Abstract
In presidential nomination contests, the number of voters participating in selecting the presidential candidates varies considerably across states. In the same election year, turnout in presidential nomination contests ranges from less than 1 % of party supporters participating in some caucuses to record breaking turnout levels upwards of 50 % in primaries in other states. This variation is attributable, in part, to the electoral rules, which vary across states, years, and parties. In this paper, I provide a comprehensive examination of the extent to which party and state rules affect voter turnout in nomination contests from 1980 to 2012. Using the normal partisan support score as the voter turnout denominator, I find that primaries, open contests, and proportional representation rules result in higher levels of turnout. I also show that within the window of competitiveness, turnout is higher in states that hold contests later in the nomination season. Overall, my analysis provides insight into the institutional structures that influence the number of people who participate in the presidential nomination process and enhances understanding about the factors that affect voter turnout.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
At times, semi-open rules are divided further into semi-open or semi-closed rules. Technically, under a semi-open rule, Independents and the party’s own members are allowed to participate and in a semi-closed state voters must either be registered members of the party or be willing to declare their past and/or future support for the party. However, many descriptions of the openness rule do not make this distinction, making it difficult to ensure that the data would be coded properly at this level. Consequently, for my purposes, semi-closed and semi-open rules are collapsed into one category labeled semi-open.
It is possible that voters may not be aware of the delegate allocation rules used by their party in their state, but the candidates should be aware of these rules and create campaign strategies to gain as many delegates as possible. There may be tighter, more intense campaigns, or campaigns from more candidates, in states that use proportional representation, rather than winner-takes-all. Candidate strategy may influence the delegate allocation rule indirectly, but the effects of the delegate allocation rule on voter turnout should still be present.
Some might think that turnout could also decrease in an open contest because more people would be eligible to participate, but they would be less interested in the outcome. Because of the way I operationalize voter turnout (which is discussed at length later on), the denominator stays constant regardless of whether the contest is open or closed. Thus, I theorize that opening up the contest should increase the turnout rate. Recent work by Hillygus and Treul (2014) also suggests that the rate of strategic voting increases in open contests, which typically leads to greater overall turnout.
The notable exception, of course, is the 2008 Democratic nomination.
Unfortunately, I cannot test the individual-level mechanisms at work because individual-level survey data that include respondents’ demographics, if respondents know the rules in their state, whether they received a campaign contact or mobilization efforts, and whether or not they voted does not exist, particularly in any comprehensive form over the course of multiple nomination cycles. By analyzing this relationship first at the aggregate level, I show that that rules affect voter turnout and then future work can explore the individual-level mechanisms underlying these relationships, if and when such data become available.
A bonus proportional rule indicates that delegates are awarded proportionally but the candidate who wins the state or district receives an extra delegate. This is sometimes referred to as winner-takes-more.
A loophole rule is winner-takes-all at the district level, rather than at the state level.
States that fall into this category may elect delegates directly, the delegates may be uncommitted, there may not be a formal system in place for translating voter preferences into delegates, or the states may use a complicated or uncommon rule. For instance, in 2008, the Republican Party in Alabama used a complicated rule to allocate delegates. If a candidate won a majority of votes statewide in Alabama, he received all 24 of the at-large delegates. If no candidate secured a majority of votes statewide then proportional representation with a 20 % threshold was used to allocate delegates. The district delegates also were awarded in a similarly convoluted fashion based on district vote results (Cook 2008). Thus, this category represents a catch-all, rather than a distinction with clear expectations.
I use the date of whichever of the following came first: a candidate secures a majority of delegates or a candidate becomes the de facto nominee. If I relied only on the date the candidate secured a majority of delegates, it would provide an inaccurate sense of competitiveness in some cases. For instance, in 2004, John Kerry was considered to be the presumptive nominee on March 4, the day after John Edwards withdrew from the race. Since Howard Dean, Wesley Clark, and Richard Gephardt all had withdrawn previously from the race, the end of Edwards’s campaign made Kerry the presumptive nominee. The race was no longer competitive because all of Kerry’s serious competitors had withdrawn, even though he did not secure a majority of delegates until April 27, 2004. However, in 2008, McCain secured the Republican nomination on March 4, when he captured a majority of the delegates to the National Convention, although Huckabee did not withdraw until the following day. Therefore, all Republican contests in 2008 held on or before March 4, 2008 are coded 1 for competitive, whereas all contests held after March 4, 2008 are coded as a 0 since the contests are no longer competitive. The dates used to create this variable can be found in Table 2 in the Appendix.
Norrander (2006) explores the dynamics in the winnowing process and finds that various candidate characteristics, institutional structures, and early outcomes affect candidates’ decisions to withdraw. As a result, the number of candidates in a race can vary dramatically over the nomination season. The variable number of candidates in the race accounts for this variation, as it measures the number of candidates in the race for each contest in the nomination season. In other words, this variable captures the candidate attrition that occurs as the nomination season progresses.
After discussing the findings of my analyses, I explain why it is not possible to measure these factors more directly and why the included variables are adequate.
The variable average state turnout is the state’s average turnout rate over the past two general presidential elections. For example, when analyzing North Dakota in 2008, I use the total raw vote in the 2004 and 2000 North Dakota presidential elections divided by the respective voting eligible population for each year. The average state turnout is the same for both parties.
Data for the Gini coefficient came from Dr. Mark Frank’s U.S. State Level Income Inequality website, which can be accessed at http://www.shsu.edu/~eco_mwf/inequality.html.
In her analysis of primary turnout, Norrander (1986) outlines four main criteria that any denominator for calculating voter turnout in nomination contests should meet: (1) the denominator should have different values for the Democratic and Republican contests because they are separate events; (2) the measure should allow the examination of the largest number of possible cases; (3) the denominator of the dependent variable should not be interrelated with any of the independent variables or bias will be introduced into the analysis; and (4) the dependent variable should be consistent across states and years.
To further demonstrate that this measure is a suitable choice, Norrander (1986) compares the empirical results from four different measures and finds that the various measures result in different, and sometimes contradictory, results. She concludes that some of the discrepancy in the results is based on the cases that can be analyzed using each measure.
The general election results data were collected from the Congressional Quarterly’s Guide to U.S. Elections series.
Since states elect senators and governors in different years, I use the past two elections in the creation and calculation of the NPSS, regardless of how recently they occurred. For instance, to calculate the Democratic NPSS for Minnesota in 2008, I average the percentages of the vote captured by the Democrats in the 2004 and 2000 presidential general elections, the 2006 and 2002 senate general elections, and the 2006 and 2002 gubernatorial general elections. For the 2008 Democratic NPSS in Kentucky, I average the percent of the vote captured by the Democrats in the 2004 and 2000 presidential general elections, the 2004 and 2002 senate general elections, and the 2007 and 2003 gubernatorial general elections.
The VEP was chosen, rather than by the VAP or the total registration figure, because the VEP reflects the number of people who could vote in the primary or caucus if they so chose. The VEP data for each state in each year was drawn from Dr. Michael McDonald’s election project website, which can be accessed at http://elections.gmu.edu/index.html.
Of course, in some years in some states, there may also be a competitive Independent or third-party candidate competing that garnered a relatively large share of the vote. While this measure does not incorporate that directly, if an Independent or third-party candidate receives a large portion of the vote, this suggests that the voters in the state may not be very supportive of the major parties and should not be expected to support their candidates routinely, which is captured in this measure. Additionally, by averaging the results of elections across races and years, the presence of an Independent or third party candidate is less problematic when creating the NPSS. Finally, the NPSS avoids problems caused by inaccurate voter registration lists.
The 1984 Republican, 1996 Democratic, 2004 Republican, and 2012 Democratic nomination contests are excluded from the analysis because the incumbent president was unopposed for renomination. Therefore, many states did not hold contests in these years. If contests were held, little attention was paid to them since there was no uncertainty about the outcome.
It is conceivable that the electoral rules are not exogenously assigned or changed. While this possibility exists, the evidence suggests that states are not primarily concerned with changing the rules to increase turnout. Instead, they are modifying the rules to advantage a certain candidate, to gain more attention form the media and candidates, or to abide by the national party rules.
Owing to the pooled dataset, I contend that the fixed effects model controlling for year is the most appropriate, so in the discussion that follows, I report those results. When the results between models differ considerably, I discuss both specifications. In order to ensure that my findings are robust to various model specifications, I also estimated a model clustering by year to account for the fact that observations within years are not independent and may violate the assumption of independent observations. The findings of this model provide further confidence in the robustness of my results, as they are very similar to the results of the fixed effects model and the pooled OLS model. Given that there are concerns about models with a small number of clusters, and the results indicate that the rules influence voter turnout in presidential nomination contests, the model is not reported. Additionally, I ran a random effects model, controlling for year, but the model degenerated to pooled OLS, likely as a result of the small number of panels (years) in the dataset. Furthermore, a hierarchical model is inappropriate for this analysis because of the limited number of years available to analyze in the post-reform era, and thus is not conducted.
As is discussed later on, additional work and investigations are needed to parse out whether the delegate allocation rule affects the voter turnout rate because voters are participating based on whether they think their vote matters or whether it depends on how candidates campaign based on the delegate allocation rule. Unfortunately, reliable indicators of candidate behavior and campaign activity across states are not available for every nomination in the post-reform era. These data may exist for some portions of the more recent nomination processes though, and future research could incorporate those measures for a subset of cases. Nevertheless, this finding is an important first step in that it demonstrates that the delegate allocation rule influences how many voters participate.
For the fixed effects model, the coefficient is 2.8 with a p value of 0.067, approaching significance; for the pooled OLS model, the coefficient is 3.11 with a p value of 0.040.
In order to ensure robustness, I also specified the models with other control variables, including variables capturing the number of days between the contest and the previous contest, the percentage of the state that has graduated high school, and a dummy variable indicating a southern state. All of these variables produce statistically insignificant coefficients, and the main findings of the models are statistically and substantively similar. I also controlled for lagged nomination turnout instead of average turnout in the general election, and the models with this alternate control produces consistent findings about the impact of the electoral rules on voter turnout.
The openness of the contest is not statistically significant in any of these models. However, this is explainable by the fact that this rule changes infrequently within a state, if at all. Thus, there is little variation in this variable once state fixed effects are taken into account. The coefficients on the other rule variables (days between Iowa caucuses and contest, proportional representation and primary) are substantively and statistically similar to previous models, providing additional confidence in my findings.
References
Abramson, P. R., Aldrich, J. H., & Rohde, D. W. (1998). Change and Continuity in the 1996 Elections. Washington: CQ Press.
Abramson, P. R., Aldrich, J. H., & Rohde, D. W. (2002). Change and Continuity in the 2000 Elections. Washington: Congressional Quarterly Press.
Abramson, P. R., Aldrich, J. H., & Rohde, D. W. (2006). Change and Continuity in the 2004 Elections. Washington: CQ Press.
Aldrich, J. H. (1980). Before the convention: Strategies and choices in presidential nomination campaigns. Chicago: The University of Chicago Press.
Amy, D. J. (1993). Real choices/new voices: The case for proportional representation elections in the United States. West Sussex: Columbia University Press.
Amy, D. J. (2000). Behind the ballot box: A citizen’s guide to voting systems. Westport: Praeger.
Balz, D. (2008). Feb. 5 primaries to pose a super test of strategy. The Washington Post.
Barber, B. R. (1984). Strong democracy. Participatory politics for a new age. Westport: Greenwood Press.
Beck, N. (2006). Time-series-cross-section methods. http://as.nyu.edu/docs/IO/2576/beck.pdf.
Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E. (1960). The American voter. Chicago: The University of Chicago Press.
Cook, R. (2004). The presidential nominating process. Lanham: Rowman and Littlefield.
Cook, R. (2008). Race for the presidency: Winning the 2008 nomination. Washington: CQ Press.
Corrado, A., & Gouvea, H. (2004). Financing presidential nominations under BCRA. In W. Mayer (Ed.), The making of the presidential candidates 2004. Lanham: Rowman and Littlefield.
DiClerico, R. E., & Davis, J. W. (2000). Choosing our choices: Debating the presidential nominating process. Lanham: Rowman and Littlefield.
Downs, A. (1957). An economic theory of democracy. New York: Harper and Row.
Geer, J. G. (1989). Nominating presidents: An evaluation of voters and primaries. Westport: Greenwood Press.
Green, D. P., Kim, S. Y., & Yoon, D. H. (2001). Dirty pool. International Organization, 55(2), 441–468.
Hillygus, D. S. & Treul, S. A. (2014). Assessing strategic voting in the 2008 US presidential primaries: The role of electoral context, institutional rules, and negative votes. Public Choice. doi:10.1007/s11127-014-0183-1.
Jewell, M. E. (1984). Northern state gubernatorial primary elections: Explaining voting turnout. American Politics Research, 12, 101–116.
Kamarck, E. C. (2009). Primary politics: How presidential candidates have shaped the modern nominating system. Washington: Brookings Institute Press.
Kenney, P. J. (1983). Explaining turnout in gubernatorial primaries. American Politics Research, 11, 315–326.
Kenney, P. J., & Rice, T. W. (1985). Voter turnout in presidential primaries: A cross-sectional analysis. Political Behavior, 7(1), 101–112.
Mayer, W. G., & Busch, A. E. (2004). The front-loading problem in presidential nominations. Washington: Brookings Institute Press.
Moran, J., & Fenster, M. (1982). Voter turnout in presidential primaries. American Politics Research, 10, 453–476.
Norrander, B. (1986). Measuring primary turnout in aggregate analysis. Political Behavior, 8(4), 356–373.
Norrander, B. (1992). Super Tuesday. Lexington: University Press of Kentucky.
Norrander, B. (2006). The attrition game: Initial resources, initial contests and the exit of candidates during the US presidential primary season. British Journal of Political Science, 36(3), 487–507.
Norrander, B. (2009). What political scientists may (or may not) know about presidential nominations. In J. Citrin & D. Karol (Eds.), Nominating the president: Evolution and revolution in 2008 and beyond chapter 5 (pp. 95–108). Lanham: Rowman and Littlefield.
Norrander, B., & Smith, G. W. (1985). Type of contest, candidate strategy, and turnout in presidential primaries. American Politics Research, 13, 28–50.
Pateman, C. (1970). Participation and democratic theory. Cambridge: Cambridge University Press.
Patterson, T. E. (2009). Voter participation: Records galore this time, but what about next time? In S. S. Smith & M. J. Springer (Eds.), Reforming the presidential nomination process (pp. 44–63). Washington: Brookings Institute Press.
Ranney, A. (1977). Participation in American presidential nominations, 1976. Washington: American Enterprise for Public Policy Research.
Salisbury, R. H. (1975). Research on political participation. American Journal of Political Science, 19(2), 323–341.
Schier, S. E. (1982). Turnout choice in presidential nominations: A case study. American Politics Research, 10, 231–245.
Wattenberg, M. P. (2002). Where have all the voters gone?. Cambridge: Harvard University Press.
Wolfinger, R. E., & Rosenstone, S. J. (1980). Who votes?. New Haven: Yale University Press.
Acknowledgments
I thank Sarah Treul, Paul Goren, Jason Roberts, Joanne Miller, Kathryn Pearson, Brian Southwell, Ben Ansell, and the anonymous referees for their valued comments on the article.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Rights and permissions
About this article
Cite this article
Jewitt, C.E. Packed primaries and empty caucuses: voter turnout in presidential nominations. Public Choice 160, 295–312 (2014). https://doi.org/10.1007/s11127-014-0185-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11127-014-0185-z