Poll Puddling and Quinnipiac

The new Quinnipiac poll has Obama leading McCain in Pennsylvania, Florida, and Ohio. If those numbers are accurate and hold up, it would obviously indicate that Senator Obama is in great shape, picking up 47 electoral votes to add to his position. However, I noticed that Quinnipiac hid a bit of its methodology. In opinion polls, the demographc weighting is an important element in their accuracy. The gist of the Quinnipiac polls was not really anything new; young voters prefer Obama and older ones prefer McCain, women lean towards Obama while men lean towards McCain, and so on.

It’s also important to notice that Quinnipiac never defined what they mean by ”likely voter””. A respondent who voted in the primaries, for example, is far more significant as an indicator of true voter mood than someone who just says he thinks he will vote in the fall, still months away from the mid-June poll.

I also noticed clear discrepancies in the Quinnipiac poll. Quinnipiac, for example, says President Bush has a 22% Job Approval rating in Ohio, against 29 percent from a University of Cincinnati poll.

I was also preplexed by Quinnipac’s announcement that President Bush enjoys only 60% from Republicans, when the polls I have seen show around 80% from Republicans. The stated numbers demonstrate a lean to the Left, from what I see. How much of lean depends on how you account for the discrepancies. The problem is while the skewing is evident, it appears to be a result of the methodology rather than simply the weighting. Therefore, the numbers may be taken to reflect the situation only in a general manner and with the normal caveats that Quinnpiac is a good poll, but hardly the most reliable.

A poll which misses by an average of 7 points is not where I would look for an election barometer.