might be stronger in IRV winners than in middle-of-the-road Condorcet winners.
But when electing a council, the question “Do we want ‘bold leadership’ OR 'the best moderator'?” is a false dichotomy. We can have several reps who are strong advocates for various interest groups AND a chairperson who excels at building consensus. Her role is very different from the chief executive's.
Many democratic reformers prefer the “council and manager” form of government in which a Chief Administrative Officer is hired by the council to execute its decisions and run the bureaucracy. There is no need for a strong mayor. This CAO has great influence through reports and testimony to the council. But the broad-based council has all final power. The views of the CAO should align with the council’s usual majority. So an ensemble council should elect a Condorcet winner, whereas an unbalanced council should elect an IRV winner.
The process of governing presents another danger: often the way to find the best course is to go a little to the right, then a little to the left. Plurality and IRV tend to do that when electing successive chief executives. Condorcet winners should change course only as much as voters do; small shifts in the electorate do not lead to policy flip flops. Which is usually a good thing, but occasionally some organizations may need sweeping changes.
Condorcet rules might worsen the extreme gridlock some Latin American governments have suffered. The legislature is elected by proportional representation, which may elect reps from several parties, none of which win a majority of seats. The voters elect the president directly, following the U.S. model. The president then is not closely tied to the legislature (unlike a prime minister elected from and by the legislature). Some presidents find it very hard to negotiate with the splintered legislature. (see Gary Cox, p. 59 of “Instability? A response to Robert Richie and Steven Hill's 'The Case for Proportional Representation'” ) The president's veto power is key; absent that, this gridlock cannot occur.
Part of the solution, according to Cox, is tightly linking the election of the president with election of the legislature, so the president's “coattails” help give her a legislative majority and make reps seek her favor in coming campaigns. Their elections should be simultaneous and should use equally eccentric voting rules so the president and most reps come from the same party or at least from the same side — so electing this president by IRV is usually best.
(Another part of the solution is to limit the effective number of parties the president must negotiate with. Reducing the number of reps elected from any district may reduce the effective number of parties. (But see “Rethinking Duverger’s Law” by Rein Taagepera and Bernard Grofman. They say effective number of parties relates more to the number of issue dimensions dividing a polity.) Three-seat districts are safest here because they usually produce one or two reps from each of the two major parties and none from smaller parties.)
The chance of a tie (or “voting cycle”) may run about 10% in most electorates. It is easily fixed by any of the numerous “Condorcet-completion rules” such the Schulze method, Tideman's Ranked Pairs, or one of the Condorcet-Hare hybrids.
The final argument against trying the Condorcet rule is that it has never been used by any government, not for elections, and not for policies. (Several organizations do conduct pairwise Condorcet elections.) Its use would threaten the majority of politicians because their views are not central in their districts.
The next two pages look at some of the ways people manipulate most voting systems.