Thursday, February 23, 2012

The Limits of Campaigns

Nate Silver made the case this past Sunday in the NYT Magazine that Barack Obama will run a populist campaign because, all else equal, it works best for him in the electoral college.

Here's the problem with that: Silver's analysis depends on a massively unrealistic estimate about campaign effects. If the president goes populist, Silver speculates:
So let’s conduct a thought experiment. Suppose that against Romney, Obama does 10 points better among white voters whose households make less than $50,000 per year. The trade-off is that he does 10 points worse among whites making $100,000 or more and 15 points worse among whites making at least $200,000.
10 points! There's no way that an incumbent president can move voters that much by any combination of 4th year rhetoric and promises.

Fortunately, we have a bit of evidence to look at. Silver says, and I agree, that the president has shifted to a more populist tone in recent months, with policies to match. Since Gallup keeps week-to-week crosstabs on its daily tracking poll, we can see what effect, if any, the shift has. The categories don't match up perfectly -- Gallup only gives us ethnicity and income (and education) variables separately. But if there's any effect, it should show up, I'd say, in the income breakdowns. Gallup gives us four categories, by monthly income: the ones we're interested in are below $2000 (or $48K  $24K a year*) and above $7500 ($90K/year).

So: Obama gave his jobs talk to Congress on September 8. In the five weeks preceding his speech, the average approval rating difference between the "working class" and "rich" groups was 7 points (I'm using five weeks because the number jumps around quite a bit, as small-sample crosstabs in polls will do). And in the most recent five weeks, that gap has...been reduced, to under 6 points. Overall approval rating in the lowest income group went from 46.4% in August to 50.4% over the last five weeks; in the highest income category, the shift is from 38.4% to 44.8%.

Granted, these are not exactly Silver's groups, but the fact that they moved together -- and that there was more movement among the highest income group -- is certainly suggestive. Of course, it's unlikely that the president's rhetoric had much to do with the shift, which (I think Silver agrees) was primarily caused by improving economic conditions, along with getting farther away from the ugly debt limit fight. But that's the whole point; presidents really can't do much by rhetoric and position-taking alone to affect approval, or, as we get closer to November, vote choice.

Now, none of this means that Silver's advice to the Obama campaign is necessarily wrong. After all, even if the realistic maximum effects are closer to 1% than to 10%, well, Al Gore would have been very happy to get another 1% in Florida in 2000. Of course, whether the electoral college math works the same way at 1% or 2% instead of 10% is another question, as is the important question of whether the tradeoff Silver speculates about would really play out evenly overall, so that the campaign would merely be shifting where support comes from. So I'd still advise a bit of caution.

The larger point, however, is just to be very careful about anyone who claims very large campaign effects in general election presidential campaigns. The truth is that if they existed, we would know about it. Campaign effects in my view do exist...but we're talking about very modest stuff, and my guess would be that's even more minimized when we're talking about an incumbent president, who is mostly going to be stuck with the reputation he's built over the previous campaign and his time in office.

*Update: apparently multiplying by 12 was too much for me; I had this wrong originally. Sorry about that. I don't think it particularly changes the point, so otherwise I'm going to leave the post alone.


  1. Silver has a 'structural' bias toward treating campaigns as Making a Difference. But perhaps he chose 10 point swings for the sake of illustration? Two points doesn't sound like very much, even though it could easily be decisive in a close race.

  2. I'm not sure we should accept a statistical argument from someone who thinks $2,000 a month is equal to $48,000 a year.

    1. Argghhh. That's really ugly, isn't it. I guess I need to go in and fix...

    2. And now fixed. I'm leaving, as I noted above, the rest of the post alone. Sorry about that, and thanks, Scott, for the catch.

  3. J.B.,

    There's not really any claim in the article about the magnitude of campaign effects. It is termed a "thought experiment" and designed for illustrative purposes.

    The meaty part of the article is simply in pointing out that white working-class voters tend to be concentrated in swing states, whereas wealthy whites very much are not. It does seem to me that there is some tenuous evidence in the polls that Romney could have some issues among working-class whites, and in the Midwest, at least as compared to how he performs against Obama nationally. (Despite the headline, the article is really more about Romney/Santorum than Obama). Even 1 or 2 points worth of difference is enough to impact the electoral math since that's roughly the margin at which the electoral math matters to begin with.

    1. Hey, Nate! Welcome to Plain Blog.

      Fair enough, and I hope I didn't overstate your claims in what was overall an article I like (in, by the way, an edition of the NYT magazine that was outstanding).

      Truth is, though, that I'm pretty skeptical that there's any kind of effect here at all. I do think there's a plausible case for a Romney/Santorum difference, although I'm not entirely convinced that it would be the case (and, in the event, I think there's a very good chance that Santorum as nominee would turn into a disaster, making this kind of analysis irrelevant). But for an incumbent president...I definitely believe that there can be GOTV-type campaign effects produced by organization, but I'm pretty skeptical that rhetoric & positioning can make any difference at all.

    2. If y'all are interested, I'm curious whether campaign effects might make a difference in niches, if not necessarily en masse. To illustrate, there's a new Santorum ad that attacks Romney using his own words. Most of it is pretty bland boilerplate stuff, Romney not disciplined on spending, sucking at the trough, compromised like all other politicians. A big yawn.

      The first two quotes seem a bit different, particularly in light of Nate's observation about Romney's struggles with working-class whites. The first is Romney saying he's not aligned with the NRA; the second is that he's pro-choice.

      The question: if we set aside the general stuff about spending and liberalism as just noise, might there be a specific impact in niches, such as the 2nd Amendment or Pro-Life voters paying extra attention to the Santorum ad? Has anyone studied the impact of issue-based campaign framing, and whether its impact flows through from the niche to the overall results, such as Santorum surely would like to happen in the Midwest?

  4. Jonathan,
    Have you read Lynn Vavreck's book "The Message Matters?"

  5. I think the voters in families with incomes above $200 K are more switchable than the rest of the electorate. According to the exit polls, they switched in significant numbers from Gore in 2000 to Bush in 2004, and then switched in large numbers to Obama in 2008 (Obama was the first Democrat to carry the highest income group polled since exit polls have been around, while Kerry lost this group by more than 3 to 2). These voters have more at stake in government decisions, and tend to be better educated and to follow politics more closely than less prosperous voters. McCain's choice of Palin likely hurt the Republican ticket with the highest income group, as did his somewhat erratic performance around the Lehman bankruptcy in September 2008.


Note: Only a member of this blog may post a comment.

Who links to my website?