I need to partially withdraw what I said a month or so ago about negative political advertising and turnout: apparently it's not the case that negative campaign ads depress voter turnout. At least according to The Effects of Negative Political Campaigns: A Meta-Analytic Reassessment by Richard R. Lau, Lee Sigelman, and Ivy Brown Rovner in The Journal of Politics, Vol. 69, No. 4, November 2007 (PDF at this address, via NPR).
What these authors did was to take the findings from 111 different studies completed between 1984 and 2006 and mash them up together into one big study (a "meta-analysis" because [jump]
an analysis not directly of data but of other analyses). The 111 studies presented evidence that went in all directions: that negative advertising is more memorable or less memorable; that it arouses interest in the campaign or diminishes it; that it decreases positive feelings about the target candidate or actually increases them (by "backlash"); and so on.
They found that overall, attack ads had more of a negative effect on voters' feelings about the attacker than anything else; but not enough of an effect to change the results. And they found that negative advertising decreases people's intention of voting but increases the actual turnout, but again not in a way that really matters. So I have to step back and say, OK, not proven. QNED.
However: all of these studies tend to fail in typical social science fashion to take some important things into account, for example:
Putting it more concisely, here's a hypothesis that hasn't been tested:
 |
| The original Demon Sheep, Tony Blair, in a Tory ad from 1997. Guardian. He won his election. |
What these authors did was to take the findings from 111 different studies completed between 1984 and 2006 and mash them up together into one big study (a "meta-analysis" because [jump]
an analysis not directly of data but of other analyses). The 111 studies presented evidence that went in all directions: that negative advertising is more memorable or less memorable; that it arouses interest in the campaign or diminishes it; that it decreases positive feelings about the target candidate or actually increases them (by "backlash"); and so on.
They found that overall, attack ads had more of a negative effect on voters' feelings about the attacker than anything else; but not enough of an effect to change the results. And they found that negative advertising decreases people's intention of voting but increases the actual turnout, but again not in a way that really matters. So I have to step back and say, OK, not proven. QNED.
However: all of these studies tend to fail in typical social science fashion to take some important things into account, for example:
- the pure quality of the ad: a helpless idiocy like Carly Fiorina's famous Demon Sheep ad of 2010 is not going to help its candidate very much;
- the verisimilitude of the attack: I cannot imagine why voters were ready to believe the attacks on John Kerry's war record in 2004, but they were;
- the attack's target audience: while lies about Kerry's war record were effective, truthful aspersions on George Bush's draft dodging were not.
Putting it more concisely, here's a hypothesis that hasn't been tested:
Attacks on Democrats (if at all believable) decrease Democratic turnout (and may or may not increase the Republican); attacks on Republicans do not affect Republican turnout (and may or may not increase the Democratic).That would explain why it works, anyway, and who it works for.
