Don’t let anyone tell you they know how the Iowa caucuses are going to go.
With just a few weeks left until Democratic voters take their first steps toward choosing a nominee for the 2020 presidential election, the focus is as intense as ever on every new poll that comes out of an early primary or caucus state.
This was made even more acute by the dramatic dearth of polling during December when pollsters avoided being in the field due to unusually situated holidays. When the polls returned, the results held fairly close to the pre-holiday consensus: a close four-way contest between former Vice President Joe Biden, Vermont Sen. Bernie Sanders, Massachusetts Sen. Elizabeth Warren, and former South Bend, Indiana, Mayor Pete Buttigieg.
Every new poll that has one candidate rising a point or two will be at serious risk of being overinterpreted. Make no mistake; this is a four-way race in Iowa at the moment. (Tuesday night’s debate in Des Moines seems unlikely to shake up that status.) Nationally, Biden is well out in front of the pack, but that has not translated to a similar lead in the Hawkeye State. With polling averages that place all four candidates’ share within a less than 5 percentage point range, anyone who feels confident in how this is going to go hasn’t looked at the history.
Polling in Iowa is notoriously tough. There’s a reason why Ann Selzer of the Des Moines Register poll is held in the highest esteem in polling circles: She works hard to navigate the ins and outs of the difficult polling challenge of gauging how Iowans might caucus, and does so in a state where you have to imagine most Democratic voters have stopped answering their constantly ringing phones altogether.
There’s also a spotty track record of the polls “calling” much of anything in Iowa.
There was, of course, the stunning victory of John Kerry in 2004, doubling the vote total of Howard Dean, despite polling averages showing them within the margin of error on caucus day. In 2008, on the Republican side, polls heading into caucus day showed Mike Huckabee and Mitt Romney within the margin of error—Huckabee took the state by 9 percentage points. For the Democrats in 2008, Hillary Clinton, John Edwards, and Barack Obama were locked in a close three-way race according to polling averages, but Obama wound up winning decisively.
In 2012, polling averages on caucus day had Romney in the lead with Ron Paul nipping at his heels and Rick Santorum trailing by 6.5 points. Santorum and Romney finished within a tenth of a point of one another.
The 2016 caucuses were not much better, with the final polling average showing Trump up nearly 5 points over his rivals, while in the end, it was Texas Sen. Ted Cruz, who won by 3.3 points. The Democratic side was a bit more on the mark, with final polling averages showing Clinton up by 4 points, and a final result of a razor-thin margin for Clinton over Sanders.
All of which is to say, it is hard to find an Iowa result over the last decade and a half that looked exactly like the polling averages. This doesn’t mean the polling is broken. There are a few cases of results within the margin of error that nonetheless only look like a “miss” if you think of polls as exact measures and don’t think probabilistically. Nonetheless, there are more than a few years where the polling consensus either missed late movement or was off simply off the mark.
Let this serve as a reminder that Iowa is a hard state to pin down. The unusual caucus rules and the intense focus on the state for months and months leading up to caucus day all can lead to an unexpected outcome.
