Thursday, February 14, 2013

If campaign effects are so minimal, why even try?

Today marks the beginning of my gig as a regular blogger for Pacific Standard. I'll be doing a weekly column there on, you know, politics 'n stuff. (Thanks to Marc Herman for helping to make this happen.) Anyway, my first piece is a discussion about campaign effects, related to this blog post I wrote a few weeks ago. I try to explain why the Obama campaign, which was purportedly steeped in political science research and obsessed with empirical testing of its activities, spent so much time and money on activities (such as a massive ad campaign during the summer of 2012) that likely had zero effect on voters.

One explanation that I really didn't get into in the piece (although a friend wrote to discuss with me) is that a candidate who understands all this minimal effects research is nonetheless trapped in a collective action dilemma. The candidate has obligations to her supporters and to the other candidates on her ticket. If all those people (and reporters, party activists, donors, and other observers) believe that campaign activity matters a lot, then the candidate sends a terrible signal by doing nothing. What may be, on an empirical level, a very smart decision to not waste money on a particular ad campaign is nonetheless seen by others as, at best, a dereliction of duty and, at worst, an outright betrayal. So until these empirical findings are accepted by a very broad swathe of political observers (don't hold your breath), it will still be in candidates' political interests to keep paying for things with little or no payoff.


  1. Campaigns don't matter because both sides campaign, with roughly equal effectiveness. Campaign effects are more visible when there's more of a visible difference in the level of campaign activity (eg funding disparities, candidate quality, etc.)

  2. "Also, pulling in a tug of war has no effect. But even if a team accepts this finding, no team wants to not pull because that's what their backers expect of them."

    That's roughly what "campaigns have no effect on the outcome" amounts to.

  3. BTW, the implication of all of this is that campaigns and political consulting is a giant grift. If it doesn't insult the intelligence of campaign operatives and consultants, it insults their integrity.

  4. Dana, this post certainly wasn't intended as an insult or an allegation that campaign consultants are grifters. While I think many consultants overstate the impact of campaign activities, I don't doubt that they believe in their value and perceive their work to be actually helping those candidates attain office rather than just fleecing them. I've known and worked with (and for) a number of consultants, none of whom I would describe as grifters.

    But I think the minimal effects finding is more than just the tug-of-war analogy you describe. Obama massively outspent Romney on ads last summer, and it's hard to show any direct impact of that on vote shares. It wasn't that equal campaigns were canceling each other out; there was a substantial asymmetry, with little to show for it. Bush didn't have much to show for his one-sided campaign against Gore in California in 2000, either. The point is that even when campaigns aren't evenly matched, we still see very modest impact, if any, from many campaign activities, and what we see is often quite ephemeral.

  5. I think a big problem here is making statements about "campaign effects." I think you're looking at the presidential election, where there's little problem of familiarity with the candidates and quite a bit of information known about them by the electorate, and talking about "campaign effects" as if campaigns for any other office have little effect on the final outcome. Even if one stipulates campaigns have little impact on a presidential race, I think it's obviously false for statewide and downballot races.

    Furthermore, it doesn't account for a campaigns' ease or difficulty in achieving objectives. It's pretty easy for Democrats to get to 40% of the white vote in Minnesota. It's impossible in Mississippi. To get from 15% to 20% in Mississippi may be possible, but would probably cost more and require a more effective campaign than getting to 55% of the white vote in Minnesota. And to do that in a year like 2008 was, for Dems, a heck of a lot easier than it was in 2010.

    As I said on Twitter, there's an easy antidote to the notion that campaigns have little or no effect on results: get access to the canvassing, phone, online, mail and TV plans of a campaign, match it up with the opponent's, and see whether there was movement in one direction or the other if one campaign was engaging voters earlier or far more vigorously than the other. Having seen that dynamic repeatedly, I just can't accept any blanket or loosely applied claim that campaigns barely matter, if at all.

    [In fact, here's another set of examples to look at: Senate races in 2000 and 2006. In both cases the DCCC outspent by the NRSC by a solid margin over the last 10 days or so in multiple races where Democrats won upsets by narrow margins. You could probably figure out the same thing RE IE expenditures in a lot of Congressional races in 2010, where a well-delivered blow of fairly modest size was probably a necessary factor in the defeat of up to a couple dozen of the Democratic incumbents.]

    Finally, I don't believe any insult was intended or that you were even aware of it. But I think it's there, which is why I said implied.

  6. I think your collective action explanation is likely best. This reminds me of the NFL 4th down decisions issue. Perhaps they are analogous? Paul Romer's paper on optimal 4th down decision (iirc) demonstrated pretty definitively that NFL head coaches weren't making optimal decisions. But why?

    The going theory was more or less confirmation bias. Since those responsible for hiring and firing of coaches (GMs/owners/fans) preconceived notions for correct behavior were of a particular school, only guys with unimpeachable credentials had the incentive to make the right choices. Bill Belicheck clearly had those credentials and as his 4th down decisions bore out over time, the common wisdom changed somewhat and the NFL has become more efficient.

    Similarly, the moves respectively to read option, uptempo and spread principles were most implemented by already successful coaches rather than unknowns or those with seemingly better incentives. Those without such credentials affect their future potential jobs as blame is more likely to land on someone who obviously violates common wisdom.

    This also explains why innovation is so slow in these arenas. For those attached to the mainstream party choosing apparatus, you're rewarded for maximizing effort along CW lines. Outside of that apparatus, you'll only see innovation among those who think they can actually win. That happens pretty rarely, right? Howard Dean probably counts. Ross Perot?

    In any case, I don't think it's obvious that the party insiders know something the academics don't. I think the incentives may well favor a very slow trickle of that knowledge into the CW.

  7. I'd say your both right and wrong. Within the Obama campaign the basic idea was that the ad spending was an inoculation against similar spending and a way of codifying certain narratives that were necessary but not sufficient to move the electorate. For example, the least amount of movement in the polls after the first debate was in the swing states. Now most of the movement in those polls was dissatisfied republicans admitting to their preference but the larger issue is that ad's engage the electorate such that other campaign actions can have an effect.

    So for example, many of the adds on tv were designed to sell the narrative that Romney did not care about job loss and made money on it with the converse being that Obama did care. That allowed the field staff to make the argument that things though not great were better and Obama needed more time to get the economy working because there was already a pre narrative that Romney didn't care about that issue.

    Now what I'm saying is that you can't disentangle the effects which is probably anathema to you. During the campaign a lot was done that in individual studies had shown to have effects in boosting turnout. But measuring those effects in each election are going to be enormously difficult based on whatever inhibitory factors go into voting in an individual election while an individual study showing that if you get people to plan ahead when they are going to vote may give you a 4 to 7 percent increase in voting as opposed to a placebo group of similar individuals you won't be able to test it when the campaign is going on unless you have access to individual voting records and the contact made by each campaign. Of course then you also have the issue that most contact is made by volunteers who may or may not stick to the script and who knows what effect that had.

    So in conclusion, I guess my point is that campaigns have an effect, but we don't know exactly what that effect is and if individual pieces that when measured separably do or don't have an effect in concert with other campaign actions.