|
Article Excerpt
This study compares the leading presidential primary vote forecasts for a common set of candidates and nomination campaigns through 2008. Primary vote forecast models are valuable diagnostic tools. They tell us something about what we know and do not know about the nomination process. Forecasts are based on theoretical arguments and empirical measures that generate an expectation for an outcome. If the outcome differs from the prediction, then the theory and/or the measures are wrong, or something intervened between the prediction and the event that changed the outcome. Forecasts thus help further our understanding of the factors and processes that yield presidential nominees.
To the extent that nominations can be forecast with data from before the caucuses and primaries, we can infer that the critical period of the campaign occurs prior to rather than during the primaries. (1) In this scenario, the caucuses and primaries play a plebiscitary role, confirming the results of the campaign that occurs before the primaries. If predictions using data from prior to the caucuses are inaccurate, while predictions incorporating the results of caucuses and primaries are accurate, then the caucuses and primaries would appear to have independence from the pre-primary campaign. If both are inaccurate, then the logical conclusion is that the theory underpinning the models is wrong, and we need to rethink presidential nominations.
The next section briefly reviews the leading presidential primary vote forecasts. Then I estimate the leading models for a common set of candidates and nomination campaigns from 1980 to 2008 to compare these models on an equal footing. The final section compares these pre-Iowa forecasts with "momentum" models that incorporate the results of the Iowa caucus and New Hampshire primary to predict the remaining primary vote.
Presidential Primary Vote Forecasts
Forecasting models implicitly assume that the critical period in the presidential nomination process occurs before the primaries (e.g., Cohen et al. 2008b; Hadley 1976; Steger 2000). Only candidates who emerge from the pre-primary campaign as viable options for primary voters have a realistic chance of winning the nomination. Viable candidates are those who gain media exposure, money, the backing of party insiders, and support in national polls before the caucuses and primaries begin (see, e.g., Butler 2004; Cohen et al. 2008b; Haynes et al. 2004; Hull 2008; Mayer 1996; Steger, Dowdle, and Adkins 2004). If the pre-primary competition for resources and support affects candidates' ability to compete for and win votes during the primaries, then models accounting for these effects should predict and at least partially explain the presidential primary vote.
Mayer (1996, 2003) found that candidates' support in the last national Gallup Poll taken before the Iowa caucus was a powerful predictor of the aggregate primary vote (APV) in contested presidential nominations from 1980 to 2000. The idea is that candidates' personal and programmatic appeal is integrated into aggregate-level, national polls of party identifiers and leaners--the people most likely to vote in the presidential primaries. Mayer (2008) refined his argument, noting that polls have their greatest predictive power when there is a clear leader in pre-primary national polls (one with at least 40% in polls while other rivals have less than 30%).
Adkins and Dowdle (2000, 2001, 2005) forecast the presidential primary vote in open nominations from 1980 to 2004. (2) Their results affirmed the importance of mass partisan support before the primaries, but also showed that the amount of money raised or spent is not as important as how much candidates have on hand when the primaries begin. (3) Steger (2000) and Steger, Dowdle, and Adkins (2004) found partisan differences, with candidates' cash reserves significantly affecting the Democratic primary vote, while pre-primary national Gallup Poll standings did not; they also found that candidates' Gallup Poll standings predicted the Republican primary vote, but cash reserves did not.
Steger (2007, 2008a, 2008b) and Cohen et al. (2003) found that endorsements by party elected officials were a significant predictor of the presidential primary vote for both political parties. Endorsements by party elites reflect an insider game, as opposed to a mass partisan effect, with candidate endorsements serving as cues to attentive publics about which candidates are viable and desirable. Such cues may be important when voters cannot use party labels to differentiate candidates and when prospective voters have little information about candidates' policy positions (Jamieson, Johnston, and Hagen 2000).
Steger (2008a) produced marginal improvements in the models through two modifications. First, he found that the predictive power of Gallup Polls and cash reserves varies with the competitiveness of the pre-primary nomination campaign rather than by party, as he had argued previously. Gallup Polls predict well in noncompetitive campaigns in which there is a front-runner before the Iowa caucus, while cash reserves help predict winners in competitive races without a clear front-runner before the Iowa caucus. Second, he found that more liberal Democratic and more conservative Republican candidates received relatively larger shares of the primary vote than would be expected, controlling for candidates' prior elite and mass partisan support and cash reserves (see Steger 2008a for elaboration on these points).
Post-New Hampshire Momentum Models
Forecasting the presidential primary vote accurately is difficult in large part because the primary vote is distributed across time, with the results of early caucuses and primaries affecting the vote in subsequent primaries (Bartels 1988). Candidates who gain momentum by beating expectations for the vote share in early caucuses and primaries receive a surge in media coverage, fund-raising, and support in subsequent primaries (Bartels 1988), while others drop out as they fall behind in the delegate count or run out of money (Norrander 2006). (4) Forecasts based on factors from before the primaries cannot account for the effects of positive or negative momentum during the caucuses and primaries, resulting in under- and overpredictions of vote shares for candidates who beat or lag expectations.
Adkins and Dowdle (2001), Steger, Dowdle, and Adkins (2004), and Steger (2007) developed forecast models that incorporate the Iowa caucus and New Hampshire primary results to predict the remaining primary vote. These models use the variables of the pre-Iowa forecasts and various measures of the results of the Iowa and New Hampshire contests to predict the post-New Hampshire primary vote. Including the results of these contests corrects for under- and overpredictions of the pre-caucus forecasts for candidates who gain or lose momentum, respectively. See the appendix...
|
