The 2016 presidential campaign has already begun, and, inevitably, it is spawning conversation about what happened in 2012.  So it will be important to get the story about 2012 correct, or at least to ground that story fully in the available evidence.

Consider this Twitter conversation among the New York Times’ Maggie Haberman, former George W. Bush adviser Matthew Dowd, National Journal’s Ron Fournier, political writer Robert Draper and pollster Kristen Soltis Anderson:

There is an empirical question here: Did the ads sponsored by the pro-Obama super-PAC Priorities Action USA actually shape the views of voters and the election outcome?  During the 2012 campaign, some commentators were completely certain that these ads mattered.

Focus groups, contra Robert Draper’s tweet, will not give us much purchase on this question.  People are not good at reporting what actually influences their decisions.  The fact that a small group of Ohio voters could remember the “coffin” ad (which was actually known as “Stage”) does not tell us  that this ad mattered.  Even if Ohio voters had said it was the most powerful piece of television that had ever seen in their lives, this would not constitute good evidence that the ad actually mattered.

In The Gamble, Lynn Vavreck and I investigated the effect of the advertising in the 2012 presidential election and in particular attacks on Mitt Romney’s time at Bain Capital (see also our  here, here and here).  Here is what we found:

  • When one side had more ads on the air than the other side, this created small shifts in the polls depending on how big the advertising advantage was.  However, neither side had a consistent advantage for very long.  As Anderson noted later in that Twitter thread, this meant that each side’s money was mainly cancelling out the effect of the other side’s money.  And so the polls were pretty stable.  If you don’t believe me, ask Obama senior strategist David Axelrod: “What’s been interesting to watch is that our data has been remarkably consistent really from last spring forward, and our battleground polls really didn’t fluctuate much. There were times when it would dip to where we had a 2-point lead in the battleground states. There’s one poll over the course that we had a 1-point lead.  By and large, we’ve been 3 and 4 points ahead in the battleground polls.”
  • The effects of the ads wore off quickly.  Most of the effect was gone after 24 hours.  This is consistent with previous published research on political advertising, which has found that the impact of advertising decays rapidly.  Thus, it is unlikely that the ads during the summer — when “Stage” first aired — made much difference on Election Day.
  • The attacks on Romney’s time at Bain Capital were supposed to make it seem like he didn’t care about the average American.   But we found no evidence that either the Bain ads or the news attention to Bain Capital, which spiked in July 2012, affected these perceptions of Romney.  In fact, these perceptions really didn’t change much over the course of the entire campaign.

Of course, the 2012 campaign doesn’t offer us the perfect experiment we might need to determine the impact of a single ad like “Stage” or a particular message spread across several different ads.  There were lots of ads being run at the same time, making it difficult to separate out their independent effects.  And it’s not implausible that campaign messages could matter — see my discussion here and especially Vavreck’s The Message Matters.

But if the message matters, it’s likely not because a single ad had some special sauce.

[Correction: I originally said that Maggie Haberman worked for Politico.  She left Politico last month go to work at the New York Times.]