Expert Commentary

How did the primary vote forecasts fare in 2008?

2009 study from DePaul University in Presidential Studies Quarterly on changing variables in forecasting U.S. presidential primaries.


Political scientists have been deeply interested in whether or not models can be constructed to predict presidential nomination contest outcomes, particularly using polls and other data points from the period before actual voting. Such a valid model might give credence to the idea that the “critical period of the campaign occurs prior to rather than during the primaries,” as DePaul University professor Wayne P. Steger writes. This would suggest the true power of the “invisible primary” period — and demonstrate that primary voters do not matter very much.

Steger’s 2009 study published in Presidential Studies Quarterly, “How Did the Primary Vote Forecasts Fare in 2008?” scrutinizes the methodology of leading political science prediction models and ultimately concludes that new dynamics — particularly with the rise of digital technology — are emerging that must be accounted for. The traditional models take data from late December a year or so before the general election, or the January of the election year; they attempt to demonstrate that, indeed, there is an informal but powerful pre-primary contest that winnows out candidates without sufficient media coverage, money, endorsements from party elites, and a baseline of public support as demonstrated in polling results.

The study’s findings include:

  • The variables outlined above — poll ratings, money, media coverage, elite endorsements — can forecast primary outcomes with some accuracy when there is a clear front-runner; however, they are less adept at predicting outcomes between candidates who are more equally matched.
  • Moreover, momentum is not sufficiently accounted for in the forecast models, and the “Iowa and New Hampshire contests can have massive effects on the selection of the eventual nominee.” The pre-primary polling, of course, misses the momentum a candidate — such as Gary Hart in 1984 or John McCain in 2008 — gains upon winning one of these early contests.
  • Models developed to explain elections from 1980 to 2000 are becoming outdated because of deep changes in the system. Indeed, the 2004 and 2008 presidential nominating processes “suggest that the underlying dynamics of presidential nominations campaigns that held sway during the 1980s and 1990s are giving way to something else.” While nominating rules play a role in outcomes, the biggest game-changer is the Internet, which allows operatives to generate grassroots support, organize supporters and raise funds from smaller donors.
  • In addition, the 2008 Democratic nomination contest, in which superdelegates played a decisive role, suggests that the role of party rules needs to be reexamined and perhaps elevated in importance for forecast models.

The study concludes that in this new era of Internet-powered political organizing, “front-loading does not necessarily undermine democratic choice in later primaries [and] it suggests a transfer of influence from party insiders and fund-raising networks to massive numbers of party activists.”

 Tags: elections, polls, presidential primary, Iowa/New Hampshire