Poli-Sci Perspective is a weekly Wonkblog feature in which Georgetown University's Dan Hopkins and George Washington University's Danny Hayes and John Sides offer an empirical perspective on the issues dominating Washington. In this edition, Hopkins looks at the evidence on whether political advertising moves voters. For past posts in the series, head here.

In the weeks before the 2012 election, pundits debated the impact of President Obama's team's unorthodox advertising strategy and Mitt Romney's side's last-minute “ad bomb.”  Now, with the election returns in, we can begin to assess just what the presidential candidates and their allies got for the hundreds of millions of dollars they spent on TV advertising.

The short answer: surprisingly little.  In all likelihood, even major shifts in advertising would have produced only minor shifts toward the candidate benefiting from those shifts.

First, though, let's get a sense of the strategies actually employed by each side by looking at where they spent their TV money.  Drawing on data from the Washington Post's ad tracker, I mapped U.S. counties in a group of 38 states that includes almost all of the swing states.  (States are omitted simply to make the patterns in the included states, where most of the action is, a bit more visible.)  I used varying shades of blue and red to represent the edge that Obama and his allies or Romney and his allies had in all national television advertising after April 2012.  Counties in purple saw even numbers of advertisements from both sides, either because neither side was on the air (as in uncompetitive states like Texas and Arkansas) or because the advertising by the two sides was quite even (as in parts of southwestern Virginia and western Ohio).

The largest Democratic advertising advantage was in the Denver media market, where the Obama campaign and its allies put more than 32,500 ads on the air.  The comparable figure for the Romney side was just over 23,000 advertisements, giving Obama a sizable advantage of over 9,000 total advertisements.  (Keep in mind, these are raw numbers of advertisements over the entire campaign, not Gross Rating Points.)

Romney's biggest advantage was in the Charlotte, N.C. media market, where his campaign and allies had an advantage of over 6,000 advertisements.  Regression analyses reinforce what a quick glance at the map suggests: Within swing states, the candidates and their allies advertised more heavily in media markets where they had higher levels of baseline support. Note Obama supporters' heavy advertising in the Cleveland media market, for instance, and Romney backers' advertising in Jacksonville, Fla.  Romney and allies did have the advantage in pseudo-swing states that got attention in the campaign's final days, including Michigan and Minnesota.

But even at those high levels, does advertising influence presidential voting?  In an article in American Politics Research, political scientists Michael Franz and Travis Ridout used county-level data to estimate the persuasive effects of TV advertising in 2004 and 2008.  They found that TV advertising was more influential in 2008 than in 2004. In 2004, with an incumbent on the ballot, having an advantage of 1,000 ads during the campaign increased a candidate's county-level vote share in non-swing states by 0.19 percentage points.  In 2008, the comparable figure is more than three times as large, at 0.60 percentage points.

As earlier research by Gregory Huber and Kevin Arceneaux points out, there is a good reason to focus on voters who share TV markets with swing states but live outside them. These voters are not experiencing other parts of the presidential campaign, meaning that we can recover a cleaner estimate of the effects of TV advertising.   (Examining the 2000 presidential election, they, too, conclude that television advertising can persuade voters.)  So construing swing states broadly to include Michigan, Pennsylvania and Minnesota, I then approximately replicated Franz and Ridout's analyses for non-swing states using 2012 county-level election data.  According to my analyses, the estimated effect of a 1,000 advertisement Democratic advantage is 0.14 percentage points, with a standard error of 0.05 percentage points.

That estimated effect is quite close to Franz and Ridout's estimate for 2004, another case in which an incumbent sought reelection.  It also fits with Lee Drutman's analyses of spending in 2012 U.S. House elections.  So voters in 2012 appear a lot less persuaded by TV advertising than we were in 2008.  While the candidates and their allies were spending more than $400 million dollars apiece on national television advertising, they were shifting only a fraction of the voters that comparable expenditures in 2008 would have moved.

Consider a concrete example.  Even had it been the Romney campaign who had a 9,500 advertisement advantage in Colorado, this analysis implies that Romney would only have netted roughly 2.7 percentage points, which is only half of the margin by which he lost that pivotal state.  And the Obama camp's advantage in Colorado was the largest enjoyed by either side during this cycle.  As in 2004, voters in 2012 were quite hard and quite costly to move.  So while the pre-election polling made 2012 seem close, post-election analyses underscore how unlikely it was that a change in advertising strategy alone could have made Romney into the one preparing an inaugural address.