Explain this to me, I don't get it?

[Rich in KCK]

I'll be honest I didn't pay that much attention in government classes but I did in Math classes.

Just looking at the two states near me Kansas, and Missouri I notice one thing in particular. How the hell can they project a winner this early? Really I don't get the math. In 2004 almost 2.5 million people voted in MO, so far only around 800,000 votes have been reported with a difference of only 20,000 votes between the two front runners. With only around 30% of the votes counted how do they pick a winner so early? It's crap like this that make me hate the media and politics.

[noxiousdog]

I'll be honest I didn't pay that much attention in government classes but I did in Math classes.

Just looking at the two states near me Kansas, and Missouri I notice one thing in particular. How the hell can they project a winner this early? Really I don't get the math. In 2004 almost 2.5 million people voted in MO, so far only around 800,000 votes have been reported with a difference of only 20,000 votes between the two front runners. With only around 30% of the votes counted how do they pick a winner so early? It's crap like this that make me hate the media and politics.

It depends on polling and the counties that are reporting.

Currently I see Bush with a 120,000 lead in MO. In order to make up that difference, Kerry would have to have a complete reversal of the current count. If polling is consistent across the state, it makes it unlikely to impossible for him to come back.

[The Meal]

Rich, a lot of it is from work they've done before the election. History, polling, that sort of thing. Basically, you or I could have called Wyoming last week without a single vote being cast, for instance. There is some journalistic integrity this year in that they do *not* call elections until the polls are closed, but other than that, probably 50% of the states out there were *known* to be solid for one candidate or the other ahead of time. And they're pretty good with breaking down expected values by county, so as you start to get certain data in from some precincts, they have a good feel for what sort of data to expect from the outstanding voters.

Mathematically it *is* a crock that they call states "solid" before they've received 51% of the votes, but realistically they can make really really good guesses even with only 1% of the vote in.

~Neal

[Grifman]

I'll be honest I didn't pay that much attention in government classes but I did in Math classes.

I guess you missed statistics then, right?

It's quite simple. You take a sample that is representative of the general population and extrapolate from that sample to the population as a whole. That's the whole basis of statistical analysis.

Grifman

[Rich in KCK]

It still irritates me even though I know you're right about being able to call a state a week ago or longer, hell I live in one of them. It's just that I was looking at the BBC NEWS site when I posted that and it bothered me that some states that had a larger variation between Bush & Kerry than MO did were still considered up in the air while MO wasn't. Living in Kansas it is discouraging to know that the state always (13 out of 14 past elections) votes the same way. It further pisses me off that my location on a map gives others a preconceived notion of how I think. I think that is what really gets to me, news sites posting opinions based on preconceived notions and history instead of cold hard numbers. I guess I'm just too analytical.

[Rich in KCK]

I'll be honest I didn't pay that much attention in government classes but I did in Math classes.

I guess you missed statistics then, right?

It's quite simple. You take a sample that is representative of the general population and extrapolate from that sample to the population as a whole. That's the whole basis of statistical analysis.

Grifman

That's kind of my point. How do you know that your sample is truely representive of the general population's mindset. You can guess that it is and use history to back up a theory but it can never be considered fact. I'm 33 years old and have never once had someone poll me about any issues of importance. I guess I'll just go read a book and wait a day or possibly longer until they really know who one what. I'm not wired right to watch this stuff unfold.

You're right I didn't take stats.

[Grifman]

I'll be honest I didn't pay that much attention in government classes but I did in Math classes.

I guess you missed statistics then, right?

It's quite simple. You take a sample that is representative of the general population and extrapolate from that sample to the population as a whole. That's the whole basis of statistical analysis.

Grifman

That's kind of my point. How do you know that your sample is truely representive of the general population's mindset. You can guess that it is and use history to back up a theory but it can never be considered fact. I'm 33 years old and have never once had someone poll me about any issues of importance. I guess I'll just go read a book and wait a day or possibly longer until they really know who one what. I'm not wired right to watch this stuff unfold.

You're right I didn't take stats.

You match demographics - you get a sample that has the same mix by sex, age, geography, income, etc. I don't know all the factors, but by matching demographics the assumption is that people of the same demographic group vote similarly - you infer a mindset with your demographics.

Grifman

[D'Arcy]

How do you know that your sample is truely representive of the general population's mindset.

You pick a random sample. A random group of people is representative of the population in the same way as a glass of water taken from the ocean is representative of the ocean. There's always a margin of error associated with such a method, expressed in a confidence integral: The actual distribution will be within X% of the sample distribution Y% percent of the time.

The larger the sample, the better the prediction. Projections are based on truly random samples of actual votes, so they're quite reliable. With 800,000 votes counted, the projection you mentioned is less than 1% off more than 99% of the time.

Pre-election polls, on the other hand, are usually not based on representative samples. Sure, the people are randomly chosen from the phonebooks; but people can lie or refuse to answer, or they might be absent, or a self-proclaimed household spokesperson answers for them. These samples aren't random, they're self-selected, and the raw data typically reflects that: It's totally off the mark. Nobody would buy it. So the Pollster-Augurs have to adjust it according to their gut feelings, which they call "weighting". The goal is to make the data look more like the result of the last election so that it appears more credible. The method is about as scientific as reading the future from the flight of the birds, and about as prone to become a self-fulfilling prophecy.