1. "Likely Voters": Rasmussen narrows its field of respondents to "likely voters," whereas most other firms use registered voters or all adults. "Likely voters" don't usually include young, minority, or marginal voters more likely to support Obama.
2. The four-category question: instead of "approve or disapprove," Rasmussen asks respondents whether they strongly approve, somewhat approve, somewhat disapprove, or strongly disapprove. This yields fewer "I don't know" answers, which, typically, leads to higher disapproval numbers
3. Automated polling: Rasmussen doesn't use live interviewers. People could feel more comfortable saying they don't like Obama to a machine, or hardcore partisans could be more likely not to hang up the phone. Regardless of theory, firms that use automated polling tend to give Obama worse numbers.
This article available online at:
http://www.theatlantic.com/politics/archive/2009/12/the-rasmussen-approval-curve/31096/
