<$BlogRSDURL$>

Thursday, September 30, 2004

The Persuadables

As this was a news cycle lead yesterday, I thought I'd throw my $0.02 USD on it as it can be easilymisleadling.

First, if 20% of all voters are "persuadable". then AP/Knowledge Networks obtained 6645 respondants to arrive at is sample size of 1329.

What I find implausible is that these 1329 were all recontacted three weeks later. Something does not pass the sniff test. Maybe, just maybe, this explains the math of the 937 "remaining persuadables. These follow-up interviews demonstrated 24% of the initial 1329 had become committed to one or the other candidate (although 18% more committed to Bush! Sorry could not resist the confusion by percentages.) This totals 319 and would leave a reamining 1,010 who are persuadable. 1010 does not equal 937. Where did those 73 respondants go?

Of this 937, 22% are labeled "purely undecided", sub-sample n subscript pu. ( Maybe Dan Rather can offer me some help on subscripting in a blog?)

Sub-sample n subscript pu is 206. To take us back to the beginning, this means of the 6645 original respondants,206 or 3.1% are purely undecided. Does it make any difference what these 3.1% think when Bush is ahead 6-8% points? Even if they all broke for the challenger, Bush would still win.

To break down this 206 even further, as this poll does, into when asked who would handle Iraq best, 20% choose neither candidate. 20% of 206 is 41 (0.16% of original sample)! Can we really say anything about what 41 people will do? Other than saying they will probably never decide!

On a pure methodological basis, I wonder how reliable the operational defintion of "persuadbale is. If it is not very reliable, then this entire poll is GIGO.

Comments: Post a Comment

This page is powered by Blogger. Isn't yours?