The gender gap has opened since the debates began and now roughly mirrors the pattern in 2000. Currently half of female voters favor Kerry, compared with 41 percent of men. That nine-point difference is nearly double the disparity in late September. At the same time, 55 percent of men support Bush compared with 46 percent of women, again about twice the size of the pre-debates gender gap.
The survey suggests that the debates also may have slightly shifted the issue agenda in a way that could benefit Kerry. According to the poll, the economy and Iraq are slowly emerging as the top two voting concerns while concerns about terrorism may be fading slightly. More than one in four likely voters -- 27 percent -- say the economy and jobs make up the single most important voting issue determining their vote. Among these voters, Kerry has a 2 to 1 advantage over Bush.
Over the past three weeks, Iraq has gained in importance, with 26 percent naming it as their top voting issue, up six points since late September. Among these voters, Kerry holds a lead of 58 percent to 38 percent over the president.
Since late September, the proportion naming terrorism as their top voting issue has dropped from 24 percent to 19 percent. Nine in 10 of these voters support Bush.
The survey also suggests that Kerry's attacks during the debates on Bush's handling of Iraq, the war on terrorism and the economy did little to change overall perceptions of the two candidates' personal qualities.
Bush is viewed as the stronger leader (57 percent to 37 percent) and more honest and trustworthy (49 percent to 39 percent), virtually unchanged since the debates began. Bush also continues to be regarded as the candidate who has taken the clearest stand on the issues (55 percent to 36 percent), though the gap has narrowed by 14 points in the past three weeks.
A total of 2,402 randomly selected adults, including 2,130 self-described registered voters and 1,656 likely voters, were interviewed Oct. 14-17 for this tracking survey. The horse-race results are based on the four-day sample, while other results are based on interviews conducted Oct. 15-17. The margin of sampling error is plus or minus three percentage points.
Senior polling associate Christopher Muste contributed to this report.