But do they work? Kinda.
An article written about some research conducted on the subject reports:
I can't find the original research and I should point out that it's not been peer reviewed, so there's no way to judge the methodology, but on the surface it seems sound -- and interesting. To continue, quoting the authors:
Their preliminary findings: Voter-choice was not altered by exposure to the email, either firsthand or secondhand. Yet, still Obama's "feeling thermometer scores'' dropped in the condition where subjects read the email, and the temperature for him rose in conditions where they read a rebuttal from Obama's Web-site.
"Our results suggest that these types of personal candidate attacks can have an affect on citizen preferences,'' Professors Phillip Hardy and Mary Walsh write in a draft of their report graciously shared with the Tribune. "The preliminary findings from our study point to (1) decreased feeling thermometer scores toward the target candidate; (2) a diminished sense of empathy and morality associated with him; and (3) fear of Obama's religious background.
The bad news is they do seem to work, at least at a vague, hard-to-define, attitudinal/affective level. Not only is that enough to get you publication, if statistical analysis holds up, but it's damned important too.
It's scary to imagine these things actually working, but apparently they do inside the head, under the hood. Whether they translate into behavior, like voting, that's always tougher to find in any kind of research. I suppose next Tuesday will be the ultimate test.