Calling the race with < 1% reporting? HOW!?

[sidster]

Can someone explain to me how can MSM call Maryland for Obama with
less than 1% reporting, and they are even showing Clinton ahead
of Obama in numbers?!

 

[Wolf]

A) Why are you watching Fox?
B) Whoa.

 

[zadrock]

From the exit polls - if they are one-sided enough, they can call it just based on that. The fact that the first precincts to report have more Clinton votes should be viewed as an aberration, assuming their exit polls are accurate.

Z

 

[sidster]

From the exit polls - if they are one-sided enough, they can call it just based on that. The fact that the first precincts to report have more Clinton votes should be viewed as an aberration, assuming their exit polls are accurate.

Z

I'm genuinely curious about these exit polls. How are they conducted?
How many polling places are sampled? Basically I'm trying to figure out
how biased these exit polls can be?

 

[Mesogen]

How do they get the exit poll results out so quickly?

 

[Bruno]

Either way, exit polls or not, it is crap.

 

[kathy88]

I'm with previous poster.

WHY are you watching Fox?

Thanks for getting the pic

 

[pacelli]

This is fucking bullshit man.

 

[zadrock]

I don't know how they do the exit polls or where they go - but the major news media was embarrassed by what happened in 2000. You can bet that if they call a race, they are at least 99% sure that they are correct.

Look at what happened in VA. The vote came in, Huck was in the lead at first, then McCain took a narrow advantage. With 50% reporting, and McCain holding a 1-2% lead, the news media called it for McCain. Then, his lead expanded substantially. Clearly the exit polls showed that the remaining areas to be counted were heavily McCain.

This isn't the first time that a race has been called for the apparent "loser." When Cardin and Steele faced off for the Senate seat in MD, they nearly called the race for Cardin, despite the fact that he was trailing at the time by a fair margin. The only reason they didn't call it for him was because of the strange demographics - Cardin is a white Dem and Steele is a black Repub - and the rural areas of MD (Republican areas) were counted first, giving Steele a lead. The assumption by the news media and the exit polls showed that Cardin would make up the difference and then some in Baltimore and the DC suburbs, heavily Democratic areas. But they held off longer than normal because of the possibility that the blacks would vote for a black candidate (even a Republican) over a white Democrat. That didn't turn out to be the case - they called the race shortly after Cardin took the lead and the early returns from Baltimore confirmed they were voting heavily Democratic.

Z

 

[RonRules]

Can someone explain to me how can MSM call Maryland for Obama with
less than 1% reporting, and they are even showing Clinton ahead
of Obama in numbers?!

It's actually OK to do that, provided that the sample is taken across all precincts.

When they take a nationwide poll, they only call around 600 to 1000 people. That'1000/300,000,000 or 0.0003%

It's called statistical sampling. You only need a small sip of the spoon to taste the soup. You don't need to eat the whole thing!

I would say that 1% is probably risky, particularly if a large group of precincts have not reported yet.

I need to add that in the case of the Obama/Clinton above, they are surely using the exit polling to make the Obama decision.

 

[RonRules]

How do they get the exit poll results out so quickly?

They talk to people thoughout the day. They have the whole story written up and going to press the second 8:00 PM rings.

It's fine, but a bit risky if the election is tight.

 

[sidster]

They talk to people thoughout the day. They have the whole story written up and going to press the second 8:00 PM rings.

It's fine, but a bit risky if the election is tight.

I understand what you are saying with the analogy of taking a small
sip to be able to taste the soup, but if you have a roach at the bottom
of your soup bowl, you won't see it 'til you are about 80% done, unless
you stir the soup around a bit before you start sipping it....

That said, I can't imagine exit polls only taking account 1%... they gotta
have people in good percentage of polling places... and as others have
mentioned, I have not seen any exit polling ever in my voting "career"
(albeit a short career thus far) ... then again I'm in CA and maybe CA
demographics are much different than places like Maryland?

 

[pacelli]

It's actually OK to do that, provided that the sample is taken across all precincts.

When they take a nationwide poll, they only call around 600 to 1000 people. That'1000/300,000,000 or 0.0003%

It's called statistical sampling. You only need a small sip of the spoon to taste the soup. You don't need to eat the whole thing!

I would say that 1% is probably risky, particularly if a large group of precincts have not reported yet.

I need to add that in the case of the Obama/Clinton above, they are surely using the exit polling to make the Obama decision.

RE: Statistical sampling. Yes, a statistical test could predict a trend based on initial data provided it was randomized. I guess they'd randomize the precincts first, and then sample randomly from those precincts.

My question is: Is it ethical to come up with a projection and broadcast the results to the public?

I think there is a psychological impact when a viewer sees the projection and "% reported" is so low. I think it is much healthier to leave the stats out of it, and let the numbers work themselves out. I enjoy watching the horserace of it =)

 

[zadrock]

Sure it's ethical. They do claim that it's a "projection."

Look, we do hate the news media here, but they are extra EXTRA careful with these projections - remember 2004? They wouldn't call Ohio for Bush, even with 99% reporting and Bush holding a small, but not insignificant lead. I really think they've got this figured out. They learned their lesson in 2000.

Z

 

[kathy88]

It says LESS than 1%. That's just shitty for everyone involved.

It makes too many people question the process. (Not that it doesn't need questionning....)

 

[Molly1]

They are hardwired to diebold.

 

[JordanQ72]

RE: Statistical sampling. Yes, a statistical test could predict a trend based on initial data provided it was randomized. I guess they'd randomize the precincts first, and then sample randomly from those precincts.

Actually, exit polling is far from location or target randomized. They do very specific non-random stratified sampling that they combine with past historical data for weighing.

 

[sidster]

CNN's Maryland Democratic Exit Poll (linky) (for reference...)