IN SEPTEMBER of 1984, an ABC/Washington Post poll asked registered voters whether they preferred a Democrat or a Republican to represent their congressional districts. By a 15-point margin, respondents favored Democrats. On Election Day 1984, Democrats lost 14 seats in the House. In 1996, a similar question produced a 14-point edge for Democrats; in that election they gained 9 House seats. The lesson is that polls are important tools for understanding politics. Except when they’re not.

Washington is buzzing with 2006 poll numbers, many of which are self-contradictory. For example, according to a Time magazine poll in March, 49 percent of respondents disapprove of the job Congress is doing, but 63 percent approve of their own representative. When asked which party they would like to see control Congress, respondents gave Democrats an 11-point edge, but when they were asked about the job the parties are doing in Congress, Democrats and Republicans had nearly identical ratings: 39 percent approval and about 50 percent disapproval.