1. Exit polls have a much larger intrinsic margin for error than regular polls.
2. Exit polls have consistently overstated the Democratic share of the vote.
3. Exit polls were particularly bad in this year's primaries.
4. Exit polls challenge the definition of a random sample.
5. Democrats may be more likely to participate in exit polls.
6. Exit polls may have problems calibrating results from early voting.
7. Exit polls may also miss late voters.
8. "Leaked" exit poll results may not be the genuine article.
9. A high-turnout election may make demographic weighting difficult.
10. You'll know the actual results soon enough anyway.
Read in detail Ten Reasons Why You Should Ignore Exit Polls here .

Pam's House Blend/ Darkrose writes:
Since Wednesday, people have been tossing around the number "70%". That's apparently now being taken as gospel, that 70% of African Americans supported Prop 8. Homophobia is a serious problem in the black community--as in every single ethnic, racial and religious community in the country. But using the results of a single, non-representative poll to claim that 70% of all black people in the country--or even in California--are homophobic defies not only logic, but basic math. The guys at 538.com, who have been my heroes this election cycle, have detailed why exit polls are inherently flawed. And Shanikka's diary on Kos should be required reading for anyone who really wants to understand why the methodology for the [CNN] poll--and remember, there's only one that's being cited--was so flawed as to be almost useless.
CNN doesn't say where they polled, but I'd bet money that they were in Southern California, probably in LA County. The demographics of LA County do NOT represent those of the entire state. (As an aside, the single county with the largest black population in the state is Alameda County, where Oakland is. Alameda County went No on 8.)
The poll also didn't count the estimated 3 million voters who turned in ballots by mail. My wife and I were among them, because our polling place is in an inconvenient location with bad parking. I didn't have to check a box stating my ethnicity; there's nothing that would trigger a "Warning! Black Voter!" alert when my ballot got to the elections office. Absent this information, the only source of the 70% number is a single poll of 2,240 people.
Shanikka's diary goes into a lot more detail than I have room or math skills for, but she makes the case that with African Americans as 6.7% of the state's total population, a sample that has the black vote as 10% of the total voting population is inherently suspect. She calculates that if black voters really were 10% of the total voting population, then that would be disproportionate to the number of eligible black voters in the state to begin with.
The number that interested me most, though was 2,240, the total number of people sampled in the exit poll.
10% of 2,240 = 224
70% of 224 = 157 (rounding up)
In other words, based on the 157 people who were approached by pollsters and who told them that they'd voted yes on 8, people are extrapolating to say that 1,582,000 people are all homophobes.
One last note: if despite the evidence, you want to take the CNN poll as representative, the numbers indicate that 3,185,452 white people voted for 8. That's almost a million more people than the entire black population of the state of California. I'd say there's plenty of homophobia to go around. Perhaps rather than trying to figure out who to blame, other than the people who put 8 on the ballot and the people who funded it, we might want to focus our energy on getting this legal obscenity overturned. READ MORE
Facts belie the scapegoating of Black people for Proposition 8
Daily Kos/ Shanikka writes:
"... virulently racist idea that Black people are to blame for the passage of Proposition 8 here in California. It is an idea grounded in utter myth, a complete lack of knowledge about anything related to Black people's presence in California, and just plain old scapegoating."
Shanikka discusses the following myths:

Shane aka QTC
Editor and PublisherQueerTwoCents
You Might Have Missed