Wikipedia defines the spoiler effect thusly:
The spoiler effect is the effect of vote splitting between candidates … with similar ideologies. One spoiler candidate’s presence in the election draws votes from a major candidate with similar politics thereby causing a strong opponent of both or several to win. The minor candidate causing this effect is referred to as a spoiler.
For example, suppose a Democrat is running against a Republican. The Democrat has 51% support while the Republican has 49% support. Then, a Green Party candidate enters. The new distribution of support is Green – 3%, Democrat – 48%, and Republican – 49%. Because the Green is seen as splitting vote with the Democrat, it is seen as spoiling the election, and causing the Republican that actually has minority support (in a head-to-head match with the Democrat) to win.
To avoid this spoiler effect, voters are forced to engage in strategic voting. This means that people who prefer the Green might vote for the Democrat instead, solely to ensure that the Republican loses. This spoiler-effect-driven impetus to strategically vote makes it such that many people can’t vote for who they actually want to win without sabotaging their own interests.
Often, people claim that Instant Runoff Voting solves this problem.
In IRV, voters rank the candidates in order of preference (giving a “1” to their most preferred candidate, a “2” to their second most preferred candidate, and so on). When votes are tabulated, voters’ first preferences are counted up. If, after this initial count, a candidate has a majority of support, the election is over and that candidate has won. If no candidate has a majority of support, then the candidate with the lowest level of support is eliminated, and the second preferences of the voters who ranked the eliminated candidate first are counted. This process of eliminating candidates and recounting the overall vote with voters’ second (and even third, fourth, and fifth) preferences is repeated until someone has a majority.
FairVote explains the rationale for why this is supposed to solve the strategic voting problem:
Because instant runoff voting (IRV) is designed to secure a majority victory, it assures that the so-called “spoiler effect” will not result in undemocratic outcomes. IRV affords voters more choices and promotes broader participation by accommodating multiple candidates in single seat races. Voters may support their favorite candidate without fear of splitting a base of support or swinging the election to their least favorite candidate. Thus it solves the problem of choosing between the “lesser of two evils” and encourages greater participation from voters and candidates, while fostering cooperative campaigns built on a more robust discussion of issues.
But this analysis only holds when the so-called spoiler candidate has a low level of support. Returning to the hypothetical three-way election above, it would likely be the case that IRV would play out so that the Green is eliminated after round one, and then the Green voters’ second preferences would all be Democrat, delivering a Democratic victory by a margin of 51% to 49%.
However, suppose the election margins are somewhat different. Suppose that voters’ first preferences line up such that the Green has 33% support, the Democrat has 32% support and the Republican has 35% support. Under IRV, the Democrat is eliminated and voters who ranked the Democrat first have their second preferences counted. But what are those second preferences? It’s conceivable that the voters are split, with 16% ranking the Green as their second preference and 16% ranking the Republican their second preference. This would mean that, at the second count, the Republican wins over the Green by a margin of 51% to 49%.
You might think this isn’t a problem. After all, the Republicans won a majority of votes. But changing the scenario a little reveals the issue. Suppose that the first preferences lined up this way instead: Green 32%, Democrat 33%, Republican 35%. In this case, the Green candidate is eliminated on the first count, and its voters second preferences are counted. Presumably the vast majority of the Green voters ranked the Democrat second. For our purposes here, let’s assume all of the Green voters ranked the Democrat second. This would mean that, on second count, the Democrat wins in a landslide by a margin of 65% to 35%.
So, as you can see, the election outcome swings dramatically based on which candidate gets eliminated first, which is determined by as little as 1% of the electorate choosing whether to rank the Democrat or Green first. If the Democrat gets eliminated first, the Republican wins. If the Green gets eliminated first, the Democrat wins.
The net effect of all of this is that you still have to strategically vote under IRV. When your candidate has a very low level of support, it doesn’t particularly matter. But once you’ve built your candidate up to the point where it might actually be competitive, you are then forced to abandon that candidate when choosing your first preference. Not abandoning the candidate at that moment risks the possibility that your second preferred candidate (here the Democrat) will get eliminated first and that this order of elimination will cause your third preferred candidate to win. That is, you once again find yourself risking the spoiler effect if you don’t strategically vote against your preferred candidate.
Edit: Rob Richie of FairVote wrote in to note that although this sort of thing is possible, it doesn’t happen that often.