Sadly, exit polls aren't an exact science...

[Drunken Irishman]

They have margins of errors and can be manipulated based on many factors - turnout in certain areas and certain groups who are more open to taking the exit polls. The latter is always a concern because some groups might not feel as inclined to fill out the survey and it could change the overall impact of the exit poll results.

1992's Republican Primary is a good example of this. In early primaries between Bush & Buchanan, exit polls generally showed Buchanan doing better than the official results. The discrepancy is tied to the belief that Buchanan's supporters were more willing to fill out the exit poll surveys and therefore skewed the overall results.

It's why networks rarely call a race solely on the exit polls, unless the margin is sizable enough that it's almost a given the race will go a certain way (think in presidential races when, with 0% of the results in, the networks will call a state like Utah).

Exit polls are not the definitive answer on a race. They can characterize a race, but that's about it. It's ultimately up to the raw vote total that will tell the network whether the exit polls are right or if they're off. Remember that as we enter November. Exit poll data is also wildly inconsistent from certain points of the day - with the theory out there being that early exit polls favor Democrats, while the later numbers favor the Republicans. I don't know how valid that is, but certainly exit poll data can be adjusted during the day and, more importantly, once the raw vote total comes in, as they update the numbers to reflect the true total.

That appears to be what happened today. It also happened in the primaries. Initial exit polls from Alabama, I believe, had Romney beating Santorum. But as the raw vote total came in, and the exit polls adjusted to that data, Santorum led in the finalized exit poll numbers. So, I guess it is what it is. Sucks, I know, but always look at exit polls with a cautious eye. They're not there to tell you who won. They're there to tell you who might win - just like poll surveys.

[bigdarryl]

1. No shit we still haven't leaned our lesson from 2000

After that stolen election the networks were all on agreement to not rely on them.But I see there right back to there old shit

[snot]

2. As I understand, they used to be a lot more accurate than they've been in recent years –

and since 2000 or so, the discrepancies seem to have steadily widened, despite efforts to "correct" the exit poll process.

[Drunken Irishman]

3. I think it depends...

In fairness, 2000 was really the first 'competitive' election since the 1976 presidential election. So, the exit polling data wasn't nearly as valuable in '80, '84, '88, '92 and '96 when we had blowout victories across the country.

I do know that in '76, the networks prematurely called Illinois for Carter based on exit polling. In the end, Ford barely carried it and even though it didn't ultimately decide the election, it was a crucial state that I think allowed the networks to call the presidency for Carter maybe before they would have had Illinois not been called.

[TheWraith]

4. Truth is, their accuracy was always overrated.

But since attention began being paid to them around 2000, there's been an increase in the extent to which some people refuse to take them. Mostly Republicans, out of some mixture of paranoia and/or assholishness, refusing to tell anyone who they voted for.

Exit polls can be fairly accurate if you map them against real results to check the gap, but they shouldn't singlehandedly be relied on.

[October]

5. Yeah, dems lose no matter the exit polling.