But see, Teixeira:
<http://www.emergingdemocraticmajorityweblog.com/donkeyrising/archives/000579.php>
July 31, 2004 Before We Pronounce on the Bounce, Let's Measure It Properly Shall We?
Newsweek is first out of the box with a poll and story on Kerry's bounce from the convention. There's just one slight problem: the data they report in the story don't really measure Kerry's bounce at all.
Check this out. Newsweek's last poll before the convention was on July 8-9 and Kerry led Bush, 51-45. In the current poll, conducted July 29-30, Kerry leads Bush, 52-44. From this they conclude Kerry got a "baby bounce" from the convention (1 point on support level, 2 points on margin or, if you look at their Kerry-Bush-Nader data, 2 points on support level, 4 points on margin).
But, as their story sheepishly admits, half of their poll was conducted on Thursday night, before Kerry had delivered his acceptance speech! Moreover, their results differ on the two nights, with Kerry leading by 2 points in the pre-acceptance speech data and by 10 points in the post-acceptance speech data.
What possible excuse can there be for presenting these data as measuring Kerry's bounce from the convention, when the effect of the most important event of the convention isn't included in half the data? Perhaps there is one, but I can't think of it.
And that's not all that's wrong with their bounce measure. To make their sin even more egregious, the previous poll they use as a point of comparison is way too long ago (July 8-9) to be a real before/after comparison. What if the race was closer before the convention than it was on July 8-9? Then using July 8-9 as a point of comparison would further contribute to understating Kerry's bounce from the convention.
And in fact that appears to be the case. In the Gallup poll, Kerry was leading 51-44 on July 8-11 but only 49-45 on July 19-21. So using July 8-9 as the comparison period probably knocks several more points off Kerry's bounce.
In short, Newsweek's analysis is totally bogus. Before we pronounce on the bounce, I suggest we wait until we've got some data that actually measures it.