People have been trying to get the media to stop reporting polls this way for years, without result. They just won't listen.

Thanks for the link.

It would certainly be useful if reporting on polls would follow the lead of scientific papers and, instead of deciding upon a "margin of error", simply tell us what the probability is that the lead shown in the polling is meaningful. Or the p-value that the suggested lead is simply random chance.

I've yet to see a poll with McCain above 43%, and most have him lower. Obama has led McCain in every poll I've seen. At some point, the accumulation of polls should lead to a greater confidence in their collective message, namely that Obama is ahead.

Have you ever taken a stat course?
The reason for the 'statistical tie' term is that Obama's lead is less than the margin of error. What this means is it shows Obama up by 3 points this time. If they took another sampling, it might show McCain up by 2...or it might show Obama up by one or down by one. Because of the possible errors, a slightly different result will be obtained each time the same population is sampled.

Have you ever learned to read, El Viajero?

Have you, Viajero? First of all, if the margin of error is 2.5% and Obama is up by 3%, then his lead isn't less than the margin of error. But even if it were, the various possible outcomes from additional polls are not equally likely. With this lead it's certainly possible that there's a tie (just as there would be with any other lead), but it's much more likely that Obama is ahead -- just not 95% likely.

Not to rain on everyone's parade, and I certainly wish that polls were better reported (any lead shows a >50% probability the leader is actually ahead), but... a 3% lead is within a 2.5% Margin. Obama could be down 2.5%, McCain could be up 2.5%, both of those would be within the first standard deviation of probability. And in that case, McCain is ahead.

At the very least, the spread between the two isn't what's polled (even if it's what we call care about), so you can't simply apply the MOE to the spread.

Yikes, Viajero, you're way off. You didn't click the link to Kevin Drum, did you? MOE means there's a 95% chance that each number is within 2.5 percentage points, in this case, of the reported figure. It does NOT mean it's equally likely that McCain is actually ahead by 2.

There's an 85% chance that Obama's 3 point lead means he's actually ahead, according to Kevin's table. Go ahead, click it, learn something.

(Shock Mouse, I'd recommend a click as well)

The biggest problem is that the sampling error is only the error that's easy to calculate. The error resulting from the fact that those surveyed aren't a perfectly random sample of the actual voters and that their answers don't perfectly correspond to their votes may be larger than the margin of error, but it's ignored.

I'm just personally annoyed that there's this meme out there that somehow John McCain had gained ground or that Barack Obama is slipping.

If you check out the RCP or Pollster numbers, you'll see that Obama has maintained a constant lead that has actually widened a bit over the last week. (And the one recent McCain +1 poll was Gallup/USA Today poll with an extremely problematic likely voter screen; Obama was still ahead among RVs.)

Now, looking ahead, ask yourself, who is going to get a bigger, better convention bounce: the charismatic speaker who rises to the occasion with an ascendant, energetic party or the cringeworthy awkward speaker with the dispirited, discredited party, whose convention will feature a prime-time speech by one of the most unpopular presidents of the last hundred years?

I don't think it's time to get smug or complacent, but the mainstream media/RNC Obama concern-trolls, the Clintonista I-told-you-so-ers, the predicable Kossack chicken-littlers are full of shit. Obama's doing well and will likely soon being doing even better.

KC, completely agree with you there, flawed methodology is way more important to polls than is MOE. Averaging polls hopefully wipes away some of that but it's hard to say just how much. Unfortunately, that's basically admitting they're more for entertainment value than anything else. Still, I'd much rather have Obama be consistently ahead, as he has been.

Yikes, Viajero, you're way off. You didn't click the link to Kevin Drum, did you?

Yikes! when you're wrong, you're wrong. I did not follow the link and thought the margin of error was 5%.

Busted! My bad. Owning up to my error.

And of course the likelihood of multiple polls with multiple methodologies all showing a lead for Obama with McCain stuck in the low 40s is what ?

I am actually with the critics on this.

Yes, in theory, the margin of error just means there's still a 95 percent chance that the poll is close to being correct.

But in practice, what KC said is right. There are all sorts of errors made in polling, and the margin of error simply measures the likelihood that a truly random sample does not accurately reflect opinion.

The truth is we rely way too much on polls, there are plenty of examples of election results that were not consistent with the polls, and we should be careful about reading anything more than general trend information out of poll results.

Good on ya for being open to correction, El V.

Wow- Props to Viajero. I think that's the first time in internet history that someone's admitted to being mistaken about something on a comments thread. I was honestly taken aback by how refreshing it was.

"The reason for the 'statistical tie' term is that Obama's lead is less than the margin of error. What this means is it shows Obama up by 3 points this time. If they took another sampling, it might show McCain up by 2...or it might show Obama up by one or down by one. Because of the possible errors, a slightly different result will be obtained each time the same population is sampled."

Except that Gallup runs a poll every day, and every day Obama is up between 3 and 6 points. You're correct for a _single_ sample, but we don't have single samples in the daily tracking polls. We actually have dozens of samples and the MOE vanishes when all of those samples reaffirm the same result.

it's funny how people keep repeating ideas- isn't the last post what I wrote above?

Let's all remmeber the 17 point lead Dukakis had about this time of the election year.

Things change. I wouldn't get too excited. Something in the air tells me it will be McCain. Telltale signs:

1)Unpopular war, troubled economy, etc., you would think the Democratic candidate would poll much higher at this time. He isn't.

2)Outspent 4:1, McCain still polls closely
3)The voters that will decide this election are not yet involved.
4)This issue with Russia could pull McCains numbers way up as voters take another look at the inexperience issue with Obama.

All I'm saying is it's not the shoe-in everone thought it would be and that gives me pause.

Then of course there is the larger problem which is that Presidents aren't elected by the national popular vote. Obama is in a very strong position by most tabulations of the electoral vote.

1988 isn't 2008 primarily because 1988 was the height of the GOP revolution, and 2008 is consigning the GOP to being a party of the South. No one thinks its a shoe in. Obama certainly isn't with his ground game (far superor to even Bush 2004) taking that risk. But to imply 2008 is similar to 1988 is missing the 20 years that have fundamentally reshaped the landscape. I would argue that even Dukakais would have a shoot in today's environment. Every Democrat would. That's how bad it is for the GOP.

The Dukakis 17 point lead was taken immediately after the successful 1988 Dem convention and before the GOP convention. The only thing the Dukakis 17 illustrates is the post-convention bounce.

It's apples and oranges. Time to put that silly meme to bed.

Also, these polls for sure are under representing young voters who do not have land lines (I'm assuming the polls are conducted by telephone) and are much more likely to support Obama.

Young people run their mouths, but they don't seem to make it to the polling place like the old do.