Zombie Politics: The Terrible Power of Negative Advertising

by John Sides on February 21, 2012 · 23 comments

in Campaigns and elections

Continuing the series, here’s another zombie idea: negative advertising “works.”  Either by hurting the candidate who is attacked or by turning off voters from the campaign altogether.  The sheer volume of negative ads certainly keeps this zombie living.

As does the terrific string of cliches attached to so much writing about negative ads.  Consider this piece.  We have effluvia: “tsunami of slime,” “toxic.”  Boxing: “win the fight with a knockout punch.”  Gore: “expensive and brutal evisceration,” “bloody victory.” And, of course, war, war, war: “ammo,” “counter-offensive,” “sharpening their arrows,” etc.  (I am not even one-fifth of the way through the piece.)

As does many examples of campaign reporting that discuss tactics—like negative ads or the micro-targeting of ads—with only vague statements about their effects.  Often from people whose very profession involves convincing candidates to pay them to make ads!

And so you get this from a forthcoming New America Foundation event:

Mudslinging isn’t pretty. But research — and conventional wisdom — says negative political ad campaigns work. Indeed, the tone early on in the 2012 contest suggests that accentuating the positive will not be the hallmark of this election cycle. Should that be of concern to us all? Are negative ads corrosive to our political discourse, or are they, in fact, a vital means of informing the electorate? Join us to consider how political messaging has evolved to its current state, as well as its impact on our broader culture. (And yes, we’ll be airing many past and current commercials.)

No, that’s not what “research” says.  I wrote two pieces on what the research about candidate advertising says—here and here—and I’ll quote from their discussion of negative advertising.

It is virtually a truism that negative advertisements make the candidate being attacked look bad, and the candidate doing the attacking look good. In 2008, Mark Penn hewed close to this conventional wisdom, asserting that despite research showing that voters dislike negativity, “clever” negative ads work. He wrote, “When reality and research differ, it is the research that is wrong.” Unfortunately, he doesn’t really know the reality. The most comprehensive meta-analysis of research into negative advertising found no conclusive evidence that they work:

“All told, the research literature does not bear out the idea that negative campaigning is an effective means of winning votes, even though it tends to be more memorable and stimulate knowledge about the campaign.”

Take the “Daisy ad.” Perhaps the most infamous negative presidential ad of all time didn’t appear to move either Lyndon B. Johnson’s or Barry Goldwater’s poll numbers. And don’t be fooled by accounts suggesting that a negative ad had some subtle effect on a race — “changed the narrative” or another similarly squishy phrase. Votes, not narratives, are what wins elections.

There has been further research since that meta-analysis was published—see, for example, here or here—that finds some effect of negative ads on either turnout or voter choice under particular conditions.  But I’m sure there is other further research floating around that hasn’t found such an effect (to say nothing of what is languishing in file drawers).  Which is to say: we haven’t remotely arrived at a place where “research” suggests that negative ads “work.”  I’m not asking that reporters or commentators or foundations say that it “never works.”  I don’t think the literature suggests that either.  I’m just asking for some engagement with this research.  I think good reporting demands it.

Die, zombie, die!


anon February 21, 2012 at 8:42 pm

I guess the intuition from those who believe in its efficacy is along the lines of: “Campaigns spend millions of dollars on negative campaigning, and have been doing so for a long time. Even in the absence of strong academic findings showing it works, it’s unlikely that 100s of years of collective experience in the hands of campaign managers is wrong on this.”

A question. Is there any experimental work that looks at framing effects & negative advertising? We know it really matters in other contexts. Is there any reason we should not believe the same here? Is there something particular about elections, in other words, that mean how a voter feels about someone does *not* effect how they are likely to vote?

John Sides February 22, 2012 at 10:33 am

Anon: Some of the work cited in the meta-analysis linked above is experimental in nature. And I don’t think anyone is claiming that how voters feel about candidates doesn’t affect how they vote. All I’m claiming is that negative ads don’t necessarily have large effects on how voters feel and thus how they vote. More specifically, I am claiming that media coverage imputes more power to negative ads than the evidence will support.

Josh Schorr February 21, 2012 at 9:20 pm

I studied political science in undergrad and grad school, regularly check the Monkey Cage, but still cannot dissuade myself of these “zombie” beliefs. What to do?

Example: Romney spends millions in Iowa and Florida. Gingrich loses. Media incessantly report effectiveness of Romney attacks. Just gets stuck in my head.

How does your 538 article square with Tali Mendelberg’s The Race Card? Her analysis shows the pronounced effect of racially-biased political advertising, specifically Willie Horton amid the ’88 election. Thoughts?

Adam Thompson February 22, 2012 at 2:04 am

Might I suggest that while negative advertising generally doesn’t work, perhaps racially motivated attack ads work more effectively. Of course, my rationale for this draws upon critical race theory, which would have it that there are institutional forms of racism that effect discourse, among other things. Whether or not proper political scientists (not fake ones like myself) would use critical race theory is beyond me.

In any case, thank you for the book recommendation, I’m sure it will come in handy for my multiculturalism and public policy class (which is working very hard to ignore issues of systematic racial bias in multicultural discourse); or at least the presentation I’m doing.

John Sides February 22, 2012 at 10:39 am

A few thoughts:

1) There is a debate over the theory that Mendelberg advances in that book. See, for example, this piece by Huber and Lapinski: http://huber.research.yale.edu/writings.html#12. (There is a further dialogue with Mendelberg in that same issue.)

2) But assume that Mendelberg is correct. What she shows in the book is that the Horton ad primed racial resentment for a brief period, until it was criticized as racist. But making racial resentment a more powerful predictor of the vote doesn’t actually mean that there was a net gain for Bush. He could easily have gained votes among those higher in resentment but lost those among those lower in resentment. What analysis I’ve seen of the tracking polls suggests that there was some movement toward Bush after the ad aired but it’s difficult to peg that to the ad as opposed to all of the other campaign events.

I am planning on writing a longest post on the Horton ad, in any case.

Adam Thompson February 22, 2012 at 5:40 pm

Thanks for the critical response piece to Mendelberg.

Looking forward to your piece on the Willie Horton ad.

Jack February 21, 2012 at 10:52 pm

For my money, Kahn (now Fridkin) and Kenney’s 1999 APSR article clarified the debate in a very useful way. They find that “mudslinging” — or personal attacks — demobilize, but policy appeals can actually mobilize voters. While these effects are really only pronounced among the ideological moderates and less politically aware, the distinction between these types of negativity fundamentally altered my own understanding of the effects of negativity. I do think that it can “work,” but the conditions are much narrower than conventional wisdom seems to indicate.

John Sides February 22, 2012 at 10:32 am

Jack: I would say that other studies haven’t entirely confirmed the Kahn and Kenney mudslinging finding. See, for example, Brooks and Geer:


In short, I don’t think that even incivility or mudslinging necessarily has detrimental consequences.

Jack February 22, 2012 at 3:53 pm

My only response would be that I don’t necessarily buy the argument that there are different effects for each type, but that it’s important to note the difference if you want to join the normative debate into which this topic tends to devolve. Geer’s own work (In Defense of Negativity) looks at this from a normative perspective and argues that negativity is, in fact, good for democracy, as it provides more information. I think even Geer would agree that mudslinging isn’t very informative. The Kahn and Kenney paper was definitely groundbreaking, if only for explicitly examining the distinction between policy negativity and (personal) mudslinging.

Rob February 22, 2012 at 7:31 am

I’m uncomfortable with this “zombie” metaphor. Has political science really taught us something so incontrovertible about negative advertising (or class and voting, or whatever) that we can dismiss conventional wisdoms as zombies? As you note, “we haven’t remotely arrived at a place where ‘research’ suggests that negative ads ‘work,’” nor have we arrived at a place where research suggests they “never work.” Looking over your 538 posts on this topic, as well as the meta-analysis, it’s not clear to me that we’ve “arrived” anywhere at all, really. Which is fine–that’s the nature of social science. But why then should we feel comfortable dismissing departures from that research as “zombies?” I think the points in your last paragraph are well-taken: journalists who cite “research” should engage with that research in a meaningful way, and should be aware of whatever tentative consensuses it may have reached. But “die, zombie, die?”

John Sides February 22, 2012 at 10:23 am

Rob: The zombie in this case is the idea that “negative ads are very powerful.” That’s the notion that the research does not support, and that notion deserves to die. What should replace it is something like “the evidence on negative ads suggests that it has no consistent powerful effect on vote choice or turnout.”

RBMJR February 22, 2012 at 10:21 am

Perhaps the effect of negative ads are muted in traditional campaigns where both (or all) sides are slinging the mud relatively equally but in cases like Iowa, South Carolina, and Florida where the negative information flow is one-sided they are far more effective? As Zaller has argued, when there is no cross-current flow of information media effects may be much more significant than we tend to think.

John Sides February 22, 2012 at 10:30 am

RBMJR: Absolutely it matters whether there’s an imbalance in ads. See my first Moneyball post linked above. The problem is that most media accounts of negative ads elide these nuances.

And suppose that the one-sided flow is entirely *positive* ads. Would those be less effective than a one-sided flow of negative ads? I’m not sure they would. But I think media accounts assume otherwise.

Scott Monje February 22, 2012 at 11:40 am

As Jack points out above, the term “negative” can cover a multitude of sins. Has anyone examined the relative truthfulness or specificity of ads? I can imagine that it would be difficult to discuss, say, Newt Gingrich and his career in a neutral way without out it sounding like a negative ad (but then, maybe that’s just me). Are negative ads more effective to the extant that they can be tied to reality, or to specific events with undeniable consequences, as opposed to vague slurs?

Dana Houle February 22, 2012 at 12:28 pm

I don’t think you can say the Daisy ad didn’t work, because in practical terms, as a form of advertising, it was never tried. It ran exactly once. It was a PR move, it was NOT a tool of advertising, where repetition and targeting matter.

That’s what I see as a big problem with this discussion. You can’t only look at content of an ad without looking at the volume of repetitions behind it, who it was targeted at, and the context of the race. There are plenty of races, probably the majority, where the only ads are positive. They’re in the safe districts, where incumbents raise a good chunk of cash and spend it in a safe way: making them look good. They can do this and this alone because their opponents have no campaign, or at best are woefully overmatched in terms of resources and the underlying partisanship of the district or state. To know if negative advertising “works,” you have to throw out all those races, and look pretty much only at highly competitive races, because those are the only races where you’re likely to get negatives that are seen by many people (either through electronic media or via direct mail).

John Sides February 22, 2012 at 12:59 pm

Dana: Running the Daisy ad only once probably did limit its impact. But I’ve seen plenty of people impute significance to it nonetheless. Hence the need to push back.

I very much agree with everything else you say: we can’t understand the effect of ads without taking into account context. My first Moneyball post was an attempt to identify some key aspects of context.

However, I’m not sure that competitive races are the ones we’d expect to see negative advertising work. In those races, you’ll see so much advertising from both sides (negative and positive) that it’s going to be hard for either candidate to get an advantage.

Dana Houle February 22, 2012 at 1:19 pm

But isn’t that the problem, that negative advertising is only done in competitive races, or where a candidate is doing a hail mary pass in an non-competitive race to try to tighten it up?

And I may agree with the folks who say the Daisy ad was significant; I’m just shifting it’s significance from advertising to something else. Swift Boats was the same way; it eventually got some money behind it, but it was primarily something that was put on TV with a very small buy but was then pushed around to journalists, pundits, talking heads and the like. It influenced media coverage of the race, but people’s awareness of it and the way it changed anyone’s perceptions was not because they experienced it as a paid advertisement.

John Sides February 22, 2012 at 2:27 pm

Dana: That’s precisely the problem. It’s what makes it difficult to isolate the effects of ads generally or negative ads in particular or news coverage about ads. Which is why I’d suggest that commentary and reporting about political ads be more circumspect.

Dana Houle February 22, 2012 at 3:19 pm

I don’t disagree with that comment, but that’s a bit different from the post. Saying “be circumspect” is different than saying the belief that negative ads “work” is a zombie that needs a stake driven through its heart.

For the record, I think negative ads are a tool, an often powerful tool, that when used properly and in the right context, can be very effective. I also believe that few tools are better than a hammer at pounding a nail in to wood, but that if you have to screw in a screw, the hammer won’t work very well. But because its not helpful with screws doesn’t mean the hammer doesn’t work, but it also doesn’t mean a hammer will drive a nail in to a piece of cast iron. I think the discussion and research should be not whether they work, but trying to isolate the conditions where they do and don’t work.

John Sides February 22, 2012 at 5:00 pm

Dana: See my comment to Rob above. The meme that negative ads “work” — stated, as it so frequently is, without caveat or nuance — is what needs to be killed. Circumspection would be a welcome change.

Your view is perfectly plausible. I think campaigns ads, and negative ads, can sway voters — and sometimes enough voters to be consequential for the outcome. There’s polisci research to back that up, cited in my posts at 538. The problem is that this research rarely gets mentioned in media reports. If it did, there might be less hyperbole about the effects of ads.

Dana Houle February 22, 2012 at 12:45 pm

Also, ad campaigns are different between Pres/Sen/Gov and Congressional and lower. In the latter category, campaigns rarely run more than 3 distinct spots; they have to get all their content delivered in 84 seconds of programming. But the other races, they tend to plan out their media campaigns in much more detail, with some ads as inoculation on the candidate’s potential weak spot (like an ad on a campaign I did where we emphasized the candidate’s blue collar roots, young family and boy scout-ish persona to blunt potential attacks about him being a very rich banker from a community that most of the rest of the district saw as full of obnoxious plutocrats, and to at least have voters think better of him when they learned he was from that disliked community), or to muddle an issue (like in 2000 when Spencer Abraham ran ads about prescription drugs that were mostly bullshit but which so confused the issue that what going in to the race was probably Debbie Stabenow’s best contrast was defused by essentially confusing the electorate). Such complexities, in my opinion, make it harder to isolate what is and isn’t a negative ad, and the effects of the opponent in parrying or even preventing an attack. It’s like assessing the value of battering rams, but not accounting for some doors being made of thin wood, some being made of thick iron, and not accounting for whether the person on the other side of the door was able to pile things against the door to make it harder to budge. It’s the same battering ram, but its effectiveness depends much on the what it’s being used to batter.

Filisfog June 25, 2012 at 12:35 pm

am interested in political advertising research.
Good article

Filisfog June 25, 2012 at 12:36 pm

very good article

Comments on this entry are closed.

Previous post:

Next post: