Steve Rosenthal and Karyn Bruggerman of the Atlas Project decry use of negative ads, using Massachusetts as a case study:
...maintaining a positive profile proved to be as much of a boon to candidates as going negative hurt their opponents. Scott Brown’s polling numbers dropped only after he began to go negative against Warren in September, and Kaine pulled away from George Allen in the polls only after he began airing positive ads in late August.
Candidates who ran positive ad campaigns also appear to have exited their races with higher favorability ratings than their negative counterparts. Looking at final polls conducted by Public Policy Polling and Rasmussen on these same Senate races, only Brown, Warren, Kaine, Heidi Heitkamp, Jon Tester and Rick Berg could boast favorability ratings over 50 percent in the final days of the campaign. Brown was even shown with a favorability rating of 58 percent in a post-election WBUR poll. Brown, Warren, Heitkamp and Kaine were notably four of few candidates examined here who kept their message largely positive. It seems possible, but unlikely, that this is purely a coincidence.
I prefer academic research when it comes to issues of what's effective and what isn't in terms of ad effectiveness.
And the academic research says, yes, negative ads are effective and provide voters a service in that in it disseminates information to voters they may not have had before.
There are, however, a few caveats: in order to be effective and viewed as fair by the voters the information must be recent, salient to the race and the voters, and, interestingly performed in moderation:
The findings reveal that students were most likely to vote for the sponsoring candidate when the ad was viewed only three times and least likely when they were exposed to the ad five times.
In another test, 306 university students viewed advertisements for unknown candidates during a 30-minute television program, with varying time intervals between ad repetitions.
Participants then filled out questionnaires to evaluate the sponsor and the attacked candidates as well as the likelihood of voting for them.
The results showed that larger time intervals between repetitions of the ad boost the evaluation of the sponsoring candidate and disfavor the evaluation of the target candidate.
The results stayed the same even with increased repetition.
Essentially: keep it relevant to the voters.