Hillarious

[Phony Moderate]

Am i the only one who finds it hillarious that the networks call the safe states using exit poll data, and even call some of the battleground states before even a 1/4 of the votes have been counted (for example, Ohio for Obama in 2008, Florida for Gore in 2000 etc)? It's hillarious. The exit polls are JUST polls, they COULD go wrong. At least 75% of the vote should be counted in every state in my opinion before it is called.

[Franzl]

Am i the only one who finds it hillarious that the networks call the safe states using exit poll data, and even call some of the battleground states before even a 1/4 of the votes have been counted (for example, Ohio for Obama in 2008, Florida for Gore in 2008 etc)? It's hillarious. The exit polls are JUST polls, they COULD go wrong. At least 75% of the vote should be counted in every state in my opinion before it is called.

Nope...just mathematical probabilties. There's nothing wrong at all with calling safe states with exit poll data...especially if that exit poll data confirms pre-election polls. Exit polls can be off...but they're not off by 20 points.

And even the closer states (like Ohio in 2008)...why on Earth do you need 75% of the vote? All you need there is a couple of votes from representive precincts...and you analyze the swing there compared to previous election.....compare that to other precincts from other parts of the state.....and it's not very difficult to make a projection with near certainty.

[Frozen Sky Ever Why]

For some reason the media tends to call blue states before red states. In 2004, They called NJ & CT immediately but waitied to call MO & AZ.

[Franzl]

For some reason the media tends to call blue states before red states. In 2004, They called NJ & CT immediately but waitied to call MO & AZ.

There's a simple explanation. Exit polls favor Democrats on average.....meaning, for example, in 2008, Pennsylvania was D+10.....but the exit poll showed D+15.

Arizona, though, was about the same margin for the Republican, but the exit poll showed a virtual tie....making it impossible to call very quickly.

[Lunar]

I assumed this thread was about Hillary

[Queen Mum Inks.LWC]

Am i the only one who finds it hillarious that the networks call the safe states using exit poll data, and even call some of the battleground states before even a 1/4 of the votes have been counted (for example, Ohio for Obama in 2008, Florida for Gore in 2008 etc)? It's hillarious. The exit polls are JUST polls, they COULD go wrong. At least 75% of the vote should be counted in every state in my opinion before it is called.

Nope...just mathematical probabilties. There's nothing wrong at all with calling safe states with exit poll data...especially if that exit poll data confirms pre-election polls. Exit polls can be off...but they're not off by 20 points.

And even the closer states (like Ohio in 2008)...why on Earth do you need 75% of the vote? All you need there is a couple of votes from representive precincts...and you analyze the swing there compared to previous election.....compare that to other precincts from other parts of the state.....and it's not very difficult to make a projection with near certainty.

Franzl's right.

I assumed this thread was about Hillary

That's what I actually figured at first too.

[The Mikado]

Am i the only one who finds it hillarious that the networks call the safe states using exit poll data, and even call some of the battleground states before even a 1/4 of the votes have been counted (for example, Ohio for Obama in 2008, Florida for Gore in 2008 etc)? It's hillarious. The exit polls are JUST polls, they COULD go wrong. At least 75% of the vote should be counted in every state in my opinion before it is called.

Not THAT wrong.  If a candidate is up 30% with a 2% margin of error (exit polls are usually around that), why would you hesitate to call Utah for McCain?  I mean, the odds of being wrong at one tail end of 7.5 confidence intervals...I'm not even going to look it up.

For some reason the media tends to call blue states before red states. In 2004, They called NJ & CT immediately but waitied to call MO & AZ.

Usually, but remember how late into the night they called PA for Kerry in 2004?

And MO last time was called days later than it should've been, but it was DEAD even for election night.

[Smid]

This article is a few months old, but still relevant. It's written by a political journalist over here who is one of the main election-night pundit/psephologist and is one of the most accurate at calling elections in the Australian media (probably not the pre-eminent journalist for behind-the-scenes deals or day-to-day governance, but in terms of predicting elections and commenting on elections, he's one of the best).

http://www.abc.net.au/elections/qld/2009/guide/predictingelections.htm

He's updated it to discuss the impact of the redistribution (redistricting) prior to the Queensland election, but since presidential elections don't incorporate redistricting, that's not overly important. Still makes for an interesting read, though.

Basically, he says that the swing is unrelated to the size of the booth/precinct, so he works off that rather than the actual vote. Of course, we don't have exit polls in Australia (well, actually, I think we did recently - but given that our federal electorates are a quarter of the size of your congressional districts... the inaccuracy makes it pretty much pointless).

[jimrtex]

This article is a few months old, but still relevant. It's written by a political journalist over here who is one of the main election-night pundit/psephologist and is one of the most accurate at calling elections in the Australian media (probably not the pre-eminent journalist for behind-the-scenes deals or day-to-day governance, but in terms of predicting elections and commenting on elections, he's one of the best).

http://www.abc.net.au/elections/qld/2009/guide/predictingelections.htm

He's updated it to discuss the impact of the redistribution (redistricting) prior to the Queensland election, but since presidential elections don't incorporate redistricting, that's not overly important. Still makes for an interesting read, though.

Basically, he says that the swing is unrelated to the size of the booth/precinct, so he works off that rather than the actual vote. Of course, we don't have exit polls in Australia (well, actually, I think we did recently - but given that our federal electorates are a quarter of the size of your congressional districts... the inaccuracy makes it pretty much pointless).

Australia has mandatory voting, so you don't have to worry about turnout.  In the USA, if a precinct goes 90% Democrat, what is more important is whether 1500 or 1000 people vote, rather than any swing.

In his example of Fitzroy, there actually does appear to be a small relationship between swing and box size, with the smaller boxes (< 500) swing towards Labor more than the larger boxes.  In the progressive count during the election, this produced an error of about 2.6% from the final result when 21.5% of the vote was counted (2/3 of the boxes).  It didn't matter in an election when one party got 70% of the vote.

Exit polling is less accurate than actual counts of the vote.  You first have to select which precincts to poll.  In Fitzroy the standard deviation in swing among polling booths was 5.3%.  The interviewers then select 1 out of N voters leaving the poll.  This could possibly introduce bias from interviewers/voters.  Voters rarely exit a polling place any single file.  Will interviewers reliably chase after a voter who comes out of a side door and heads toward a different parking lot or threatens to mace the interviewer.  In Florida 2000, only about 50% of interviews were completed.  Again you have a possibility of a bias.

And US elections may be more personality driven, so that they are more like the case of the Queensland case after a redistribution, where there was a credible contest.

And TV networks may ignore what their data actually shows.  They can report that Sen. Blotto has been re-elected and that they project he will end up with 67% of the vote (it might be 95% chance of being between 65% and 69%),  If this is an outlier, and Blotto actually ends up with 63% or 71%, they won't break in to announce the mistake.  Instead, they will steadily update their projected percentage.  By the end of the night, they will be almost perfect.

But if the 1 in 20 races that they miss on has a different winner, people will remember.