Wednesday, June 14, 2006

Survey USA: DeWine Analysis

My problem with the Survey USA polls in the past is that they frequently fail to represent the Ohio electorate. I'm no statistician and math isn't exactly my forte, but even I can see a few problems with these polls. This poll does have some problems, but I think there are still a few things we can learn from it.

For instance, we are to believe that the political ideology of the state is 35% conservative, 38% moderate, and 16% liberal (which does sound about right to me) while the party breakdown looks like this: 39% Republican, 39% Democrat, and 20% independent (which is absurd). For this to be so, no more than 4% of Ohio Republicans have to be "moderate" which all you have to do is take a look at our state government to see that it is filled with "moderates" which represents a heckuva lot more than just 4% of the Ohio Republican electorate.

What can the DeWine camp take from these numbers? Well, he needed to reach out to Pierce and Smith voters and didn't. Pierce and Smith garnered roughly 28% of the Republican primary vote. DeWine has managed to convince half of them to vote for him. 13% say they are voting for "Other" (one can assume there will be Smith and Pierce votes) and 1% say they are still "undecided".

DeWine's strongest numbers comes from the age group that is least likely to vote, 18-34 and he is doing horribly with the over 50 demographic. DeWine also does very well in the Hispanic community, which is also the least represented in the voting rolls while getting trounced in the African American community (although that is only 9% of the vote). The racial factor is pretty much a wash if you ask me.

The education numbers are kind of interesting. DeWine does well in the middle 55% - those who have some college or a college degree - but is getting beat up by the outliers - those with no college and those with graduate school under their belt. These two outlier groups tend to be the most liberal, which let's remember, is only supposed to represent about 16% of the Ohio electorate.

The Bush approval rating numbers are fairly worthless, in my opinion...

The income numbers indicate that DeWine is having trouble connecting to the less wealthy. Which makes sense with the over 50 age demographic numbers, but not the 18 to 34. These numbers also fail to make sense when you consider the education spread as well. The less educated (which tend to be younger) tend not to like DeWine while the young demographic is DeWine's bread and butter.

There are some holes in my logic, I admit that; but on the whole, there are plenty of reasons to be skeptical while still finding some value in these numbers. I expect the DeWine race to be fairly close - within ten points, I'd imagine.


Joe C. said...

If something doesn't make sense, either you're not smart enough, or someone's not telling you something...from my previous post on RAB:

If you look at the crosstabs of likely voters, there is no way that 22% of conservatives would vote for Strickland, which is more than a third higher than than conservatives protesting DuhWhine (16%). Likewise, DuhWhine only getting 26% of moderates vs. Brown is unrealistic (Blackwell gets 22%). This sample is skewed by liberals (supposedly only 16% of the sample)-- probably Democrats-- posing as moderates and conservatives. If you adjust the numbers giving Blackwell just the same % of conservatives as liberals for Strickland, he is still down outside the MOE, but by a more likely 53%-47% (rounded). This may be a little high considering a TEL backlash (the true deficit is probably between 8%-10%), but it is alot more realistic than SurveyUSA's numbers.

If you apply the same logic for DuhWhine and give him 40% of moderates, it is a 50/50 race; 50% of moderates gives him a lead, 55%-45%.

These above imputed numbers are more likely than the wild 10- and 20-point swings suggested.

Matt Hurley said...

Joe, I saw these comments over at RAB and nearly posted an update with a link to them. I couldn't agree more...