While many Kossites have been understandably getting upset and somewhat depressed by the results of the latest USAToday/CNN/Gallup poll, in fact this poll is far weaker substantively than it appears, and provides a very doubtful basis on which to assert that Bush has survived the Clarke crisis unscathed.
Go to the poll site on USA Today at
http://www.usatoday.com/news/politicselections/nation/polls/usatodaypolls.htm
and scroll down to the very bottom, where you read Results are based on telephone interviews with 1,001 National Adults, aged 18+, conducted March 26-28, 2004. For results based on the total sample of National Adults, one can say with 95% confidence that the margin of sampling error is ±3 percentage points.
In other words, the 3% margin of error relates to the entire sample of adults polled. Nowhere on the poll site does it say what percentage or how many of the total number polled were registered voters, and similarly what percentage or how many of the registered voters were likely voters. The likely voters -- the strongest measure of accuracy, historically -- are thus a subset of a subset of the total number polled. The margin of error for this could be much larger than for the whole sample. The total sample size is pretty standard for a national poll; generating news reports based on a much smaller subset of that sample without even saying how big it was is pretty close to the line of deceptive if not unethical reporting. Even the extensive Gallup discussion of the poll, stretching over four web pages, fails to state the size of the registered and likely voter subsets:
http://www.gallup.com/content/?ci=11167
It's more likely than not that this turned out to be an unrepresentative sample -- especially given that they don't disclose what the margin of error is on either registered or likely voters -- rather than indicating any swing to Bush.