But let's not overplay our hand in attacking Pew.

Remember that this is the only time Romney has led in a Pew poll. It's been a very good pollster for Obama so far this year. We may just have to accept the reality that Obama had a bad debate and that showed up in the poll.

It's not like we're dealing with Rasmussen here. This is a respected pollster. They may be overdoing the Romney numbers, but let's hope that once the Romney bounce subsides we will see things come back to normal.

In fact, the danger for Romney is that if he doesn't maintain this bounce, he will end up going back down and it will look like he can't maintain popularity with the people for more than a few days.

They feel obliged to pick up the phone when someone calls? Higher percentage have land lines?

Kidding (maybe) :p

Is not productive. Remember how a few weeks ago the Reps were saying the poll numbers were
Skewed?

are from the south. Is that a representative sample?

The numbers *may* be legit, but reflect the time-frame in the immediate aftermath of the debate, with only 15% polled on Sunday.

When you look at the balance of people polled, this particular poll does seem skewed, with heavy southern and older voters, much fewer from the other regions and age groups.

Which is also why a massive surge in youth voting has the potential to win elections out of nowhere.

I have significant experience in polling (research) and there's an easy answer for the seniors question. Pew, like many polling firms, randomly call people. Typically certain types of people are more likely to answer the phone than others. Seniors are usually retired and have more time and availability to answer the phone. Therefore the sample will show more seniors.

So does this skew the results? No. Why? Because of a practice called weighting. Weighting helps to correct the variables of randomness by making the "worth" of certain demographics match the population.

Example, if the senior Likely Voter population nationally is 30% and the polling data yields 45% seniors in the sample, the data will be weighted "down" to match the 30% national population Likely voters for seniors.

The same is for the younger male demos. Young men are the least likely group to answer the phone and take a poll. Let's say the Men 18-34 likely voting population is 8% of the electorate. But the polling data only yields 3% in the sample, the data will be weighted "UP" to match that 8% number.

If polling companies did not weight the data then the poll results would be extremely unreliable. This includes Nielsen TV ratings and Arbitron Radio ratings. Weighting MUST be used to correct the randomness factors of polling.

Why not quota poll? In other words once you reach the 30% of seniors threshold stop talking to those people. Some polling companies quota sample. However, once that happens it is no longer random sampling and is less reliable.

That the preferred practice is to weight to the relevant census data NOT some guesstimate/projection of what the electorate will look like?

Companies like Arbitron, Nielsen, Scarborough Research, etc. use census data (or annually released projections modeling census data) for the demographic information. They use census data for gender, age, ethnicity, geographic data.

These above companies, and others, are trying to mirror a population. The problem with the national or regional election polls is that you are not polling the total population. The election populations are either defined by registered voters and/or likely voters. I'm not sure the accuracy of the benchmark estimates for the two different universes. You would think they would broadly be representative of the general population but you'd probably find younger demos less likely to vote, older demos more likely to vote, etc.

But they create the universes and then weight the data to those universes. Most likely they use voting rolls as there benchmark to weight the data.

I can make a poll showing that mitt has 95% of the female vote...if I only ask female voters in Alabama between the hours of 2 and 4 PM.