Fun with Polls - Round 1

Rich Lowry posted the results of an Ipsos/AP poll that says that in a face-off between Hillary Clinton and John McCain, Hillary would beat McCain by 9 points (50-41).

Now, I hate most political polls since they are accepted at face value, with no concern for what the internal demographic makeup of those polled might be. Overweight the poll with Democrats and you can skew the results one way, overweight it with Republicans (Yeah, right!) and you can skew the results another.

So, does Hillary *really* lead McCain by 9 points? Maybe. But I would argue that the internal demographics of this poll skew it to her advantage a little too much to really believe those numbers. Ipsos' demographics are somewhat vague, but there is enough info to make some interesting observations. They also don't break out results for "Registered Voters", let alone "Likely Voters".

(2004 demographics were obtained here)

1) First, the make up of the poll is 45.7% Dem/Lean Dem, 34.5% GOP/Lean GOP, and 19.8% Unknown (Independent?) Now we may have had a greater amount of Democratic voters participating in the primaries, but that is mainly due to the fact that they still have an ongoing contest. In an actual cross-section of actual "likely voters" in this country, there is no way that Dems outnumber GOP voters by 11%. Maybe a few, but not 11. In the 2004 election, which was "extremely important" for the Dems to win, the breakout was 37%-37%. Hence the first portion of the poll skew.

2) It appears to me that they slightly overweight younger voters. In 2004 ages 18-24 was 8%, 25-29 was 9%. For this Ipsos poll, ages 18-35 made up 30% of those polled. While they don't line up exactly, that still seems like a bit of a disparity to me. Thus skewing the poll with a bloc of voters that breaks for Dems.

3) This poll has an unemployment rate of 17%. Holy stagflation, Batman! Thats a lot of unemployed voters. Gee, I wonder if they are going to break for the Dems?? For a bit of context, in 2004, those with incomes below $15k (which by my math would include $0, or unemployed) was only 8% of all voters. More skewed....

4) In terms of education those polled that have a HS degree or less make up 52% of the voters. Now there are no 2004 numbers from my source to compare this to, but methinks that this is out of sorts. The last US census shows that only 46% of those over 18 have a HS degree or less. A small disparity. Additionally, I would suspect that those who have not finished HS might be less inclined to vote, but those that do would break for the Dems. A little bit more skewed....

5) The racial demographics also seem a little skewed. Those identified as 'White' make up 74% versus 77% in 2004. Meanwhile those identified as 'Hispanic' make up 13%, whereas they only were 8% of the 2004 electorate. While the Hispanic demographic is growing, 5% in leass than 4 years?? Whites historically break for the GOP, while Hispanics break for the Dems. Even more skewing.

Now, none of these items are egregious enough that they would call the poll into question, but a pinch here, and a pull there, and you can make the numbers dance in a particular direction. That is why I never trust these types pf polls, because more often than not, their internal demographics do not reflect reality when it comes to 'Likely Voters'. So, if you oversample those demographics that break for the Dems, then you produce a poll that will be skewed towards the Dems.

Ultimately a poll like this is fairly irrelevant anyhow. National polls mean nothing. A Presidential election is not national election. It is lots of many state elections to select someone to lead the confederacy (oops, bad word) of those states. Bring me 50+ good state polls, tally up Electoral Votes, and project a winner, then I'll buy it.

But Don't bring me Hillary 50 - McCain 41%. It tells absolutely everybody, absolutely squat!

No comments: