Michael Hancock won yesterday's election for mayor Denver, beating opponent Chris Romer in the runoff by a 16-point margin. This comes after a poll last week showing Hancock with only a ten-point lead. An earlier poll showed only a four-point lead for Hancock, and Romer actually had more votes than Hancock in the early May first-round election. How did Hancock do this?
It's a good question, but the consensus answer from Denver's political elites -- that Denver residents turned against Romer because of his negative advertisements -- almost certainly has to be wrong.
For one thing, for advertising to have moved a race from a four-point gap to a 16-point one would be a shockingly large campaign effect. Advertising effects are difficult to measure, although Alan Gerber, James Gimpel, Donald Green, and Darron Shaw came up with a pretty good experiment. In their large-scale field experiment, they found that roughly $2 million in TV and radio advertising could shift roughly five percent of the vote as measured by tracking polls, although this effect was very short-lived, lasting only a few days to a week. Conversely, Hancock's lead over Romer was not only larger than that but grew over time. Keep in mind that the Gerber et al experiment was ideal in many senses. Most campaign advertising is far less effective and is often countered by the opposing campaign.
Second, this purported campaign effect was in the wrong direction. Romer's negative ads (and yeah, they were pretty negative, although to my knowledge they didn't fabricate anything) are alleged not only to have had no effect on Hancock's numbers, but to have boomeranged and pulled down Romer by 12 points. This is highly improbable. In a multi-candidate race, it's possible for a negative ad to be effective but for voters to lose interest in the candidate running the ad, as well. (Arguably, this is how John Kerry won the 2004 Iowa Democratic caucus -- Gephardt and Dean, who had better initial organization, tore each other down with negative ads, and the voters transferred to Kerry.) But in a two-candidate race, it's far less likely that the attack ads will harm the candidate running them, unless the ads are completely off the wall (which these really weren't). It's not impossible for this sort of boomerang effect to happen, but it's rare, and the notion that Denver residents are so fragile that they can't see a fact-based negative ad without clutching their pearls and fainting is rather silly (and, personally, at least slightly offensive).
Let me suggest a more plausible story: Romer's internal polling numbers last month suggested he was in serious trouble. Despite his win in the nine-way first-round race, there was a clear ceiling on his support, while many other voters who had not voted for Hancock initially were at least open to the idea of doing so. Romer realized the only way he could win this was to pull down Hancock's favorability ratings. Hancock, meanwhile, provided Romer with some convenient flubs, and Romer made the best out of them. Flubs rarely cost a candidate an election (recall George W. Bush?), but that was the best Romer had to work with. The ads just didn't work. Meanwhile, Hancock stuck with his ground game of turning out supporters, and that did work.
I don't have the direct evidence to back this up (at least not yet), but this strikes me as eminently more likely than one of the largest campaign effects in human history occurring in the wrong direction.
Update: Some solid analysis here from Patrick Doyle. He makes some reasonable suppositions about the stability of the vote from the May to the June elections, although an exit poll would be really helpful.