The Historical Accuracy of the Des Moines Register Poll in Iowa

Relevant Cx question: Will the last Des Moines Register poll accurately predict the winner of the Iowa caucuses?

After months of stump speeches, dozens of debates, and more front-runners than any other primary fight in recent memory, the contest for the Republican presidential nomination will finally get underway on Tuesday, January 3. Iowa caucus-goers across the state head to their local caucus site to choose their candidate (or, as Roger Simon sarcastically quipped, “ruin things by actually voting”). Iowans take their first-in-the-nation role pretty seriously, and the caucuses have in the past elevated lesser-known candidates and torn down supposedly inevitable nominees.

The most influential poll in the state comes from the Des Moines Register, Iowa’s largest newspaper. The DMR will release its last poll of the caucus season on Saturday, Dec. 31, at 7PM central time (but who’s counting?), and will be dissected by pundits and the press for any insights into what voters are thinking.

The unique nature of the caucus—in short, they’re local meetings in which voters can try to persuade others to switch their vote—makes forecasting the contest difficult. Organization on the ground is crucial for turning out the kind of supporters who can win over their friends and neighbors. Also, independents or Democrats can switch registration on election day to vote in the Republican primary. The most recent CNN/TIME/Opinion Research poll did not include independents or Democrats in their sample, something Nate Silver took notice of yesterday.

However, historically the DMR’s poll has been a fairly accurate predictor of who will win the Iowa caucuses. I went back and looked at the results of the final-week polls for the 2000, 2004, and 2008 elections to see how they compared with the actual outcomes. This 2007 graphic from Politico covers the 2000 and 2004 outcomes on both the Democratic and Republican sides (with the exception of 2004, when President Bush ran unchallenged). Click to see full-size:

Notice that the DMR correctly predicted the order that the candidates would finish in both years. If anything, these results show that the DMR tends to underestimate the amount of support for the front-runner, although that could be a function of uncommitted poll respondents flocking to the likely winner in the days leading up to the election.

The DMR nailed it in 2007, too, when almost no one was projecting that then-Senator Barack Obama would defeat Hillary Clinton. On the Republican side, they accurately picked up on the surge in support that was flocking to former Arkansas governor Mike Huckabee, as well. I whipped up some charts to show how the results and the poll numbers stack up from the last time around (again, click to enlarge). Legend is the same as above:

As you can see, the Des Moines Register poll picked up on the groundswell of support for the Obama campaign that swept him to victory in the Democratic caucuses in 2008. Again, as in 2000 and 2004, the respondents who were undecided going into the final week of the Iowa campaign look like they eventually settled on the front-runners.

Of course, these results are a small sample. I’m sure other polls have better track records of predicting specific races. But the DMR poll is the one that the national press corps will be waiting for on New Year’s Eve, and the one that’s talked about on the Sunday shows the next morning. The results will be something to keep an eye on in the final days of the Iowa campaign.

With all that in mind, here’s our question on the topic:
Will the last Des Moines Register poll accurately predict the winner of the Iowa caucuses?