Advertising imbalance, p. 171, The Gamble, Sides & Vavreck
One of the main building blocks in the Beltway media's "Dems in disarray" theme for this year is that Republican outside groups, especially the Koch brothers-linked
Americans for Prosperity, are out to a huge lead in running TV ads. AFP alone has outspent the main Democratic outside groups (Senate Majority PAC and House Majority PAC) in early advertising in competitive congressional races by an over 2-to-1 ratio, and that's before factoring in ad spending by the more establishment-flavored outside groups like American Crossroads or the U.S. Chamber of Commerce.
That's been going on for a while, though, and you'd think that things would steadily be getting worse for the Democratic Senate races, in the same way that the constant drip-drip of water steadily erodes anything. The odd part is, the Democrats' poll standings aren't getting worse, and as Markos pointed out several times in the last month, in selected cases they're actually getting better. The Kochs and friends aren't getting the bang for the buck on their spending that they'd expected; they're spending and spending and not moving the needle.
There are lots of possible explanations for that; for one, presidential approval numbers have ticked up a bit in recent weeks, probably thanks to better economic news and strong ACA enrollment numbers. That might rub off on overall Democratic fortunes, but not in any profound way; Obama approvals have only increased from the low-40s to the mid-40s. And another possibility is that some of the AFP's more notorious ads have simply fallen apart under media scrutiny, most notoriously the anti-Obamacare ad starring Julie Boonstra in Michigan, which might blunt their impact somewhat.
But there's a better explanation, one that's apparent to people who've quantitatively studied the effect of TV ads: They just don't work, or more accurately, they work for only a short period of time. Their effect on overall poll numbers is small and, more importantly, the effect starts wearing off almost immediately. AFP and their allies can juice the poll numbers a bit by keeping ads on constant rotation, but as soon as they take their foot off the accelerator, the races quickly revert back to the mean. Over the fold, we'll look at the arguments and the numbers behind these claims, drawing primarily on one of the best retrospectives of the 2012 election, The Gamble by John Sides and Lynn Vavreck.
If you haven't read The Gamble, it's well worth your time to do so; most significantly, it's an antidote to the usual breathless but anecdotal post-campaign write-ups like Game Change. In fact, it's almost a direct rebuttal to books like Game Change, because, as Sides and Vavreck methodically demonstrate throughout, there really aren't any game changing moments in a presidential campaign; the alleged game changers don't do much beyond causing short, small fluctuations within a narrow band of polling. The trajectory of that narrow band instead depends on a handful of data like economic conditions, presidential approval, and incumbency. Debates don't matter, gaffes sure as hell don't matter and ads matter only to the extent that neither side can surrender the airwaves—and if both sides advertise thoroughly and competently, it's largely a wash.
Sides' and Vavreck's research drilled down to look at polling in individual media markets to see how polls in markets where one candidate was advertising more compared with polls in markets where both candidates were advertising in parity. They found that a candidate with a 100-GRP advantage in a market (about one more ad per capita more than the opponent aired) could expect to gain almost an additional point of vote share, compared with a market where the ads were evenly matched.
So that's clear evidence ads work, right? Well, you have to think in terms of the point of diminishing returns. One percentage point is so small an advantage that it hardly matters, except in a super-close election. In addition, there were two further problems: one, there were few markets where, at any moment, one candidate had gone dark and the other was advertising hard; this happened only around 20 percent of the time in any given media market. The second problem was that the ad effects rapidly disappeared.
We found no statistically meaningful impact of ads aired the five days before the respondents were interviewed. Only the ads aired on the day closest to the interview mattered. Most of the effects of the ads were gone within a day, consistent with what other studies have found. This fact decay meant that any boost from an advertising advantage was a very temporary one. And because neither candidate had a consistent advertising advantage, day in and day out, it was not surprising that the polls were so stable. (p. 130)
So apply that knowledge to the situation this year, where in many markets, we currently have AFP or other GOP groups advertising, while the Democratic Senate candidates and their allies are saving their powder, most of the time. Where AFP is running its ad blitzes, if a pollster happens to show up that day, it'll register. But it'll only register to the extent of pushing the margins a point or two in the GOP direction ... and if a pollster shows up some
other time when AFP doesn't have the ad accelerator floored, then the effect doesn't show up at all. That would go a long way toward explaining the status quo remaining (or small Democratic gains) in the key Senate races. And when the battle is fully joined after Labor Day, with Democratic candidates and Democratic outside groups spending constantly on ads, that temporary, sugar-shock GOP advantage won't return.
If you want to see an illustration of those principles at work, look back up at the chart at the top of the article. It's a little hard to initially interpret, but the range of gray dots for each day is the ad spending in each media market. The trend lines are the national averages of each of those markets; the fact that the trend lines constantly hover around the "no advantage" midpoint shows how the two campaigns repeatedly fought to a draw on advertising on average, even though there could be big disparities in individual markets. They also add a few arrows pointing to periods where there was a bigger-than-usual move in the polls. Notice, though, how those arrows point to days when the ads were near parity. The poll movement (which, except for Romney's post-first debate bounce, were small anyway) doesn't correspond with big moves in advertising.
One other detail you might notice is that Romney's campaign had a small lead in ads in the closing weeks. You might think that that, indeed, gave Romney a one- or two-point boost at the finish line. It may well have done just that (maybe it kept Obama from getting, say, 52 percent instead), but Sides and Vavreck point out that thanks to early voting, nearly one-third of all the voters had already voted by that point, meaning that final blitz was wasted on them.
At any rate, that late Romney ad surge was still far too small to have the desired impact, according to Sides' and Vavreck's calculations. That's true even if he tried to leverage his firepower most effectively, targeting only the minimum number of states to get to 270 electoral votes. (And, as they point out, their calculations don't take into consideration the physical impossibility of buying enough time on the airwaves to exert that kind of advantage. In other words, there are only 24 hours in the day. That explains the Romney camp's late advertising push into Pennsylvania, which occurred largely because there simply wasn't any ad time left to purchase in places like Ohio.)
It was not common for Romney to have even a 3:1 advantage on the day before the election. The most he had anywhere was about a 9:1 advantage. Based on our model, he would have needed to expand his advertising advantage in Florida by an additional 4:1 margin, in Ohio by an additional 16:1 margin, in Virginia by an additional 22:1 margin, and in New Hampshire by an additional 31:1 margin. (p. 221)
"Wait a minute..." you might be saying. "What about LBJ's Daisy ad? What about Bush 41's Willie Horton ad? Those totally turned those elections around." Well, there's no question those ads were some of the most memorable moments of those campaigns, but there's not much polling evidence that they were true turning points, at least not evidence as granular as what we have in 2012. (Also, an economistic model points toward LBJ and Bush winning those elections anyway, regardless of what kind of ads they ran.) What, for instance, was the ad that came closest to being regarded as a "game change" in 2012? It was probably OFA's "
Firms" ad, the one released in mid-July that juxtaposed Mitt Romney singing "America the Beautiful" against statistics about Bain outsourcing. Did it move the numbers? Look below, and you'll see that it barely was a blip, along with basically everything else that happened during the summer of 2012:
Poll standing of Obama and Romney in spring and summer of 2012, p. 120, The Gamble, Sides & Vavreck
But that raises yet another question. If ads matter so little, and the cake is baked largely by factors that are outside of the campaign's control, why even bother? Surely there are better things to spend that money on. Well, it becomes a game theory problem, similar to Mutually Assured Destruction during the Cold War. You can't risk
not playing the game.
Our view is that candidates should and indeed must campaign vigorously. It is precisely when one side out-campaigns the other that the polls may shift in its direction. But the combination of two vigorous campaigns is often a tie, as the two sides offset each other. It may seem perverse for a candidate to spend more than a billion dollars only to ensure a tie, but to do so otherwise amounts to unilateral disarmament.
We have also shown that most of the effects of campaign advertising wear off quickly. This might suggest that a campaign should wait and spend most of its money in the final week. We would not advise that either. For one side to cede the airwaves to the other side until the end would create a very different campaign... In such a campaign, the candidate who did advertise might build up enough of a lead in the battleground states that a late and powerful blitz by the other candidate would be insufficient. (p. 240)
Now, granted, there are some differences between a presidential race and a Senate race that might keep us from applying lesson from
The Gamble uncritically. In a presidential race, by the time the election rolls around, both candidates have nearly 100 percent name recognition just through the sheer ubiquity of the campaign, but in a Senate race, there's a need to keep introducing the candidates to the voters, and also more of a tendency for voters to rely on whether there's a "D" or "R" after the candidates' names.
In addition, there just isn't the same level of money in Senate races that allows both candidates to run the ad engine at 4,000 RPM for half a year, so there might be larger and more prolonged ad disparities than the perpetual deadlock we see at the presidential level. And some states are small enough that retail politicking can still make a difference for below-the-radar, bolstering candidates with personal appeal and deep roots (a big advantage for Heidi Heitkamp and Jon Tester in 2012, and hopefully one that can help Mark Begich and Mark Pryor this year).
But in general, the lessons are the same: Ads move the needle, just a very little, and only for a short period of time. You can't opt out of advertising altogether, but there's not much sense in freaking out about being out-advertised half a year out from the election, since the advantage that buys you is more mirage than permanent.