November is still a long way off. I recall several similar polls in the months leading up to Bush's decisive victory over John Kerry.

You're surely right about it being a long time off 'til election day, but don't kid yourself that the GOP is in the same position now that it was in January 2004. I mean, it's not even close to the same as back then. Check out the ABC/Post poll from late January 2004
http://abcnews.go.com/images/pdf/945a1BushandSOTU.pdf

Bush and the Democrats were _tied_ in terms of "trust to handle the issues"; Bush's overall approval rating was 58/40 -- or a net +18%, as contrasted with the -14% it is now.

I am always curious about how those poll numbers are achieved, since I have never been asked for my opinion on any such poll, and I have yet to meet anyone who has. Is the ABC News/Washington Post poll really reflective of the overall attitude in America?

Lawyerchik...if a poll is able to draw a truly random sample, and its sample size is large enough, than to within a very small margin of error, yes the numbers are actually very believable.

It's a statistical question, really. Statistics tells us that if you take a large population and draw a truly random sample from it...as the size of that sample increases from 1-1000, the margin of error on that sample decreases dramatically, and beyond a sample of about 1000, the margin of error hardly decreases even for huge increases in sample size.

For a poll like this to not be representative...the only way to argue against it is to argue for some sort of bias in the poll methodology. Either its missing a significant fraction of people because of folks having only cell phones which aren't polled, or its missing a group of people because they have to work during the day, or the question is somehow leading people towards an answer (i.e. push-polling). These issues are a concern, but thus far they haven't proven to be a massive one. They may grow with time as cell phone usage (as the only phone line) increases, but at least thus far it seems that the groups which don't get sampled aren't yet large enough to fully bias the poll numbers.

So how do we know that the ABC/Wash Post polls "draw a truly random sample", with a large enough sample size"? I read the .pdf attachment that included all questions after 31 (apparently, 1-31 had been previously released, so they did not include a link to those questions or answers in the article), and that attachment only included the results.

I'm still skeptical about the validity of a survey that was conducted by telephone during the 2nd week before Christmas (Dec. 15-18, 2005), among a random national sample of 1,003 adults.

There are ways to make sure that a "random" sample is drawn...usually if a poll is taken, the pollster will take demographic information to make sure it is roughly representative. For example, if 25% of the U.S. population is African American...and you poll 1000 people and sample 100 African Americans...it's possible you have a sampling bias. However...this is not proof...in a small sample, such as a minority group, it's more likely that just by random chance you'll have a smaller group. Therefore, in many cases, you have to have multiple different things to look at before you can judge whether or not a poll is doing a good job at sampling randomly.

First, if you want to make sure the sample is random...look at the sample size. The smaller the sample...the more likely the sample is to be biased.

Beyond that, the sample should be at some level representative of the population it is sampling. If you expect the country to be roughly equal in numbers of Democrats and Republicans, but 45% of the respondents say they were 1 side and 25% say they're the other...then it's possible you've got a sampling bias. However, this is not firm...it may just be that after some big event (9/11) more people are temporarily putting htemselves in 1 party and you're observing a transient feature.

The third thing you can do to judge whether you have a biased sample is, finally, the history of the poll itself. Most of these polls have run for years, and they have a history which can be compared to a firm, large-scale number, like election results. Since that's usually what you want to examine with a poll anyway, that's a good place to look. So if a poll actually gets several elections right in a row...the odds are that they've figured out a way to get a good, random sample and ask it the correct questions. But if a poll misses consistently to 1 side or the other (i.e. the Fox News poll), then it's probable that they have a biased sample.

There's no easy answer here...and that's a product of the random variations. When you poll 1000 or so Americans, if you repeated the same poll 100 times, even with a truly random sample, you would get a variance around 2-3% in your poll...meaning if the true value was 42%, you would get a gaussian distribution with a maximum probability of seeing 42 but finite probabilities of seeing 40, 41, 43, 45, etc. as results of your poll. This is the margin of error reported. In many cases, if you really want to evaluate a poll, you just have to pay attention to enough polls to know who's doing things right historically, whether or not those polls follow a uniform trend, and what mistakes people have made in the past. There's no easy answer...but you can learn a lot from them if you know how to read them.

The only opinions that count in politics are the opinions of actual voters. Many people who blindly and mindlessly follow the ebb and flow of media hype, do not have the intellectual wherewithal to make a responsible political decision. Thankfully, these people often do not bother to show up at the polls. Besides, hypothetical candidates of polls have the luxury of not being real. Real candidates say and do things that sway voters more immediately than memories of past mistakes. Howard Dean's screaming speech in the Democratic Primary is a fine case in point. If he had won the primary and made that speech closer to November, no doubt it would have destroyed his bid for the presidency. The extreme leftists of the Democratic Party are gaining power. Fortunately, they have a tendency to say really stupid things.

So, if the unwashed masses (not political elites on the Post's hitlist) really want Democratic leadership, will someone step up and define what that is, exactly?

In other words, where's the substance?