When there is a three point plus or minus in a poll does that mean

[Bandit]

there is actually a 6% possible difference. Say Bush* leads at 51% to Kerry 46%. If the margin of error is 3% could that mean Kerry could actually have a 49% and Bush* a 48% for a swing of 6 percentage points or does it mean Kerry could have 47.5% and Bush 49.5%? :shrug:

[beyurslf]

1. It means the first. 3% MOE means it must be more than 6% different

to be outside of the MOE

[WMliberal]

2. yes. it's them admitting that their best guess

is still WAY off. 19 times out of 20.

[On the Road]

3. It Means That It Could be 3% Either Way

a 3% Bush lead in the polls could indicate a tied race or a 6% Bush lead.

Actually, it's not too likely that's it at one of the extremes. There's a 68% chance it's within half the margin of error. And the "margin of error" is itself an arbitrary point -- there's a one-out-of-twenty chance the numbers are outside the margin of error just based on a skewed sample.

The most important thing to realize is many factors are NOT captured in the so-called margin of error at all. Whether the people polled are representative of voters, whether people are honest to pollsters, whether the people answering the poll will show up on election day -- all these factors are actually more important than the margin of error.

[LifeDuringWartime]

4. 87.4% of statistics are made up

;)

[Bandit]

5. No the real number is 47.8% of all stats are made up on the spot

:shrug: :+

[warrens]

6. Well, if it is a statistically valid poll...

It could mean that Smirk has three more points than indicated and Kerry three less. or the opposite. Or anything in between. But nearly all polls are +- within a certain range, like 95% of the time in the past, it worked out this way. The other 5% it was totally off the mark.

But you have to read the methodology. Likely voters, registered voters, people of voting age all have different results, and different definitions. Likely voters can mean, you KNOW they voted last time, they TOLD you they voted last time, or they just said they were likely to vote.

Bias leaks in over and over.

And the real problem is, this is dialing for dollars. Bias leaks in by who's home when they call. Single people, who are more likely to be out, are underrepresented in most polls, married people overrepresented. And guess who happens to be more conservative.