I have heard of this phenomenon of course and I can understand how this can be. But I have pretty much concluded in past election cycles that there is a pro liberal Democrat bias in polling results.

 

I am not saying this bias still persists but I have, in the past, been used to mentally correcting published results of most polls by shifting 4-10% from the liberal Democrat to the Republican. My guess was that polling organizations skewed their results by polling people who would more likely be open to the Democrat candidate. In addition there was the idea that people might also want to respond in a politically correct way to a pollster where race was not a factor.

 

I noticed that one or two polling organizations tended to give more accurate results than all the rest.

Apparently, it makes a difference whether the polling sample is of people who are likely voters as opposed to people who are merely registered voters.

 

In Obama's case, ideology and race could combine where people believe they are being polled by a liberal.

 

 

Actually, I would like to see someone analyze African American voter motivations with respect to elections where an African American is a liberal Democrat candidate. Was it something like 97% of African Americans who voted for Obama? My impression is that if anything, African Americans are at least slightly more conservative than Whites. All you have to do is note the solid support African Americans give to defense of marriage measures such as NC's Amendment One and that the Obama came out of the closet, so to speak, on gay marriage (surprise surprise!).

 

You are correct that polls tend to over-sample democrats, resulting in a liberal bias.

 

However, the degree of bias is hard to quantify because the only way to measure accuracy is to compare the poll against the actual election results.  If the pollster are wrong, they can simply say that opinion changed as the election got closer.  Also there is a "non-response bias."  Some people just don't want to answer pollsters, which makes them less accurate.

 

Some good rules-of-thumb:

1) Polls of likely voters tend to be more accurate than polls of registered voters or average adults.

2) The larger the sample size the more accurate the result.

3) The closer you get to the election the more accurate the results tend to be (nobody want to be wrong).

4) Polls by media organizations & political parties tend to be less accurate.

 

In the most recent election, Rasmussen & Pew gave the most accurate results.