The 2008 election witnessed many states pushing their presidential primary dates earlier than ever. By the end of February 2008, 38 states had already held caucuses and primaries, up from only two, Iowa and New Hampshire, that voted by March 1, 1976. Yet despite front-loading in 2008, the contest was especially protracted on the Democratic side. Given the fluid context of primaries and observed swings in national polls, many Democratic voters likely switched candidate support over the course of the 2008 primary campaign. In this article, we examine how perceptions of early state outcomes affected voter choice and candidate momentum. Early wins may have sent signals to voters about candidate viability, an important cue for some people. Although early primaries of 2008 left many voters with a relatively brief window to assess candidates, the calendar nonetheless allowed a non-front-runner candidate to benefit from momentum and win the Democratic nomination. But what are the dynamics of such a process?
Prominent observers of U.S. presidential nomination politics suggest that front-loading should advantage early front-runners and mute a non-front-runner's chance of building momentum from an early win (e.g., Polsby and Wildavsky 2008, 108; Wayne 2008, 121). At the outset of the 2008 Democratic contest Hillary Clinton led a crowded pre-Iowa field in opinion polls for over a year, and she had a 20 percentage point lead in national polls taken immediately prior to the 2008 Iowa caucus. (1) That soon transitioned into an exceptionally close contest between Clinton and Barack Obama. (2) Obama and Clinton both experienced important wins in early states, with their relative strength in national polls changing, at least in part, as voters switched their support from one candidate to another. How stable were voter preferences as these contests unfolded? Do voters form and change preferences in response to outcomes in earlier states? Although cross-sectional and aggregate data have helped answer these questions, there are, surprisingly, few panel surveys of primary voters. (3) We examine individual-level proclivity to change candidate preferences during a single primary season; and we assess how this is influenced by awareness of previous state outcomes, perceptions of viability, and shifting evaluations of candidates. This process illuminates the dynamics of momentum.
The 2008 presidential primary represents a particularly interesting case to observe momentum because the campaign itself started so early. In January 2007, almost all major candidates had announced their presidential campaign and started an aggressive outreach effort to make their case with voters. Indeed, by October 2007, still 13 months before the general election, the Democratic candidates had already participated in 10 nationally televised debates, and opinion polls suggested the public was more familiar with the candidates than in any previous open-seat year (Fox News Opinion Dynamics Poll 2007). Despite this prolonged period to "get to know" the candidates and high name familiarity for major candidates, substantial proportions of registered voters were undecided when asked to evaluate two of the three major Democratic candidates. In mid-October 2007, 39% of registered voter respondents in a national poll could not rate John Edwards as favorable or unfavorable, and 37% could not rate Barack Obama (CBS News Poll 2007). Hillary Clinton was able to maintain at least a 20-point lead over Obama and Edwards throughout the entirety of 2007. (4) Yet in the end she fell short, we argue, owing to the momentum that Obama generated throughout a string of victories that changed his perception with voters from merely likeable, to viable. (5)
Viability and Voting in Presidential Primaries
There are key differences in models explaining voting behavior in presidential primaries and general elections. General election models often include candidate qualities, ideology, issue preferences, and most importantly, partisan identification as explanatory variables in voter decision making (Stone, Rapoport, and Abramowitz, 1992). However, unlike general elections, voters in primaries lack clear partisan--and to a lesser extent ideological--cues in determining their vote; therefore decision making may be more complex. Moreover, information about candidate policy positions in primary elections is relatively rare, leaving voter decisions somewhat of an enigma, defaulting to name recognition and personal character traits (Polsby and Wildavsky 2008).
Studies attempting to explain voting behavior in presidential primaries have investigated myriad aspects of voter decision making. Aldrich and Alvarez (1994) showed that candidate issue-stances do matter to voters during presidential primary contests in terms of vote choice. Although Wattier (1983) suggested that ideology is an important explanatory variable, many studies demonstrate that voting behavior in primaries is determined primarily by candidate preference, not ideology (based on either policies or general likeability). In her study of the 1980 presidential primary elections, Norrander (1986) found that "candidate qualities" are the most consistent and frequent correlate of who people vote for. She concluded that electability (ability to win in November) played a role in vote choice but only for states holding primaries later in the nomination calendar. Other research indeed has shown that candidate traits are the most important in determining how primary voters vote (Gopoian 1982; Marshall 1984).
Additional studies have identified the role of viability (chances of winning the nomination) and electability (chances of winning the general) as important determinants in presidential primary voting (Abramowitz 1989). Some of the research posits that, in addition to the importance of candidate evaluation, voters are concerned with being on the "winning" side and that although viability does not directly influence candidate choice, it influences perceptions of electability, which in turn strongly influences vote choice (Abramowitz 1989). (6)
Other findings suggest that voters strongly consider viability in addition to candidate preference in casting votes. Indeed, it may be that voters and donors assess candidates in terms of expectations about their prospects for winning the nomination, their prospects for being elected in November, or both (e.g., Abramowitz 1989; Abramson et al. 1992; Mutz 1995). Cross-national literature also provides systematic evidence of "strategic" or "sophisticated" voting in many multiparty (multicandidate) choice settings (for a review,, see Cox 1997). For example, we have evidence that some voters may defect from their most preferred choice and vote for a lower-ranked option if they perceive their first option has little chance of winning (Blais and Nadeau 1996; Cain 1978; Karp et al. 2002). One causal mechanism driving this is voter response to information about a candidate's electoral prospects. This can come in the form of information about a party's historic strength in an electoral district, information about candidate standing in recent opinion polls, or other sources. (Bowler and Lanoue 1992; Johnston et al. 1992).
In nomination elections, voters also utilize information from early electoral events to adjust their voting intentions in response to changes in perceptions of viability (Abramson et al. 1992; Bartels 1985). Strategic voting associated with perceptions of viability is likely to be part of a broader phenomenon referred to as momentum--the process where candidates are advantaged because they are perceived to be leading or gaining ground. Scholars are divided as to what momentum "means"--whether it reflects learning or rational or irrational behavior (Bartels 1988; Brady and Johnston 1987; Kenney and Rice 1994; Mutz 1997). That said, we expect that one important way of learning about candidate viability is mass media attention to early election results (Popkin 1991).
These studies provide evidence that factors beyond initial candidate preference--notably viability--shape primary voting decisions. However, the findings by Abramowitz et al. (1992) are only a single cross-section. This makes it difficult to know whether voters shift preferences in response to actual caucus and primary election results, whether they gravitate toward a candidate regardless of awareness of the candidate's early electoral success, or whether they simply supported the candidate from the start. We need to know whom voters supported before the first caucus (Iowa) and then examine the same voters' preferences and perceptions after the early caucuses and primaries unfolded.
Stone, Rapoport, and Abramowitz's (1992) expected utility model is especially relevant to the current analysis as a way of explaining presidential primary voting behavior--especially momentum and vote switching. This model holds that voters combine both candidate preference and likelihood of winning (in the general election) as determinants of voting behavior. In their study of 1984 Iowa Democratic caucus
attendees and Iowa Democratic convention delegates, Stone, Rapoport, and Abramowitz, (1992) found that the interaction of candidate traits, qualities, and electability was the strongest predictor of candidate support. (7) This supports the expected utility model as an explanation for voting behavior in presidential primaries, which this article extends and tests in the 2008 primary season.
Stone, Rapoport, and Abramowitz (1992) review and test three contending models of candidate choice in the 1984 Democratic presidential nomination campaign in Iowa. The first model, a preference model, suggests that ideological, issue, or candidate preferences drive vote choice. The second proposes that voter preference is driven more by interpretations of the of a candidate's chance of success in winning nomination or general election campaigns. The...