This is from ABC's The Note:

This from the Des Moines Register Poll is potentially really key for Iowa and perhaps beyond:

"Twenty-seven percent of Iowa adults surveyed said they had already voted. Kerry leads Bush, 52 percent to 41 percent, among that group of early-bird voters. Among the 73 percent who said they definitely would vote on Tuesday, Kerry and Bush are tied."

Iowa is the Upper Midwest state I've felt most confident about, but if this survey is anywhere near accurate, it's bad news. With twenty-seven percent of the votes cast, anything like an 11-point lead will be hard to make up.

UPDATE: Reader James Sherk comments:

You noted that Bush is 11 points behind among early voters in the Des Moines register poll.  This is true, but ignores the much larger sampling error for the subsample of early voters than for the sample as a whole.  So Bush might not be as far behind as this survey puts him.

Here is the article with most of the polls internals: http://desmoinesregister.com/apps/pbcs.dll/article?AID=/20041030/NEWS09/41030008/1001

Consider: We have an overall sample size of 806 likely voters.  27 percent of them have already voted.  So this means the sample size of early voters is 218 respondents.  The formula for determining the margin of error for a poll is plus or minus 1 / (square root of n). Plugging n=218 into this equation gives us a margin of error for this sample gives a margin of error of plus or minus 6.7 percent.

Also, this poll is an outlier.  Every other poll from Iowa has had Bush up, or Kerry up by less than this poll does.  So it is likely to already oversample Democratic voters.

This isn't to say that Kerry isn't leading Bush among early voters. In 2000, Gore beat Bush by 10,000 votes among early and absentee voters.  But given that this poll is the one most favorable to Kerry, and the much higher margin of error for the early voter subsample, it is probably a safe bet that Kerry isn't beating Bush by 11 points among early voters.

OH, NEVER MIND: Reader John Wixted is even less impressed:

Reader James Sherk got the margin of error story half right. The margin of error of 6.7% applies to a particular value, not to the difference between 2 particular values. First, let's consider a particular value. If I said that the president's approval rating was 50% with a margin of error of 6.7%, then you could be reasonably confident that his approval rating was somewhere between 43.3% and 56.7% (i.e., 50% plus of minus 6.7%). The approval rating is a particular value, not a difference between two values.

But (and this is the critical point) if that same poll shows Kerry with a lead of 11%, then we are suddenly talking about a difference score (i.e., the difference between the particular poll value for Bush and the particular poll value for Kerry). But the margin of error applies to specific poll values only, not to the difference between them. The margin of error for the difference between two values (which is what we really want to know) is approximately double the margin of error for a specific poll value. To see why (without going through the calculations), imagine that a poll shows Bush and Kerry tied at 50% each with a margin of error of 3.1%. This means that Bush might have 53% support and Kerry might have 47% support, a difference (53% - 47% = 6%) that is double the margin of error (even though both specific values are within the margin).

In any case, the Des Moines Register Poll shows that among those who have voted already, Kerry has an 11-point lead plus or minus about 12 points!

A number of readers what to know what happened to the missing 7% who have already voted. I assume they wouldn't say whom they voted for; but in that case, why were they included in the poll?

As I said, never mind.

Posted by John at 01:27 PM  |    |  

Trackback Pings

TrackBack URL for this entry:
http://www.powerlineblog.com/mt/mt-diespammers.cgi/8456

Listed below are links to weblogs that reference Bad news from Iowa:

» Broken Streaks from Truth, Lies & Common Sense
For the past 15 elections the incumbent's party has retained the Whitehouse whenever the Washington Redskins have won their final home game prior to the election. [Read More]

Tracked on October 31, 2004 03:44 PM

» phentermine from phentermine
In your free time, check out the sites about online poker texas holdem phentermine [Read More]

Tracked on February 12, 2005 12:03 AM