Some Poll Analysis

Watching the polls over at Real Clear Politics. From May and June, Bush was up by one or two points (and that lead seemed to be growing). The first three weeks of July, Kerry had a two or three point lead, but by the last week of July it was less than a full point. This was the week of the Convention. Begining in August, Bush has his one or two point lead back. In the most recent Survey USA poll, Bush was up one (with a +/- of 4%). So the question is, was July just part of the random variation within this +/- or was July a bump? Its hard to say, because other state polls don't show a similar pattern. Survey USA shows Matt Blunt taking the state house by five points, which clears the +/- by a point. Bond is a run-away winner getting 55% of the likely voters with 2% undecided, and 5-6% identifying "other." I think that reverse coat-tails are the new effect. So that I predict Bond and Blunt will help Bush.

One of the most interesting background results was this, "88% of Bush voters pick Bond." Given Bond's huge lead, who are the Bush-Yes, Bond-No people? Are they 9-11 Democrats? Looking at this another way, 82% of Bush supporters said they were voting for Bush, while 14% said they were voting against Kerry. Is this 9% Bush-Yes, Bond-No the same as the 14% who was voting against Kerry? If so, its probabaly a good sense of the 9-11 democrat. Likewise, 14% of Bush voters are voting for McCaskill for governor.

How many Bush voters could return home by November? Bush has more solid report. Of those voters who are certain who they will vote for (and the whole poll is of likely voters), Bush gets 50% and Kerry gets 46%. Of the merely probable, Kerry gets 55% and Bush 36%. Part of this can be explained by two phenomena. First, there are conservatives who are mad at Bush and are responding to pollsters for Kerry but will come back to Bush when the voting counts. They are sending a message. Second, conservatives tend to under-report in polling, especially in urban areas. Some of this Kerry softness can be explained by these. Another way to look at this is the voting-against column. Only 44% of Kerry respondants are for Kerry. 53% are just anti-Bush. One of the key third catagories in voting is the stay at home voter. Voters who are dissatisfied with both candidates are more likely to stay at home, even if they did appear to be likely voters.

Where is Bush strong demographically? Voters who know who they are voting for, males, voters 35-65, white and hispanic voters.

Where is Kerry strong? Probable voters, females, young voters, black and asian voters.

I've already mentioned political affiliation and the cross-tabs with other candidates, but I haven't mentioned independents. 53% support Kerry, and 39% support Bush. First, we should note that 89% of the survey responants, 573 out of 643 are certain who they will vote for. The other 70 are soft on their selection. 193 responants were Independents. 75 independents said they would vote for Bush. 102 independents said they would vote for Kerry. 1o said they would vote for someone else. 5 were undecided. There is 1 independent unaccounted for. By comparison, Blunt got 74 independents to 96 for McCaskill. Bond gets 88 independents to 86 for Farmer. Bond is a very popular Republican. So assume he has attracted every independent who would consider a Republican. That's 46% of independents. Keep in mind that 5% of so-called independents are actually partisans for 3rd parties. So an index of Bush's progress or failure might be this figure. he's at +1 with 39% of independents, with a maximum goal of 46%. If we attempt to test this by ideological self-identification, we get similar numbers. Conservatives support both Bond and Bush in similar numbers. Out of 259 conservatives, 204 support Bond, 207 support Bush. Out of 281 moderates, 115 support Bond, 84 support Bush. Bond has 41% of moderates, Bush has 30% of moderates. The gap here is slightly wider than independents, so it might well be easier to see progress. You'll notice the large number of conservatives who don't support the Republican. Missouri democrats tend to be conservative democrats. Check out Ike Skelton, Missouri Democrat.

Let's continue in the vein by looking at where a candidate's support comes from. Out of the 311 Bush supporters, 301 of them identified themselves as conservative, moderate, or liberal. So out of 311, 66.5% are conservative, 27.0% are moderate, and 3.2% are liberal. Out of 302 Kerry supporters, 288 identified themselves C, I, or M. So out of 302, 14.6% are conservatives, 61.3% are moderates, and 19.5% are liberals. So Kerry needs to capture Bush's moderates without giving up the same number of conservatives. Bush's ideal situation would be to go after moderate votes in such a way that it reveals how far left Kerry is. Since only one in eight Missouri voters is liberal, this is not the place to win elections.

Here's an interesting set. 54% of 100 likely voters with graduate hours are for Bush. 56% of 125 college graduates are for Bush. 55% of 178 likely voters with some college are for Bush. 41% of likely voters with no college support Bush. The previous month's Survey USA found a more more even match based on education, except for college graduates. The numbers for Bush in July were 44% for voters with grad hours (to Kerry's 43%), 57% of college graduates, 46% of some college (to Kerry's 47%), and 46% of no college (to Kerry's 49%). These number have a pretty wide margin of error. They figured for a sample of 100 it was +/- 10%. Putting them together, gives you a better sense in so much as it lowers the margin of error.