ARE UNDECIDEDS GOING TO BREAK FOR THE PRESIDENT?
Bush 52.3% – 295 EV | Kerry 46.7% – 243 EV
EV without Toss Up states (under 2% margin): Bush 284| Kerry 207| Toss 47
September 28, 2004
Bush's lead in the Composite Poll's meta-analysis of national polls holds steady this week, up a mere 0.1 point over his lead from last week. In fact, Bush's bounce from the convention of 5-6 points has remained steady and currently shows no signs of weakening, right as we prepare for Kerry's next big opportunity to move the numbers - this Thursday's debate.
Despite his steady lead in the national polls, Bush loses Pennsylvania and New Mexico and their 26 electoral votes, as those states barely slip back into Kerry's column. But there are contradictory results in this week's analysis. With 77 state polls included over the last week for 38 states (including DC (Kerry +67)), Bush has a lead of more than 2 points in states worth 284 electoral votes - up from 270 last week. Thus, Bush has solidified his lead to some extent in enough states to win re-election while Kerry has picked up a few points among states within 2 points to increase his aggregate total in this week's analysis. But don't take too much comfort in any movement in the 2 point lead toss-up range. Bush leads in the traditional battleground states by 2.8%. Bush leads in the 2000 Bush states by 12%. Kerry leads in the 2000 Gore states by 8%.
There's something interesting in the polls over the last three weeks while Bush has held his lead. While Bush has lead by double digits in some polls, FoxNews has shown the race closer since the convention with Bush leads coming in lower than the average of other polls, by 2 points on Sept 8 and by 3.4 this week. And in each poll, FoxNews also reports a greater number of undecided voters. Could it be that pollsters who push "leaners" to answer get better numbers for Bush than pollsters who are content to find undecided voters show a closer race? So, I reviewed all polls for the last three weeks and compared the number of undecided voters to the level of support for each of Bush and Kerry. As you seen in the below chart, as undecideds decrease in a poll, both Bush and Kerry's support increases - as you might expect. But Bush's support increases more rapidly. Ranging from a low of about 45 if there are 12% undecided to a projected high of 53% with no undecideds (projection is a least squares linear progression, represented by the line in the below chart). Meanwhile, Kerry's support ranges from about 42 to 45, a much smaller increase.
What does this tell us about undecideds and how they may break? Well, the Republican spin is clear - the undecideds are showing a tendency to break heavily for the President. But the spin you might expect from Democrats is probably more accurate and more relevant to campaigns trying to figure out where the race stands today. "Leaners" are both less solidly behind their candidate and, probably, less likely to vote. Thus, a greater proportion of Bush's voters in the latest polls may be soft.
While that should give some hope to Democrats, they should still be concerned that Bush still holds a lead even when poll victims aren't pressed for whom they lean toward. At the same time, Democrats can hope that this soft support can be turned. And the debates will give them that opportunity - while Bush tries to solidify the "Leaners'" support.
Click the graph to enlarge.
Click here for the details, numbers, graphs, maps and the Election Simulation, which now shows an 84% probability of a Bush electoral victory and 95% confidence that Bush's electoral vote total will be between 240 and 354. But you'll note that Kerry's making gains in his probability of winning Florida and Ohio - but he still has some way to go.
