Obama and McCain in "statistical tie?" Not so fast!

[wyldwolf]

From Political Wire:

An interesting debate broke out in the comments on our post reporting the latest Pew Research poll that showed the presidential race narrowing. While the pollster claimed the race was now even because Obama's 46% to 43% lead over McCain was within the survey's margin of error, this is very misleading.

The survey clearly shows the race narrowing, but suggesting it's a "statistical tie" or "deadlocked" is not accurate. As Kevin Drum noted, what we're really interested in is "the probability that the difference between the two candidates is greater than zero -- in other words, that one candidate is genuinely ahead of the other."

So instead of reporting that any result within the margin of error is a tie -- something we're often guilty of -- it's more informative to show how probable it is that a candidate is actually leading. A simple spreadsheet shows that in the case of the Pew Research poll, there is a 94% probability that Obama is actually leading McCain.

Thanks to the many Political Wire readers who took the time to make this point.

http://politicalwire.com/archives/2008/08/14/statistical_tie.html

[Hope And Change]

1. K & R.Thanks for posting!

[On the Road]

2. This Always Bugs Me --

the idea that there is some sort of magic line at 3%. Even if one candidate is 1% ahead in the polls, the poll still shows them in the lead. There's just more of a chance the lead is due to a nonrepresentative sample. On the other hand, a poll showing a lead outside the so-called margin of error is no guarantee -- there's just a small probability it's due to chance.

[dsc]

3. valid point but horrible example to use to make it

A 1% lead in a poll with a 3% MOE has only about a 1 in 4 chance of being real, vs a 4.5% lead in the same poll having about a 98% chance of being real. That is a pretty large difference.

[On the Road]

4. OK, But by Your Own Account

the candidate with a 1% has a greater probability of being ahead than the candidate 1% behind. It should still not be described as a tied race. And in daily polls like we have now, a consistent 1% lead becomes more significant over time.

[dsc]

5. you are conflating two different points

One is that a poll of polls is more accurate, that is very true. A 1 or 2 percent lead in a poll of polls or in a sustained group of polls is way more persuasive than one poll. I also conceed that a person 1% behind is less likely to be ahead than one 1% ahead though they are equally likely to be tied.

[rinsd]

6. Nice find WW

:hi: