Polarization in Less than 30 Seconds

We continuously monitored voters’ reactions over the course of a thirty second ad. As they watched a pair of target ads from a 2006 Senate race, study participants moved a slider to indicate their general approval or disapproval of what they saw or heard during the playing of each ad. The pattern of change in the position of the slider demonstrates the reinforcing effects of advertising quite clearly; Democrats and Republicans moved toward the opposite extremes as they watched the ads.

That’s from a new paper by Shanto Iyengar, Simon Jackman, and Kyu Hahn. That political campaigns tend to reinforce partisanship is no surprise, but it’s interesting to see how quickly this happens within the context of a single advertisement. I’ve pasted one set of their graphs below:


There are also some interesting nuances. For example: positive ads more quickly polarized partisans than did negative ads.

In general, Democrats responded more quickly than Republicans to their respective ads suggesting a contextual advantage for the former. Polarization was also accelerated in the case of positive ads; partisans’ dial scores generally took longer to converge in the case of negative ads. Finally, as anticipated, we found that the rate of polarization was significantly higher for the more strongly partisan of voters.

Frankly, I’ve always thought of dial groups as a cable news gimmick. (And thus hated myself for helplessly fixating on the dial group’s lines during the CNN broadcasts of the presidential debates.) It’s nice to see this technology used for something more scholarly. And, as Iyengar and colleagues note, there are yet more aspects of political advertising to explore.

2 Responses to Polarization in Less than 30 Seconds

  1. Andrew Gelman February 26, 2009 at 1:22 pm #

    Very nice. My only suggestions are to start the x-axis at 0, add a bounding line at y=100, and just put x-labels every 10 seconds (easier to read than every 5 seconds, I think).

    And the lines should be thinner.

    And “Deomcrat” in the caption” should be “Democratic.”

    Finally, the caption is ambiguous, to the extent that I don’t actually know what’s being displayed. Does each line represent a person in the experiment (a “subject,” in experimental jargon), or is each line an ad?

  2. Matt Jarvis February 26, 2009 at 5:44 pm #

    I read the graph as every line is an ad, mapping the mean response by group.