Politics: The challenges of polling according to Michael Barone

The polling numbers are not looking good for Mitt Romney.

However, polling is an inexact science.

Check out this interview with Michael Barone who is one of the experts in looking at polls.

Excerpt:
The Pew Research Center reports that only 9% of the people that it calls are responding to polls. That’s way down from historic levels, and it raises the question are those people representative of the population as a whole that they’re trying to sample? You know, one thing that polls can’t tell you is the characteristics of people who won’t be polled. So that raises some serious questions. Are we getting skewed samples? We know from the exit poll phenomenon over the last many cycles that the exit poll results tend to come in more Democratic than the actual vote does, and measured at the same precincts.

Since I've been voting in Presidential elections, 1984, 1988, 1992, 1996 and 2008 were the elections where there was essentially no doubt who would win. The only question was what the margin of victory would be. Only in 2000 and 2004 was it a question mark as to who would win going into election night as I watched.

If the Gallup 6% (as of Sept 25) Obama lead holds up over time, in particular after the first debate, then it may well be over barring something completely out of left field.

Skewing of 2 to 3% perhaps could be accounted for by two possible factors:
1. the historic tendency of Democrats to over-poll in polls as mentioned by Michael Barone in the excerpt above and
2. the so-called "Bradley effect."

If these phenomena are actual, then a 6% lead may really be only 3%.

Besides actually getting someone to talk on the poll call (mentioned above), the next challenge is to ask the person questions that probe how likely they will actually vote since a certain percentage of voters say they will vote but don't actually do so.

Rasmussen (as of Sept 26) has the race tied at 46-46. He has consistently had the race closer than other pollsters.  When the ballots are finally counted, it will be interesting to see if his "likely voter" screen resulted in greater accuracy or not.

UPDATE: Another interesting phenomena could be the self-identification of Republican leaning voters as Independent voters. Geraghty explains as follows: Perhaps these are Tea Party conservatives fed up with a GOP they find too “establishment,” etc. If the Democratic share of the vote were stable, it would just mean voters are shifting between these two other self-classifications. If this skewing is occurring than adjusting the D/R ratios downward would overstate Romney's support.