According to twitter The Washington Post/ABC Poll and The Pew Poll both have Obama-50% Romney-47%.

Interesting. Why do they poll 6% more D's than R's?

 

VOR essentially answered this, but it's worth pointing out that the reason these polls are decently accurate is because of random sampling. In theory that means that any member of the population being sampled should have the same chance of being polled as any other member of the population. Obviously, that's fairly difficult, but for the most part pollsters are pretty good at it. For instance, a typical phone poll will use a computer program to randomly select numbers to call from a database of phone numbers. That keeps it random, but it also misses people who have unlisted numbers, cell phones, no phones, screen calls, etc. (And the big question with this is whether these people are much like the rest of the population and can be safely ignored or if there is something unique to them that would cause inaccuracies, e.g. younger people tend to vote a certain way and also tend not to have landlines and thus their absence from a poll would skew it.)

 

But, I said that to say this: you can't pick and choose who you want to sample because it wouldn't be random. A poll like the one woodsrider described below would be useless because its sample isn't randomly generated, it wouldn't be representative of the broader population. So, if a poll has more responses from Ds than Rs, it's a safe bet that there are more Ds than Rs in the population. That isn't to say they are perfect, but their inaccuracies probably stem from what I described above or something specific to each firm's methodology.

 

Unless you are over sampling one way or the other what does it matter which way you swing? If Rush was doing a poll and he asked 200 D's, 200 R's, and 100 I's who they were voting for and it end up Romney 51% Obama 49% does it matter that he is as far right as you can be?

Real Clear Politics average has it Obama 48.8%-48.1%.

Real Clear Politics average has it Obama 48.8%-48.1%.

 

RCP has a good track record of accuracy. But the national vote isn't where the focus should be. It is on those 6 or 7 key states that are still up for grabs. As I have said for more than a month, Romney very well may win the national vote and lose the election.

VOR essentially answered this, but it's worth pointing out that the reason these polls are decently accurate is because of random sampling. In theory that means that any member of the population being sampled should have the same chance of being polled as any other member of the population. Obviously, that's fairly difficult, but for the most part pollsters are pretty good at it. For instance, a typical phone poll will use a computer program to randomly select numbers to call from a database of phone numbers. That keeps it random, but it also misses people who have unlisted numbers, cell phones, no phones, screen calls, etc. (And the big question with this is whether these people are much like the rest of the population and can be safely ignored or if there is something unique to them that would cause inaccuracies, e.g. younger people tend to vote a certain way and also tend not to have landlines and thus their absence from a poll would skew it.)

 

But, I said that to say this: you can't pick and choose who you want to sample because it wouldn't be random. A poll like the one woodsrider described below would be useless because its sample isn't randomly generated, it wouldn't be representative of the broader population. So, if a poll has more responses from Ds than Rs, it's a safe bet that there are more Ds than Rs in the population. That isn't to say they are perfect, but their inaccuracies probably stem from what I described above or something specific to each firm's methodology.

Why wouldn't it be random? I didn't say those respondents were hand picked.

RCP has a good track record of accuracy. But the national vote isn't where the focus should be. It is on those 6 or 7 key states that are still up for grabs. As I have said for more than a month, Romney very well may win the national vote and lose the election.

RCP has the race a toss up right based on the electoral counts.

Doesn't look good Romney fans.... Vegas lines.

 

5Dimes has Obama at 77%, BetFair is at 78%. Pinnacle has taken their line down.

Why wouldn't it be random? I didn't say those respondents were hand picked.

 

It’s not random because you’re selecting (and excluding) people based on certain characteristics, in this case party registration, in order for your sample to match your preconceptions about the demographic distribution (a truly random sample would match these distributions without any help too, so you're poisoning your results). Let’s say you call the first 500 people and 300 of them are registered Democrats. You would have to throw out 100 of those answers and continue calling people (and ignoring Democrats) until you found 200 Republicans and 100 indies. That’s not random and wouldn't tell you much about the population.

RCP has a good track record of accuracy. But the national vote isn't where the focus should be. It is on those 6 or 7 key states that are still up for grabs. As I have said for more than a month, Romney very well may win the national vote and lose the election.

 

I agree with your assessment of the national vote, but at this point, that is the aspect of the race that most interests me. I think the electoral college is pretty well determined--at least Obama having a fairly secure 270. The popular vote is much closer at this point, IMO.

FiveThirtyEight now says the President has a 91% chance of winning.

FiveThirtyEight now says the President has a 91% chance of winning.

 

 

Isn't 538 a NY Times (read DNC) poll??

It’s not random because you’re selecting (and excluding) people based on certain characteristics, in this case party registration, in order for your sample to match your preconceptions about the demographic distribution (a truly random sample would match these distributions without any help too, so you're poisoning your results). Let’s say you call the first 500 people and 300 of them are registered Democrats. You would have to throw out 100 of those answers and continue calling people (and ignoring Democrats) until you found 200 Republicans and 100 indies. That’s not random and wouldn't tell you much about the population.

You're completely missing my point. Change the %s if you like my point still stands. 500 people were randomly called, now pick a % of each.

You're mssing his point. By cherry picking respondents based on a characteristic, in this case voter ID, it is no longer a random sample but instead weighted. Incidentally, what you're describing is part of the difference between a registered voter poll and a likely voter poll.

Again, cherry picking was was not part of my example. This isn't hard guys.

Isn't 538 a NY Times (read DNC) poll??

 

It isn't even a poll.