Updated and revised, with some discussion of voter decision-making models prompted by comments from Miguel and A.M. (The first two paragraphs are revised, and the last one is entirely new.)
Lula won 60.8% of the vote in Brazil’s presidential runoff on Sunday, against Geraldo Alckmin’s 39.2. That means that Alckim actually did worse in the two-candidate runoff than he had done in the multi-candidate first round.
In the first round, Alckmin had 41.6%. Put another way, Alckmin lost votes (2.4%) equivalent to about double Lula’s shortfall from a first-round victory (50%-48.6%=1.4%). Below, in the final paragraph and continuing in the exchanges with propagators in the comments, is some (wild) speculation about individual voter decision-making that might have produced this result. It is worth noting that the drop in Alckmin’s support was not a product of turnout differentials, at least not at the aggregate level. The tournout differential between rounds–it was a bit higher in the first round, as is usually the case–amounted to only about 0.17% of votes cast (using the first round as the denominator).
Offhand, I can’t think of a previous case in which one of two candidates in a runoff performed worse than in the first round. If someone has another case, please come forward!
Miguel asks in a comment below something I have been thinking about, too. Is it possible that some voters just wanted to punish Lula by making him wait, but still preferred that he have a second term? Yes, I think that is possible. If so, it would be a form of strategic voting that I am not aware of having been addressed in the literature. Some voters may have decided to withhold a vote from Lula in the first round and strengthen his main challenger, just to force them to debate again and to get a second look at the choices. Such a hypothetical voter would not, of course, be a committed PT or other left voter (there was a PT defector running to Lula’s left in the first round to attract those voters), but rather swing voters who perhaps did not feel too good about either of the viable candidates. In the second round, their “second look” did not convince them that Alckmin was better, after all. We are talking about very small percentages here of net swing away from Alckmin, but it is still the prospect of “second look” strategic voting is an interesting prospect.
As an aside, I just noticed that this orchard reached a milestone: This is its 1000th planting! We’re also closing in on our 50,000th visitor.