6 ms·

Statistics tells you the odds that the exit poll would disagree by some degree with the actual result. If the odds show, for example, only a 1 in a 100 million

by genwin 14y ago

Statistics tells you the odds that the exit poll would disagree by some degree with the actual result. If the odds show, for example, only a 1 in a 100 million chance of it happening, an investigation should take place. That you've never hit an exit pollster doesn't say much, because the odds can be computed with high confidence by polling only a tiny fraction of the voters. Statistics is highly valuable despite not being proof. For example, it's a foregone conclusion that Obama will win today, based on earlier polls. It's not certain that he'll win, but the odds of him losing are nil, the stats show.

danielweber 14y ago
And who regulates the exit pollsters? Normally I wouldn't worry about this, but if you want to give them power to invalidate an election (or force an "investigation," whatever that means), then you shouldn't just let them run around doing whatever they think is best. but the odds of him losing are nil, the stats show. I don't even know what mindset produces this thought. While Nate Silver doesn't think Romney will win, he gives changes that are a hell of a lot bigger than "nil."
- genwin 14y ago
  Exit pollsters should be regulated, I agree, if only to avoid an expensive unnecessary investigation. The media says that Obama and Romney are neck & neck, because that gets more viewers, which sells the advertisements they depend on. (Every presidential election in your lifetime is guaranteed to be declared a "virtual tie" by the media--you'll hear those exact words every 4 years.) Those who understand stats know (with almost certainty) that it'll be a landslide for Obama in the electoral college.
  - twoodfin 14y ago
    How old are you? Within my lifetime there have been at least six presidential elections that were absolutely not reported as "virtual ties", including the last one.
    - genwin 14y ago
      Found in 5 seconds: "Gallup Daily: Obama-McCain Race Reverts to Virtual Tie" Found in another 5 seconds: "National Poll Shows Bush, Kerry in Virtual Tie" I could go on.
      - twoodfin 14y ago
        That Gallup poll is from June. The race was a virtual tie then. November reporting is a different story. And Bush/Kerry was also pretty damned close. Not one of the races I was referring to.
        genwin 14y ago
        That's just the one I found in 5 seconds. This one's from mid-August 2008: "New [LA Times/Bloomberg] poll shows McCain - Obama in virtual tie". And from Sept: "[NBC News/Wall Street Journal] Poll shows Obama and McCain in virtual tie". 3 months later, still a virtual tie! I'm sure I could find a closer one. I didn't say the media would always report that in November. I went back to 1992 and still found "virtual tie" reports. That's 5 elections in a row.
- btilly 14y ago
  Exit polls are like a canary in a coal mine. They can show evidence of manipulations that might have happened. If an investigation finds that no such manipulation seems to have occurred, then no harm is done. Unfortunately current US elections are unauditable because voting machines DO NOT produce any kind of verifiable voting trail. Furthermore the apparent conclusion of the US media is that past discrepancies are evidence that their exit polling is flawed rather than that the voting process is being manipulated. (They are biased to decide that way because accusing the winners of manipulating the vote - even if it is true - pushes them away from the appearance of neutrality that they try to keep.) Furthermore you can't do exit polling on mail-in ballots. So this potential signal of flawed elections is being lost at the same time that manipulation of elections is becoming even easier. Personally if polls are overwhelming, and elections are in line with polls, I believe the result. But I believe that there are lots of people trying to slip a thumb on the scales, so if polls are close, I am dubious about the results. I personally believe that current evidence is that the Republicans are more successful in getting their thumbs on the scales, so I'm doubly suspicious when Republicans mysteriously do better than polls indicate they should. I want us to have elections that I trust. I do not want crappy election machines. I want verifiable paper trails. I want random spot checks. We know how to run better elections than we do. We don't because nobody wants to put the work out. I want us to put the work out.
  - danielweber 14y ago
    Unfortunately current US elections are unauditable because voting machines DO NOT produce any kind of verifiable voting trail Please stop. The machine in the video (at least when deployed in NC) absolutely has a paper receipt that gets printed every time a button is pressed.
    - scott_s 14y ago
      NY also has a paper trail. I voted this morning. I filled out a form, using a pen to fill in the circles for my votes, much like standardized tests. I then fed that form into a machine. The polling station kept the form, which is the paper trail for the electronic record.
    - 14y ago
briandear 14y ago
"It's a foregone conclusion".. ?? Because of Nate Silver's statistical analysis of polls? The polls which themselves are flawed. The "statistics" are based on nothing but polls, all with margins of errors and varied methodology. A sample of 1000 people can hardly predict an entire state, because there are many factors that influence turnout that aren't captured in polls. Being the subject of a poll is generally a passive activity -- they come to you, while going to vote is an active activity. "Likely voters" is a very subjective term and thus a difficult thing to quantify.
- genwin 14y ago
  > The "statistics" are based on nothing but polls, all with margins of errors and varied methodology. Sure, but if the methodology is standard and the poll shows 60%/40% with a 3% margin of error, you can safely bet money on the outcome. Statistics works just as well for casinos; they may lose in the long run, but they almost certainly won't. > "Likely voters" is a very subjective term and thus a difficult thing to quantify. Stats can be used to determine the likelihood that those being polled will actually vote.

by genwin 14y ago

danielweber 14y ago

And who regulates the exit pollsters? Normally I wouldn't worry about this, but if you want to give them power to invalidate an election (or force an "investigation," whatever that means), then you shouldn't just let them run around doing whatever they think is best. but the odds of him losing are nil, the stats show. I don't even know what mindset produces this thought. While Nate Silver doesn't think Romney will win, he gives changes that are a hell of a lot bigger than "nil."

genwin 14y ago
Exit pollsters should be regulated, I agree, if only to avoid an expensive unnecessary investigation. The media says that Obama and Romney are neck & neck, because that gets more viewers, which sells the advertisements they depend on. (Every presidential election in your lifetime is guaranteed to be declared a "virtual tie" by the media--you'll hear those exact words every 4 years.) Those who understand stats know (with almost certainty) that it'll be a landslide for Obama in the electoral college.
- twoodfin 14y ago
  How old are you? Within my lifetime there have been at least six presidential elections that were absolutely not reported as "virtual ties", including the last one.
  - genwin 14y ago
    Found in 5 seconds: "Gallup Daily: Obama-McCain Race Reverts to Virtual Tie" Found in another 5 seconds: "National Poll Shows Bush, Kerry in Virtual Tie" I could go on.
    - twoodfin 14y ago
      That Gallup poll is from June. The race was a virtual tie then. November reporting is a different story. And Bush/Kerry was also pretty damned close. Not one of the races I was referring to.
      - genwin 14y ago
        That's just the one I found in 5 seconds. This one's from mid-August 2008: "New [LA Times/Bloomberg] poll shows McCain - Obama in virtual tie". And from Sept: "[NBC News/Wall Street Journal] Poll shows Obama and McCain in virtual tie". 3 months later, still a virtual tie! I'm sure I could find a closer one. I didn't say the media would always report that in November. I went back to 1992 and still found "virtual tie" reports. That's 5 elections in a row.
btilly 14y ago
Exit polls are like a canary in a coal mine. They can show evidence of manipulations that might have happened. If an investigation finds that no such manipulation seems to have occurred, then no harm is done. Unfortunately current US elections are unauditable because voting machines DO NOT produce any kind of verifiable voting trail. Furthermore the apparent conclusion of the US media is that past discrepancies are evidence that their exit polling is flawed rather than that the voting process is being manipulated. (They are biased to decide that way because accusing the winners of manipulating the vote - even if it is true - pushes them away from the appearance of neutrality that they try to keep.) Furthermore you can't do exit polling on mail-in ballots. So this potential signal of flawed elections is being lost at the same time that manipulation of elections is becoming even easier. Personally if polls are overwhelming, and elections are in line with polls, I believe the result. But I believe that there are lots of people trying to slip a thumb on the scales, so if polls are close, I am dubious about the results. I personally believe that current evidence is that the Republicans are more successful in getting their thumbs on the scales, so I'm doubly suspicious when Republicans mysteriously do better than polls indicate they should. I want us to have elections that I trust. I do not want crappy election machines. I want verifiable paper trails. I want random spot checks. We know how to run better elections than we do. We don't because nobody wants to put the work out. I want us to put the work out.
- danielweber 14y ago
  Unfortunately current US elections are unauditable because voting machines DO NOT produce any kind of verifiable voting trail Please stop. The machine in the video (at least when deployed in NC) absolutely has a paper receipt that gets printed every time a button is pressed.
  - scott_s 14y ago
    NY also has a paper trail. I voted this morning. I filled out a form, using a pen to fill in the circles for my votes, much like standardized tests. I then fed that form into a machine. The polling station kept the form, which is the paper trail for the electronic record.
  - 14y ago

briandear 14y ago

"It's a foregone conclusion".. ?? Because of Nate Silver's statistical analysis of polls? The polls which themselves are flawed. The "statistics" are based on nothing but polls, all with margins of errors and varied methodology. A sample of 1000 people can hardly predict an entire state, because there are many factors that influence turnout that aren't captured in polls. Being the subject of a poll is generally a passive activity -- they come to you, while going to vote is an active activity. "Likely voters" is a very subjective term and thus a difficult thing to quantify.

genwin 14y ago
> The "statistics" are based on nothing but polls, all with margins of errors and varied methodology. Sure, but if the methodology is standard and the poll shows 60%/40% with a 3% margin of error, you can safely bet money on the outcome. Statistics works just as well for casinos; they may lose in the long run, but they almost certainly won't. > "Likely voters" is a very subjective term and thus a difficult thing to quantify. Stats can be used to determine the likelihood that those being polled will actually vote.