This post has been updated.
The nature of the last few days before presidential primary voting kicks off is that a thing you wrote at 2 p.m. can be out of date by 4 p.m. On early Thursday afternoon, I looked at the state of play in Iowa with 11 days to go. Later the same afternoon, CNN and its polling partner, ORC, dropped new Republican and Democratic polls in the state.
These new polls offer a noticeable shift from other recent surveys. Loras College surveyed Iowa from Jan. 13 to 18, putting Donald Trump and Ted Cruz at about even in the state. In CNN/ORC's poll, conducted from Jan. 15 to 20, Trump is up by 11. On the Democratic side, Hillary Clinton led Bernie Sanders by five points in the polling average. CNN/ORC has Sanders up by 8.
As we have noted regularly, late, big shifts in polling — and not necessarily predictive ones — are common in Iowa and in New Hampshire. (Eleven days prior to 2012, Ron Paul lead the GOP. In 2008, Clinton and Mike Huckabee led -- Clinton wound up in third.) But it's not clear that these are big shifts in polling.
FiveThirtyEight's Nate Silver notes that the CNN methodology is based on the possibility of more than 300,000 people turning out to caucus. In 2012, the number was 121,503. The highest recent turnout for a caucus was the 2008 Democratic contest, which saw 40 percent turnout. If 40 percent of Iowa Republicans turn out this year, that's 245,000 caucus-goers.
Assuming higher turnout benefits both Trump and Sanders. In the Des Moines Register/Bloomberg polling in Iowa earlier this month, Trump trailed Cruz by five points among those who definitely planned to caucus. Among those who might go, though, they were tied. More maybe-voters showing up to vote means more support for Donald Trump.
The pattern has been similar for Sanders, who sees more support from younger voters than Clinton. In this CNN poll (the sample size of which is small enough that there aren't a lot of demographic breakouts), Sanders leads Clinton by eight points overall — but Clinton leads by 29 with the substantial over-65 population. Young people, who are less likely to own homes and more likely to work odd hours, vote less regularly. If more people turn out to vote, that means more young people — which means Sanders does better.
Update: Our polling guru Scott Clement notes that the DMR/Bloomberg poll cited above had only a slightly smaller implied turnout -- but that it and the CNN/ORC polls may also be seeing a split in how voters are identified as we saw in New Hampshire last week. In this case, though, the effect is reversed.
So how do we look at this poll? One good way is to look at the averages of recent polls in the state, which can help smooth out some of the oddities and differences found in individual surveys. Here's how the Republican and Democratic contests look, according to Real Clear Politics. (This includes the new CNN polls.)
The races are close. Since Jan. 1, Cruz's lead has fallen or been erased; Clinton's has narrowed.
More polls will come out. If one tomorrow shows Clinton up by 10 points, that does not mean that Iowa voters moved 18 points in however-many days. It probably means some movement and some differences in sample or methodology.
Perhaps Iowa Republicans moved that 11 points toward Trump from what was measured by Loras on Jan. 18 to what CNN/ORC found by Jan. 20. We saw pretty big changes at the end of the 2008 and 2012 campaigns, after all.
Or maybe we should take a deep breath and look at the big picture, which is admittedly less fun.