But it is more accurate. On Friday, Loras College released its own survey in Iowa -- and it shows Clinton up by 29 points.
It is not the case that Iowa voters went from preferring Clinton by five points on Wednesday to preferring Sanders by eight on Thursday to preferring Clinton again on Friday, this time by 29 points. One-third of Iowa Democrats did not change their minds overnight.
First of all, the Loras poll was actually completed before the CNN one, so, if anything, it went from Clinton up 29 to Sanders up eight during the two days after Loras was done, not the other way around. But it didn't do that either.
Polling since the start of the year in Iowa has been a bit confusing. We've plotted it below, showing the duration that the polls were "in the field," the industry term for when polls are actually being conducted.
It was clustered shortly after the new year, and now it's not.
So what's happening? Our polling guru, Scott Clement, points out that Loras's polls have shown Clinton above the polling average with some regularity recently. They're the light blue dots on this graph.
Does this mean that Loras is biased for Clinton? Probably not. It probably simply means that the way Loras targets its poll respondents, the system it uses to conduct its polls, means it ends up with more Clinton supporters. It's hard to say at a quick glance.
The important moral of this important story is that polling is subject to differences based on methodology. Both CNN and Loras can't be right, of course, and we don't know which is, or where in the middle (or above or below!) the truth lies. This is why more polling helps, and over the next 10 days, we can be confident there will be no shortage of polling.
Just remember: Big shifts aren't necessarily signs that the race is actually moving that much. We love to say "We told you so," but we'd rather not have to do it all the time.