So just how much is a poll lead for the Iowa caucuses really worth at this point? That’s the big question after the latest CNN/Des Moines Register/Mediacom poll was released this weekend.
For those that weren’t deeply in the polling weeds, former Vice President Joe Biden led the field at 24%. Vermont Sen. Bernie Sanders, Massachusetts Sen. Elizabeth Warren and South Bend, Indiana Mayor Pete Buttigieg followed with 16%, 15% and 14% respectively. Rounding out the top five was California Sen. Kamala Harris at 7%; no other candidate polled above 2%.
A look at polls taken at about this point in primary cycles since 1980 suggests that while Biden leads now, there’s a pretty good chance he won’t actually win the caucuses.
Assigning an exact chance to how polling this early translates to the eventual results is difficult because of the high number of candidates running and the traditional volatility of Iowa caucuses’ polling. Only three eventual winners were polling at this point between 10% and 30% – which is where the top four 2020 contenders in Iowa currently sit. Most have been above 30% (seven out of 14) or below 10% (four out of 14).
Still, there’s a clear relationship between polling at this point and the eventual Iowa caucuses’ outcome. Using some simple modeling, we can capture some of this relationship to try and understand how frequently we should expect people polling at certain levels should go on to win. This model is meant as a first word, not the final word. It controls for where a candidate is polling and where the leader is polling, allowing us to make some rough assessments of what to expect.
What’s clear is there is no runaway favorite in this field. For candidates to have a 50% or greater chance of winning the caucuses at this point, my simple historical model suggests they should be polling at about 35% or greater. No candidate is close to that right now.
A frontrunner polling in Biden’s position (in the mid 20s) should be expected to win the caucuses about 30% of the time, according to said simple historical model This is even true when attempting to control for past candidates with high name recognition like Biden. That 30% chance lines up fairly well with how often candidates polling nationally in the mid 20s have gone on to win the nomination.
Put in non-statistical talk: Biden’s got the best shot of anyone to win the caucuses. But chances are someone else will win.
Those polling at about 15%, with a polling leader in the mid 20s, should be expected to win the caucuses about 15% of the time. This means that there is a meaningful distinction between leading at around 24% and trailing in the mid 10s. In other words, we should take the 2020 polling at face value: Biden’s probably in a bit of a better position than those below.
Still, we currently have three candidates polling at about 15% in Iowa. That means that, combined, there is probably a better chance that Buttigieg, Warren or Sanders will win the caucuses than Biden – but we just don’t know which one.
Someone in Harris’ position (a little north of 5%) should be expected to win Iowa around 5% to 10% of the time. Republican Ted Cruz was at about 5% in the polls at this point four years ago, and Rick Santorum was in the same position back in 2011. Harris definitely has a better-than-zero shot of winning, though it’s probably best to categorize her as a clear underdog at this point.
Perhaps most interesting is what past contests tell us about the group of candidates polling at 2% or lower. They suggest that someone polling at about 1% should go on to win the caucuses about 3% of the time. This isn’t nothing, especially since we have more candidates than ever polling around this level. The chance that any one of these candidates joins the top tier of candidates at some point in the coming months is almost certainly higher than 3%.
In fact, there is an example of a candidate polling at about 1% at this point in the cycle who has gone on to win the caucuses. Republican George H.W. Bush polled at 3% in May 1979 and 1% in August 1979 in the Des Moines Register poll. He would go on to shock the political world by winning the 1980 Iowa caucuses. (Of course, he didn’t win the GOP nomination that time around.)
Notably, all the examples of candidates polling at less than 10% at this point and going on to win the caucuses were Republican. But I’m not sure I would read much into that lack of Democrats coming from way back to win. We’re dealing with a sample size of just seven on each side of the political aisle. Plenty of second place finishers on the Democratic side were at 6% or below at this point, including Gary Hart in 1984, Paul Simon in 1988 and John Edwards in 2004.
The big question going forward is whether one of the candidates can actually start polling in the mid-30s in the near future. If none can, then this is a race where the frontrunner’s chance of winning Iowa will continue to be below that of the other candidates combined (i.e. the field).