The Impact of Front-Loading on Presidential Primary Turnout in

Front-Loading versus Back-Loading: The Impact of FrontLoading on Presidential Primary Turnout in Michigan and
Indiana*
By
Marie-Catherine Wavreille
PhD Candidate
Centre d’étude de la vie politique (Cevipol)
ULB, Brussels
Avenue Jeanne 44, CP124
1050 Brussels
E-mail: [email protected]
Phone: +32 (0)2/650 46 56 (office)
* Paper prepared for the 7th ECPR General Conference, 4-7 September, 2013. Hosted by
Sciences Po Bordeaux.
Front-Loading versus Back-Loading: The Impact of Front-Loading on Presidential
Primary Turnout in Michigan and Indiana
Abstract:
Scientific debates about front-loading tend to focus on a variety of consequences of the
sequential nomination calendar on presidential nomination campaigns such as compressing
voters’ effective decision-making period, influencing the types of presidential hopefuls who
are ultimately nominated, giving advantage to the front-runner, making the nomination phase
of the campaign more media centered, and settling the nomination races in the early spring
(Mayer and Busch, 2004; Atkeson and Maestas, 2008). Because studies of presidential
primary elections have not focused profoundly on the impact of the front-loaded calendar on
turnout, the purpose of this paper is to add one more layer to the understanding of
participation in presidential primaries by investigating whether, and to what extent, the timing
and scheduling of delegate-selection events affect electoral participation.
Employing data from presidential primary elections in Michigan and Indiana from 1992 to
2012, we find evidence of a partial, yet effective, impact of front-loading in shaping primary
turnout. The research shows that only Republican partisans in Michigan and Democrats in
Indiana seem to have been positively affected by the increasingly front-loaded calendar. In
addition to the influence exerted by the nomination calendar, voters in Michigan were
mobilized to cast ballots in the primaries by aspects of the campaign (competition, campaign
spending, number of candidates, candidates activities, and the presence of additional races on
the ballot) and by legal structures (open primaries). What our comparison reveals is that frontloading did not have a depressive impact on voting rates in Indiana late-scheduled contests.
Instead participation in Hoosier primaries was influenced by both contextual (the organization
of state-level races, candidate spending and candidates activities) and legal factors (open
primary).
In the end, this paper strives to provide the reader with a comprehensive understanding of
how the amount of electoral participation varies as the scheduling of primary contests in
Michigan and Indiana changes.
2
I. Introduction
Repeatedly identified as highly compressed, the American nominating process has
substantially been altered notably with the emergence of front-loading. That is, more and
more states schedule their primary and caucus elections near the beginning of the delegateselection season (Mayer and Busch, 2004). Recent research suggests that primary
participation rarely average more than a third of the entire electorate (Cook 2004).
Electoral scholars have found the fast-paced calendar of primaries to transform presidential
nomination politics in various ways. As it happened, in the 2000 campaign, the contest for the
nomination on both sides was declared for all practical purposes five weeks after the primary
campaign began (see Table 1). The short time frame of the nomination season was not a
coincidence. Rather it was the result of concerted efforts by some states to front-load the
primary calendar in order to exert influence over the outcome (Smith and Springer, 2009).
Indeed, as the primary calendar has become more front-loaded with more and more states
jockeying for earlier and earlier spots on the primary schedule, it has become routine for
contenders to clinch their party’s nomination in the early spring (Mayer and Busch, 2004). In
addition, only voters in the first few states are guaranteed a meaningful voice in selecting the
major party nominees (Cook, 2004). Lastly, the fact that nominations are wrapped up early
has a depressive effect on participation with contests near the back of the calendar drawing a
lower number of participants (Atkeson and Maestas, 2008; Patterson, 2002). As the primary
electoral calendar has become more compressed in recent elections, it becomes increasingly
important to understand turnout in these primary contests.
Despite the importance of primary front-loading, little is known about its effect in shaping
primary turnout. This paper, therefore, addresses the following question: To what extent does
the timing of a state’s nomination event (i.e. how front-loaded it is) have an impact on
participation? Put differently, does moving a primary closer to New Hampshire’s first-in-thenation primary influence the number of votes cast in? Taking the literature on political
participation in presidential primaries as a point of departure, two main hypotheses are the
focus of this article. The first hypothesis combines the literature on timing and participation
and explores the impact of the electoral calendar on turnout patterns among primary voters.
According to the electoral calendar hypothesis a positive relationship exists between casting
ballot in early contests and participation in politics. The conclusion of the nomination battle
early in the campaign may explain low turnout levels among back-loaded voters. The second
hypothesis, on the other hand, focuses on patterns in partisan elections. Recent research
3
suggests that voting patterns differ dramatically between Democratic and Republican
primaries (Buell 2004). The partisan hypothesis suggests that front-loading has a different
impact on the Democratic and Republican levels of participation. Turnout, throughout this
article, is defined as the percentage of the partisan voting-age population who votes in
primary contests.
The article is structured as follows. After reviewing the current literature on the rational
choice model of voting, we provide a detailed outlook of the relationship between
participation and sequential nomination processes in the next section. Next, we introduce our
dataset in the following section before presenting and discussing our empirical analysis. The
analysis consists of two parts: first, participation in front-loaded contests is examined. Second,
we zoom in primary elections scheduled at the end of the campaign. Both parts of the
respective analyses emphasize the partial, yet effective importance of the electoral calendar.
The conclusion summarizes our findings.
II. Theoretical Framework: Rational Choice and Turnout in Primary Elections
A vast body of literature offers a wide range of theories on the determinants of political
participation (e.g., Lane 1959; Milbrath 1965; Verba and Nie 1972; Barnes and Kaase 1979;
Conway 1985; Parry, Moyser and Day 1992; Norris 2002). These models can be grouped in
three categories, corresponding to three levels of analysis: micro, meso, and macro (Norris
2002, Teorell 2006). Given that our paper is a comparative analysis with only two cases, this
section will mostly develop explanatory models at the micro level. More specifically, we look
at voter participation from one particular perspective, that of rational choice. The rational
choice model emphasizes that citizens decide to cast ballots if, in their own view, the benefits
stemming directly from their own involvement are greater than the costs (Downs, 1957). The
rational choice model has been often used in past studies to explain primary turnout (Morris
and Davis, 1975; Aldrich, 1980; Moran and Fenster, 1982; Schier, 1982). The following
pages have two objectives that I hope begin to develop answers to whether sequence matters
for the level of primary participation. The first objective is to provide a body of systematic
information about voter turnout in elections and as well as, more specifically, in primaries.
The second objective is to develop the link between the sequential nature of the presidential
nominating system and participation.
4
2.1.
Rational Choice and Electoral Participation
The choice to turn on to vote or to stay home on election day is an individual decision (Jewell
and Morehouse 2001:121). With chances of casting a decisive ballot being quite small, most
voting participants would be expected to abstain rather than vote. Yet the reality is different
with citizens expressing their preferences in large numbers. This leads many observers to
contend that the rational choice model should be rejected (Blais, 2000). Rather than
discarding the model, Blais argues that it offers a partial and incomplete explanation of
electoral participation (2000:11). In an earlier analysis, Rothenberg and Brody (1988) claim
that: “even the most broadly conceptualized expected utility models have been inadequate for
explaining turnout. The electorate appears susceptible to influences beyond the costs and
benefits of voting” (1988:255). Noting that rational choice theories face a particular difficulty
with voting, Aldrich’s study aims at solving such puzzle by presenting a richer rational choice
theoretical accounting of why people vote or abstain (1993:246). In addition, analyzing the
famous “paradox of voting” and the fact that a relatively large share of the electorate votes
even though the probability that one person’s one lonely act decides the election is
vanishingly small, Blais argues that the cost-benefit calculus only and truly affects the
decision to vote or to abstain of those whose sense of duty is weaker (2000:113). Rosenstone
and Hansen argue that citizen involvement in politics cannot be exclusively explained with
individual motives (2003:22). Instead, they reason, “the explanation of participation, to make
any sense, must move beyond the worlds of individuals to include family, friends, neighbors,
and co-workers, plus politicians, parties, activists, and interest groups” (2003:23). Through
their mobilization efforts, the costs of political participation are reduced and it hence helps to
overcome the “paradox of voting”.
2.2.
Rational Choice and Primary Turnout
Explaining primary voting levels has been widely studied in the scientific literature (e.g.
Kenney and Rice, 1985; Norrander and Smith, 1985; Norrander 1986b). Yet, primary turnout
is not truly understood. Plus, most studies were conducted preceding the advent of frontloading in the late 1980s (Ranney, 1977; Moran and Fenster, 1982; Norrander and Smith,
1985; Norrander, 1986c, 1992). Unlike general elections, primaries are staggered over a sixmonth period. And the primary season draws a far smaller percentage of voters than the
November general election. Voter turnout in contested delegate-selection events constitutes
less than 30 percent of the voting-age population (Davis 1980:134). Figure 1 presents turnout
in presidential primaries starting in 1968 when reforms introduced, although inadvertently, a
5
largely primary-based presidential nomination system. With the chief exception of 2008, the
figure reveals an almost linear descent since 30.9 percent voted in the 1972 presidential
primaries.
Figure 1: Presidential Primary Turnout, 1968-2012
Source: GANS Curtis, “2008 Primary Turnout Falls Just Short of Record Nationally, Breaks Record in Most
States”, Committee for the Study of the American Electorate, May 19, 2008. Updated by the author.
Note: Here the number of citizens eligible to vote is used as denominator to calculate turnout rates.
* 1968 turnout was based only seven states which held primaries in both parties that year.
** 2012 turnout was based on 22 states which held primaries in both parties.
2.3.
Rational Choice and Sequential Nominating Campaign
Rational choice voting models do not traditionally consider the sequential character of the
presidential nominating process. In this context, however, the expected utility of a citizen’s
vote at the time when the voter casts his or her ballot determines the decision to participate or
to stay at home on election day (Atkeson and Maestas 2008:4). This value of a vote depends
greatly on whether the ballot cast is likely to be decisive, that is, whether the chances of an
election being determined by one single vote are large. Casting a pivotal ballot, in a sequential
race, varies according to the position of voting participants in the electoral calendar. Hence
outcomes early in the sequential nomination season determine the value of votes cast later in
the campaign. Battaglini, Morton, and Palfrey (2005) claim that “more voters will turn out
early in the presidential process than late because the utility a voter receives from going early
is higher relative to the costs than a voter who goes late in the process” (Battaglini et al in
Atkeson, 2010:51). Earlier voting participants have also increased incentives to cast ballots
6
because of their capacity to message citizens in late-scheduled contests about both the
viability and electability of presidential hopefuls. Accordingly, recent research finds voter
participation rates to differ significantly over the extended nomination season. Norrander
notes that primary events scheduled in the early stages of the nomination campaign
experience higher voting rates as residents of states holding these early contests are exposed
to vigorous campaign and have an opportunity to select the nominees (2010:44). Once media
and the press have declared the eventual nominees incentives for participation fall greatly. In
the 2000 campaign, turnout for primary contests held on or before March 7 averaged 23
percent compared with 14 percent for later contests (Curtis in Buell and Mayer, 2004).
Voters in early contests are exposed to active and intense mobilization campaigns by
candidate organizations. These campaigns result in smaller costs for the voters to obtain and
gather information about the contenders and the election. In addition, voters are reminded of
the forthcoming electoral contest through news coverage from national and local media which
also advise them about the contenders. As the campaign unfolds, the amount of candidate
mobilization efforts declines. With front-loading causing White House contenders to exit the
race sooner (see infra), this creates a concentration of mobilization efforts in the few earlyvoting states but a fading or even absence of such efforts in the later states. In fact, Mayer and
Busch find New Hampshire and Iowa to receive a larger share of candidate attention in
comparison with other states (2004:25). As fewer aspirants vigorously campaign and as the
eventual winner becomes more evident, the incentives for voting participants decrease
resulting in greater participation costs.
Presidential nominations are decided through sequential elections. Regarding the effect of the
placement and relative positioning of particular state primaries within the months-long
nomination process on the benefits of voting, contradictory results are found in the literature.
Norrander’s work emphasizes that participation levels are apparently influenced by the
calendar of presidential primaries (1996:886). A study by Moran and Fenster, which
explained variations in primary voting levels within individual states, finds out that states
scheduling their delegate-selection event earlier in the season may stimulate participation
(1982:469). Norrander and Smith (1985) highlight that such influence may be indirect, and
mediated by the candidates’ tactics. Because the results of early-voting states receive extra
news media coverage and essentially determine the fates of many of the White House
aspirants, residents of those states “could calculate that his or her vote would have a greater
probability of producing the benefit of nominating a preferred candidate” (Norrander and
7
Smith, 1985:32) and thus cast ballots in higher numbers. Such effect exercised by the
scheduling of primary contests on turnout, they add, may be greater for the party not
controlling the presidency since contenders in these races are required to perform well in
order to avoid dropping out of the race (Norrander and Smith, 1985:33). Yet other authors
disagree with these conclusions (Ranney, 1977; Moran and Fenster, 1982; Schier, 1982). In a
study of the 1976 Democratic nomination process, Schier (1982) finds participation rates to
be higher at later caucuses.
As mentioned in the introduction front-loading has induced an array of phenomena, the first
one being the early withdrawal of contenders. In contrast, in the pre-reform period,
presidential aspirants remained in the field until the end of the primary campaign and usually
until the summer convention (Hagen and Mayer, 2000:32). With the advent of front-loading
one sees the emergence of a quite different pattern with White House aspirants withdrawing
more quickly. Mayer and Busch highlight that the driving factor behind these early
withdrawals is a lack of money rather than votes (2004:44). Thus as the effective number of
candidates declines as each primary passes, voters have fewer incentives to cast ballots
(Atkeson and Maestas, 2008). As the nominating season progresses, candidates who fared
poorly in the early contests exit the race for the pursuit of the Oval Office leaving fewer
choices for subsequent primary voters.
A second phenomenon associated with the emergence of front-loading starting in 1988 and
the resulting compression of more and more primaries into a brief period near the beginning
of the primary season, is the fact that nominating campaigns end today earlier with the races
being decided more quickly. Depicting front-loading as a massive transformation in the
electoral environment (2009:61), Atkeson and Maestas argue that the early conclusion of the
competitive stage of the primary with the identities of the two parties’ nominees being known
early in the primary schedule is probably the most significant form of influence of frontloading on voting (2009). This attribution of presidential nomination candidates early during
the season means that many party supporters face meaningless contests. Hence, with an
increasingly front-loaded nominating calendar and with delegates accumulating more quickly,
the eventual nominee is known before the last primary is conducted. This contrasts with the
1976 elections when both major parties had contesting races that lasted until the final day of
the primaries.
8
Table 1: Front-Loading and Stepping Down From the Race, 1992-2012
Presidential Elections
1992
1996
2000
2004
2008
2012
When Last Serious Office-Seeking Aspirant Exited the
White House Race
Democrats
Republicans
March 19 (Paul Tsongas)
August 17 (Pat Buchanan)
No office seeker
March 14 (Steve Forbes)
March 9 (Bill Bradley)
March 9 (John McCain)
March 3 (John Edwards)
No office seeker
June 7 (Hillary Clinton)
March 4 (Mike Huckabee)
No office seeker
April 10 (Rick Santorum)
Source: WATTIER Mark J., “Presidential Primaries and Frontloading: An Empirical Polemic”, Paper prepared
for “State of the Party: 2004 and Beyond,” a conference sponsored by the Ray C. Bliss Institute of Applied
Politics, University of Akron, Akron, OH, October 5-7, 2005. Updated by the author.
Table 1 present the dates when all but one of the serious office-seeking aspirants have left the
race. As we can see, the races are usually over in early March, a few weeks after the
beginning of the delegate-selection calendar in Iowa and New Hampshire. As the primary
season has become more front-loaded, it has become routine for nomination races to be settled
in the early spring. Office-seeking candidates exit persistently the presidential nomination
battle when they have little reasonable probability to catch up to the front-runner. The exit of
the last serious contenders confirms the conclusion of the nomination contest.
If timing matters for turnout, this is likely that voters casting ballot in early primaries turn out
to vote in higher numbers than in states holding their primary events after one of the
contenders has clinched the nomination. That nominating events scheduled before the major
party nominees are determined experience high percentage of voters is the main argument of
the present article.
III.Data and Methods
To test our hypothesis that turnout is reasonably higher when a state holds its contest prior to
the date when the winner of the nomination has been declared, we must track state voting
results from one nominating season to the next. This permits an examination of how primary
voting in the same state changes in back-to-back races. To get at this issue we use data from
1992 to 20121 nomination races on both major parties in the states of Michigan and Indiana.
The data is from six presidential election cycles, considering only primary elections. The
dataset is particularly well-suited for the present purpose, even though data are not available
for a series of contests since the state of Michigan had organized caucuses on several
occasions and no elections were held in the same state in 2004. Here we specify, successively,
the motives for having selected these two states in our comparative analysis. First, because
9
party primaries are administered by state governments, each state independently sets the date
for those electoral contests. There is thus a remarkable amount of variation in when states
hold their party primaries. Including in our analysis both primaries scheduled early on the
season and at the end of the calendar will allow us to examine the impact of the calendar on
voter turnout. Michigan, on the one hand, has joined in the front-loading bandwagon by
moving its primary events to earlier dates in the hope of increasing its sway on the nominating
process. As the national parties have not assigned the state an early place on the nomination
calendar, it grabbed an early date of its own in 2008, scheduling its primary on January 15,
which is barely one week after the New Hampshire contest (Cook, 2008). In contrast, Indiana
has until now resisted the urge toward front-loading its presidential primary in spite of the
wholesale movement of many other states to schedule their contest to early dates. As Table 4
shows, from 1992 to 2012, the Hoosier State has continuously held its primary in the
beginning of May. Moreover both Michigan and Indiana do not have party registration as
most states do, so citizens may participate in the primary of either party. We refer to these
states as holding open primary races. Jewell and Morehouse (2001:103) categorize Michigan
as completely open and Indiana as open (2001:103). It is likely that participation rates should
be higher in states holding open primaries than in closed primaries. The lower the costs of
voting, the higher supposedly is participation, and open primaries involve lower participation
costs because nonparty identifiers can readily participate in a primary without having to
register and face the additional obstacle of public declaration of party preference (Schier,
1982). Yet mixed results are found in past empirical studies regarding such effect. Research
on open and closed gubernatorial primaries finds that states holding open primaries can expect
significantly higher turnout (Kenney, 1983; Jewell and Olsen, 1982; and Jewell, 1977). Yet
Ranney (1977) found the opposite for presidential primary contests. Lastly, both Michigan
and Indiana held their first primary in 1916, although no primary was held between 1928 and
1972 in Michigan. Rubin (1980, 1981) shows that having a long primary history influences
turnout. In these states citizens should be more familiar with the nominating process and
consequently more likely to cast a ballot. Support for such interpretation can be found in
Kenney and Rice’s study. They found participation to be 8 percent higher in jurisdictions with
primaries implemented prior to 1968 (1985). Again contradictory results exist with Geer
(1989) finding the history of primary use to have no impact on primary voting levels.
Our concern in this research is to test the role of the date of the primary race on the amount of
political participation. The first requirement to undertake this task is to secure data on front10
loading. For space and time restrictions in what follows we discuss predominantly whether
the sequential nomination calendar influences the level of voter participation. Studying
multiple explanatory factors is beyond the scope of this paper. The independent variable
hypothetically influencing primary turnout is operationalized in the following way. Drawing
upon Mayer and Busch’s (2004) measure of front-loading, we assess this phenomenon by
examining the cumulative percentage of delegates selected by the end of each week in the
primary season. This measure is, according to the political scientists, “the single best measure
of front-loading available” (2004:269). In Tables 2 and 3 we present these percentages for the
period between 1992 and 2012. By the end of the eight week of the nomination campaign, 56
percent of Democratic delegates were awarded in 1992. In 2012, the half of delegates was
selected by the end of the 18th week, a sign of the recent step backward in terms of frontloading.
Table 2: Cumulative Percentage of Democratic Delegates Selected by the End of Each
Week during the Presidential Primary Season, 1992-2012
Week
1992
1996
2000
2004
2008
2012
1
1
1
0.8
1
0.85
1
1
1
1
9
3
0.85
2
9
26
1
14.4
4.6
2.4
3
31
46
1
17
8
2.4
4
41
60
1
17.8
59.5
4.8
5
43
74
47
56.9
66.2
4.8
6
43
74
65
73
68.8
4.8
7
56
74
71
78.6
68.8
4.8
8
56
74
71
78.6
81.4
22.9
9
56
81
79
78.6
82.5
25.6
10
62
81
79
78.6
82.5
32.6
11
68
87
79
78.6
82.5
32.6
12
70
89
79
78.6
82.5
39.3
13
72
91
86
83.8
82.5
39.3
14
75
92
88
86.2
82.5
39.3
15
100
100
89
88.3
87.8
58.6
16
92
93.1
87.8
58.6
17
92
93.1
94
66.2
18
100
95.6
95.1
68.2
19
100
98.7
71.3
20
98.7
78.3
21
100
100
22
Source: MAYER William G. and BUSCH Andrew E., The Front-loading problem in Presidential Nominations,
Washington, D.C., Brookings Institution Press, 2004, p. 6. Updated by the author for the three most recent
electoral cycles.
11
Table 3: Cumulative Percentage of Republican Delegates Selected by the End of Each
Week during the Primary Season, 1992-2012
Week
1992
1996
2000
2004
2008
2012
1
2
1
2.2
0.6
0.68
1
2
9
2
9.1
3.4
2.1
2
12
28
4
13
3.4
2.1
3
36
51
9
15.8
6.4
4.9
4
44
65
13
15.8
54.5
4.9
5
46
77
49
47.7
60.6
4.9
6
46
77
68
60.5
64.8
4.9
7
56
77
72
65.6
64.8
8.3
8
56
77
72
65.6
78.5
28.1
9
56
81
78
65.6
80.5
33.2
10
60
81
78
65.6
80.5
39.7
11
66
88
78
65.6
80.5
39.7
12
67
89
78
65.6
80.5
45.3
13
71
91
84
70.9
80.5
45.3
14
75
93
86
74.7
80.5
45.3
15
99
100
87
79.3
84.3
58.4
16
100
92
87.1
84.3
58.4
17
92
89.3
90.8
65.8
18
100
92.7
91.4
67.4
19
100
95.3
72
20
97
80.8
21
100
97.7
22
97.7
23
97.7
24
100
25
Source: MAYER William G. and BUSCH Andrew E., The Front-loading problem in Presidential Nominations,
Washington, D.C., Brookings Institution Press, 2004, p. 7. Updated by the author for the three most recent
electoral cycles.
The second requirement to undertake this task is to secure data about voter turnout in
primaries. The dependent variable in this study, presidential primary turnout, can be measured
by different ways. Calculating participation rates in states with closed primaries is relatively
simple since voters are required to register with a party if they want to express their
preferences for America’s highest office. Geer (1989) uses data on party registration. This
technique permits to treat each party’s primary as a distinct contest. However, one of the
disadvantages of such method is that it excludes open primaries from the analysis. Indeed, in
these open contests, any registered voter can cast a ballot in either Republican or Democratic
primary elections, and, consequently grouping voters by party is not possible (Jewell and
Morehouse, 2002:122). Given the open character of primaries organizing in Michigan and
Indiana, this measure based on party registration is of limited value for our research. In a
study, Norrander (1986b) mentions four different indicators commonly used to calculate
primary voting levels. A first technique divides the total number of voters in the Democratic
and Republican primaries in a state by a state’s total registered voters or by the voting-age
population (Key, 1956; Ranney, 1972; Jewell, 1977; Jewell and Olson, 1982; Kenney, 1983;
12
Kenney and Rice, 1985). This method provides one participation figure for each state. But,
given our objective to differentiate participation by party, we abstain here from using this
method. The three other ways for calculating primary turnout mentioned in Norrander’s work
produce different indicators for each major party’s primary. All three techniques share the
same numerator, that is, the number of votes cast in each major party’s primary contest, but
differ in the denominator they employ. The partisan enrollment measure, a second method
outlined by Norrander, is confined to closed primaries, as Geer’s method (Morris and Davis,
1975; Ranney, 1977; Moran and Fenster, 1982). A third technique, what Norrander calls the
subsequent general election measure, is obtained after dividing the electorate into partisan
groups based on their vote in the later November general election (Norrander, 1986:361)
(Overacker, 1926; Zeidenstein, 1970). This indicator, again, is not suitable for open primary
contests. Norrander recommends a forth method, called the normal vote measure. It employs
the normal vote, thus, “the expected value for the size of a party’s primary electorate if
everyone votes and no factor makes one party’s primary more or less attractive than the
other’s” (Norrander, 1986b:362) (Jewell, 1984a, 1984b; Norrander and Smith, 1985; Schier,
1982). In addition to these five measures mentioned supra, employing raw vote totals allows
the researcher to directly see whether participation levels have increased or not (Buell and
Mayer, 2004:208). Plus, it is argued, such technique facilitates comparison of voter
participation rates between the two major parties’ primaries.
As the above discussion indicates, calculating participation rates in open presidential
primaries is problematic. Due to either peculiarities of our data set or unavailable data, we
have decided not to apply any of the above-mentioned techniques (with the exception of raw
vote totals) but to opt for our own calculations. The dependent variable in our analysis is,
thus, the turnout of citizens by party – that is, the number of votes cast in each party’s
presidential primary divided by the partisan voting-age population. Following Jewell and
Morehouse (2002) and Norrander (2010), we divide the voting-age population in each state
between voters with Republican and Democratic leanings using indicators of party
identification provided by the annual Gallup Polls. This produces two partisan voting-age
populations, one for the Democratic Party and one for the GOP.
As regards to the numerator, as indicated above, we use the number of votes cast in each
party’s presidential primary from which we subtract the cross-over votes. Acting strategically,
Burden and Jones argue, presumes that voters favor a more viable or more electable contender
than their preferred candidate (2009:54). But voters can even be more strategic in some
13
primary elections. In these contests, that are open primaries, cross-over voting is allowed, and
not only party members but also independents and supporters of the other major party may
enter the opposing party’s primary (here, cross-over voting) and support a weaker competitor
to face their first choice in the fall general election (here, raiding) (Chen and Yang 2002). On
election day, citizens in both Michigan and Indiana choose in which party primary they wish
to cast ballots and partisans may cross party lines. As Chen and Yang put it, when primary
races are open to all voters, cross-over voting and raiding happen inevitably (2002:2).
Although not a majority, the number of voters who behave strategically cannot be determined
precisely on the basis of available data. Hence, we subtract, from the number of votes cast, the
share of the votes which emerges from cross-over voters. Including or not independents in the
definition of cross-over voting is a matter of academic dispute. For matters of consistency
(between closed and open primaries), we side with Wekkin when he raises the following
argument: “Why should independent identifiers who vote in an open primary be counted as
cross-over voters when independents who vote in a “closed” primary are not” (1988:106)?
Consequently, in our conceptualization of cross-over voting we exclude votes in either major
party’s primary by independent identifiers. Estimates of the amount of voters who participate
in the primary of the party they do not identify with in any given party’s primary election vary
somewhat. In a study of primary elections from 1980 through 1996, Alvarez and Nagler
exclude independents and find that the cross-over rate ranges from 4.2 percent in GOP
primaries and 5.3 percent in Democratic primaries in the 12 states examined with open
primaries (1997:11). Using data from the 1992 Voter Research and Surveys, the percentage of
cross-over voting they find in Michigan amounts to 4.1 percent in the Democratic primary and
3.7 percent in the Republican primary. The cross-over rate as we have defined it in Indiana
ranges from 3.6 percent in the Republican primary, and 2.5 percent in the Democratic primary
(Alvarez and Nagler, 1997:16). In spite of using not recent survey data, Alvarez and Nagler’s
study seems the only identifiable research with specific and accurate figures about overall
cross-over voting in the two states we examine here. Hence, in this report, we use these
estimates.
14
Figure 2: Presidential Primary Turnout in Michigan and Indiana, 1992-2012 (in
percent)
Figure 2 shows the levels of voter participation by party in Michigan and Indiana. We also
present, in Table 4 and 5, the raw vote totals. They provide a direct indicator of whether
voting participation has actually risen or fallen. When looking at GOP turnout in Michigan, a
clear spike emerges in 2000 when a little over 37 percent of citizens turned out to vote. In the
subsequent cycles, turnout declined and averaged 32 percent in 2008 and 33 percent in 2012.
Voters in Democratic primaries, on the other hand, had a decidedly lower turnout. The spike
shows that Indiana Democrats, which held its 2008 primary on May 6, had a little over 54
percent voter turnout - the highest turnout in our study. At this point, Clinton still had the
opportunity to win the nomination, suggesting that voters were motivated to participate
because it was still possible for them to influence the outcome. Then, when looking at
Republican contests in Indiana, a modest increase appears. Republicans had a 23 percent
turnout in 1992 as opposed to slightly more than a 26 percent turnout in 1996. After a small
flat section, turnout began to rise in 2008 when almost 22 percent participated but did not
reach the level in the Democratic primary.
IV.Results
We divide our analysis into two parts. In the first part below, we examine voting patterns in
Michigan in order to shed light on the electoral calendar hypothesis which supposes a positive
15
impact of the front-loaded contests on citizen involvement. In the second part of the analysis,
we examine, using an identical structure, participation in Indiana.
1. MOVING TO THE FRONT AND VOTER PARTICIPATION IN MICHIGAN
In the following pages, we analyze participation rates in Michigan. We first look at our first
hypothesis which contends that turnout is expected to be relatively higher in states holding
primary contests before the major party’s nominee is known. Simultaneously, we turn our
attention to Republican and Democratic contests distinctively in order to address the second
hypothesis about the differentiated effect of the electoral calendar on turnout in Republican
and Democratic nomination events. In a second step, nuances are brought by looking at other
explicative factors. And we finally conclude.
a. GOP Nominating Races: Voter Participation Jumped Starting in 2000
What is interesting for the purpose of this article is the sharp increase in turnout in 2000.
Voting level peaks at about 38 percent. It then begins to slightly fall but remains above the 30
percent. This voting pattern calls for an explanation. By the time the campaign rolled into
Michigan in 1992 more than a third of the delegates were already selected (see Table 3). Four
years later, this proportion even rose to half of the delegates. In contrast, during the most
recent electoral cycles, no more than 5 percent of the delegates were awarded by the time
voters had a chance to nominate the party’s nominee. Since then, citizens have a louder voice
in determining the party nominee as opposed to their position in previous cycles. All this
suggests that moving the primary to earlier dates has positively influenced primary
participation in GOP contests.
Table 3: Participation in Presidential Primaries in Michigan, 1992-2012 (raw totals)
March 17, 1992
March 19, 1996
Feb 22, 2000
2004***
 of Votes Cast in
Republican Primary
 of Votes Cast in
Democratic Primary
Voting-Age
Population (VAP)
449,133
585,972
6,947,000
(36%)
(31%)
524,161*
No Primary
(51%)
(/)
1,276,770**
No Primary
(4%)
(/)
No Primary
No Primary
7,177,000
7,358,000
7,541,000
16
Jan 15, 2008
Feb 28, 2012
(/)
(/)
869,169
594,398
(0.6%)
(1%)
996,499****
No Primary
(4.9%)
(/)
7,613,000
7,539,572
NB: In brackets, are the cumulative percentages of Republican and Democratic delegates selected by the end of
the week preceding the primary contest in Michigan.
*In 1996, 124,267 votes were cast for Democratic candidates in the GOP presidential primaries. Because those
voters are presumably Democratic supporters we do not include them in our calculations of GOP primary
turnout.
** In 2000, 44,850 votes were cast for Democratic candidates in the GOP presidential primaries. Because those
voters are presumably Democratic partisans we do not include them in our calculations of Republican primary
turnout.
*** In 2004, Republicans chose not to have either a caucus or a primary, since they were renominating George
W. Bush. The Democrats held a caucus on February 7.
**** Similarly, in 2012, 194,887 votes were cast for Democratic candidates in the GOP presidential primaries.
As for the 2000, we do not include them in our calculations of Republican primary turnout.
Source: Federal Electoral Commission; Michigan Department of State.
Admittedly, these conclusions are rushed. Other determinants outside of the primary calendar
did influence participation in Michigan presidential primaries. Hence, for instance, our
finding that participation was not dramatically low in the 1996 Republican contest (18.22
percent) in spite of the fact that the GOP nomination battle was virtually determined by the
time the campaign rolled into the Great Lakes State (see Table 1). First, participation in
Michigan GOP contests was influenced by legal determinants, and more specifically by crossover voting. In 2000, for instance, cross-over voters were so disproportionately numerous
when a significant number of independents and Democratic identifiers entered the GOP
primary and supported John McCain (Cook, 2004:74). According to exit polls, 17 percent of
voters in that year’s Republican primary self-identified as Democrats. But this was not the
first occasion when this strategic behavior was witnessed in the state. In fact, the Wolverine
State has a long history of voters mischievously crossing-over in the state’s open contests.
Mark Brewer, the most experienced MDP Chairman for decades, reasons: “In the 1970s,
Republicans crossed-over into our primary and voted for George Wallace” (quoted in
Michigan Radio, 2012) who won easily despite the opposition of Democratic leaders, and “the
same thing happened in 1988 when Jesse Jackson won the state” (quoted in The Washington
Post, February 22, 2012). And voters crossing party lines also occurred in 2012 with President
Obama running unchallenged on the Democratic ballot. According to exit polls, Democrats
accounted for 9 percent of the Republican primary electorate in the state. That number
matches up closer with the 2008 and 1996 figures, years when Democrats were not a major
characteristic in the GOP primary electorate. “Democrats this year weren’t feeling the same
17
warmth for Santorum and his deeply conservative and religious message” (quoted in Real
Clear Politics, 2012) observes Mark Grebner of Practical Political Consulting. As a
Washington Post reporter maintains, all of this “points to a cross-over effect that was certainly
real, but wasn’t all that exceptional” (quoted in The Washington Post, 2012).
Equally important in explaining the variations in voting rates are campaign features. That was
evident in the 2000 campaign when more than 1.2 million voters cast a ballot, a record for a
Republican presidential primary. “The only higher primary turnout for president”, Cook
remarks, “came on the Democratic side in 1972, when nearly 1.6 million voters took part”
(2008:74-75). Although moving the primary to the front and consequently before a significant
share of the delegates has been awarded might have increased turnout, an open nominating
event (without an incumbent running) and spirited competition were key determinants in
producing the strongest primary turnout. Competition draws, in general, more voters to the
polls (see Kenney and Rice, 1985; Geer, 1989). Hence with early-scheduled primaries being
more contested with greater campaign activity and campaign spending, this creates incentives
for citizens to get involved. In 2000, both leading candidates, Bush and McCain, concentrated
heavily on the Wolverine State whose February 22 primary was the first large showdown of
the campaign. Indeed, Michigan GOP primary carried “enormous stakes for both contenders,
but particularly for McCain, who need[ed] the victory to overcome the establishment support
and financial resources that Bush ha[d] assembled” (quoted in Los Angeles Times, 2000). In
short, Rosenstone and Hansen argue that citizen participation is fostered when campaigns for
America’s highest office are strongly challenged (2003:188).
Moreover scholars have argued that whether the state is home to one of the presidential
aspirants affect electoral participation (Atkeson and Maestas, 2008). The 2012 primary
campaign in Michigan was an interested case in that respect. With Romney having closed ties
to the state, the Michigan contest was particularly central for him and it became, according to
political analysts, a “must win” for the former Massachusetts Governor whose father served
as Michigan Governor in the 1960s. In the hopes of securing a win in his boyhood home state,
Romney campaigned massively and his campaign spent some $3.8 million on TV ads in the
Great Lakes State, compared to about $2.2 million for Santorum. But these campaign efforts
did not foresee the fact that Michigan Democrats crossed-over in 2012 and voted for
Santorum in an effort to prolong the GOP nomination campaign. Michigan Democrats’
animosity toward the former Massachusetts Governor and a native of Michigan was not just
18
tactical. “There also is a strong streak of lingering resentment over Romney’s refusal to
support the 2008 auto-industry bailout” (quoted in The Daily Beast, 2012) notes a reporter.
It can already be said that the front-loaded primary events have come to play an increasing
role in influencing participation in Republican nomination contests. But this is hardly the
lonely factors shaping turnout. Cross-over voting, competition, candidate activities and
spending, and whether the state is home to one of the presidential contenders are important
determinants.
b. Democratic Presidential Contests: Few Primaries Held
Primary elections in a given state being two distinct contests, they may be held in completely
different environments. And Republicans and Democrats may be influenced by different
structures and should be expected to react accordingly. Compared with GOP races, a rather
different picture emerges from Democratic primaries in terms of turnout patterns. In fact, less
than one in five citizens went to the polls in the only two primary elections held during the
time frame studied. In 1992, the level of voter participation (17.21 percent) was almost
3 percentage points higher than in 2008 (14.65 percent) when the Wolverine State held an
early-voting contest on January 15. In the early 1990s, almost one third of the delegates were
awarded by the time the primary calendar reached Michigan. Contrary to our expectations,
organizing an early nominating event in 2008 at a time when barely one percent of the
delegates were selected did not promote participation in the Democratic primary. This finding
requires further explanation.
Knowing that the voting-age population in 2008 amounts to more than 600,000 citizens as
opposed to 1992 and that the 1992 event was scheduled almost two months after the primary
in 2008, how can we explain the absence of a high turnout in 2008? Put differently, it seems
that factors beyond the electoral calendar shape participation in Democratic primaries. To
explain these counter-intuitive findings, we argue that voters were mobilized to vote in the
1992 primary by aspects of the campaign, namely candidate strategies, candidate spending
and competition. Presidential hopeful and front-runner, Clinton, allocated a considerable
portion of his time and money to the two primary contests held on March 17 in Michigan and
Illinois (quoted in The New York Times, March 5, 1992). Campaign spending is found to have
a positive impact on turnout (Geer 1989; Norrander and Smith, 1985; Ranney, 1977). In the
Wolverine State, Brown, “offered himself as the roaring voice of blue-collar outrage” (quoted
in The New York Times, March 18, 1992) and his hard-charging campaign in the state was
aimed at economic unrest. As a Chicago Tribune reporter summarizes:
19
“in 1992, there was a lot of competition for the union constituency
Buchanan hoped to reach – nearly a third of workers in Michigan are union
members. Former California Gov. Jerry Brown, a Democrat, was
campaigning in the state wearing a UAW jacket. Ross Perot was in the race,
and so was Clinton, who drew together a strong coalition of union members,
African-Americans and white conservatives” (quoted in Chicago Tribune,
1996).
Equally important were candidate strategies in the 2008 campaign. In spite of extraordinary
voter participation in the 2008 nomination process and greater enthusiasm and interest in the
delegate-selection campaign across America because of the elongated two-person race
between Obama and Clinton, slightly more than 14 percent of the Democratic electorate went
to the polls in Michigan. These mixed turnout findings lead us to suspect that timing might
not have been an important factor in shaping turnout. Rather Democratic participation was
influenced by contextual and campaign variables. Only Hillary Clinton’s name was on the
Michigan ballot among the three major contenders. Both Barack Obama and John Edwards
took their names off in deference to the national party when the Great Lakes State jumped
ahead of “Super Tuesday” by holding its primary in January. In addition, the fact that
Democratic candidates largely refrained from campaigning in Michigan also affected turnout.
With the exception of Dennis Kucinich, none of the presidential aspirants of the Democratic
Party campaigned (USA Today, 2008) in the state, honoring pledges not to campaign in
Michigan. Yet studies have found participation to be positively related by the amount of
attention candidates allocate to a particular contest (Aldrich 1980; Atkeson and Maeastas
2009; Ranney 1977). Rosenstone and Hansen show how more campaigning by White House
aspirants promotes electoral involvement because these campaigns bear the costs of
information that a citizen would otherwise have been liable for (2003). But, in 2008,
Democrats in Michigan had mainly to bear these costs.
Finally, in 1996, given President Clinton’s lack of primary opposition to his renomination,
Michigan Democrats shifted from a primary to a “firehouse caucus”. As a reporter condenses,
Democrats “have little reason to go to the polls Tuesday, except in largely African-American
Detroit, which votes on a proposal to build a new baseball stadium” (quoted in Chicago
Tribune, 1996) were held simultaneously and using the same ballots as the presidential
primary. This issue had presumably increased the interests of voters in metropolitan Detroit –
who are mainly Democrats and independents – to go to the polls and to cross-over by
participating in the contested Republican primary. This finding is congruent with other studies
20
(Norrander and Smith, 1985; Patterson, 2002) which find that the presence of additional races
on the ballot has an impact on turnout.
In contrast with Republican primaries, we find no evidence that the front-loaded electoral
calendar in Michigan has promoted citizen involvement. This confirms our second hypothesis
about the differentiated impact of front-loading on Republican and Democratic primaries. Our
analysis points out that participation in Democratic delegate-selection events was shaped by
candidate strategies, candidate spending, competition, and the presence of additional races on
the ballot.
2. FRONT -LOADING VERSUS BACK-LOADING: A COMPARISON WITH INDIANA
After having examined the participation patterns in a state which has jumped on the frontloading bandwagon in the early 21st century, we now turn to the investigation of voting levels
in Indiana where nomination events are consistently scheduled in the first week of May in
spite of the presence of a highly compressed schedule. Our aim here is to evaluate and
compare the impact of the nominating calendar on turnout in both front-loaded and backloaded contests. We first look at Republican primaries before turning to delegate-selection
events for the Democratic Party. In each section, the impact of the calendar is first analyzed
before examining the effect of additional factors. We finally conclude.
a. GOP Presidential Contests: Combining Balloting for President with Important Races
for Statewide Office to Attract Voters to the Polls
In spite of back-loaded nominating events and nomination effectively settled for weeks by the
time Hoosiers got to vote (see Table 1), turnout is not significantly low in Indiana Republican
primaries. This goes against our hypothesis raised in the beginning of this paper. Turnout
levels fluctuated modestly in the Hoosier State. Indeed, the lowest turnout rate occurred in
2000 when nearly 18 percent of voters participated. That year almost 80 percent of the
delegates were already awarded by the time the campaign rolled into the state. In contrast,
Indiana set a record in voting participation in 2012 with almost one out of three voters casting
a ballot. It should be noted that the most recent electoral cycle noticed an overall movement
backward and unusual given the front-loading trend in recent decades. Hence, when voters in
Indiana turned out to vote in early May, less than 60 percent of the delegates were selected in
previous contests. But this pattern confirming our hypothesis that turnout fluctuates according
to the level of front-loading is not persistent in all electoral cycles. For instance, in 1996, more
than 80 percent of the delegates were awarded when Hoosiers went to the polls. Yet, their
21
number was significantly important, with 26 percent of the Republican partisans participating.
This finding suggests that other variables than the electoral calendar mobilized the Republican
electorate to vote in the primary.
Table 5: Participation in Presidential Primaries in Indiana, 1992-2012 (raw totals)
May 5, 1992
May 7, 1996
May 2, 2000
May 4, 2004
May 6, 2008
May 8, 2012
 of Votes Cast in
Republican Primary
 of Votes Cast in
Democratic Primary
Voting-Age
Population (VAP)
467,615
476,849
4,198,000
(60%)
(62%)
516,514
329,462
(81%)
(81%)
406,664
293,172
(78%)
(79%)
469,528
317,211
(70.9%)
(83.8%)
412,684
1,278,355
(84.3%)
(87.8%)
635,589
221,466
(58.4%)
(58.6%)
4,146,000
4,527,523
4,660,157
4,826,231
4,945,762
NB: In brackets are the cumulative percentages of Republican and Democratic delegates that had been selected by
the end of the week preceding the primary contest in Indiana.
Source: Federal Electoral Commission; Indiana Secretary of State; and United States Election Project.
Among these variables, the presence of a favorite-son contender on the ballot may have
fostered participation in Indiana (Atkeson and Maestas, 2008). Indeed, with former
Indianapolis mayor and then-four-term Senator Richard Lugar running for the GOP
presidential nomination, Republican voters in Indiana would have been expected to show
greater interest in the 1996 contest. Yet, by the time Hoosiers got to vote, Indiana’s native son
had already suspended his campaign. In addition, GOP Hoosiers were mobilized to participate
in the primary by another factor outside of the presidential race: the presence of additional
races on the ballot. In contrast with Michigan, the Hoosier State holds its state primary on the
same day as its presidential primary. And thus voters residing in states with a contested and
interesting congressional election have incentives to go to the polls. Indeed, the 2012
campaign cycle saw Indiana turnout further rise to 28.03 percent of the GOP voting-age
population. Nonetheless Romney locked up his party presidential nomination several weeks
before the Hoosier State’s primary, leaving Indiana voters without the national fanfare of the
22
deeply contested 2008 Democratic race (see infra). Hoosiers were however mobilized to
participate in the primary by factors outside the presidential race. Indeed, the Republican
Senate race between Senator Richard Lugar and Indiana Treasurer Richard Mourdock for the
seat Lugar first won in 1976 and held ever since apparently did grab voter interest. This
finding is consistent with the literature on electoral participation (Norrander and Smith, 1985;
Patterson, 2002). Indeed, Norrander argues that in spite of an already settled presidential
nomination, congressional races might promote participation in contests scheduled in the
latter half of the primary season (1986c:39). Wattier, in a study on front-loading, reports that
while turnout dropped below 20 percent after the last credible challenger ended his or her
campaign in the 2000 nomination race, it increased toward the end of the season because
those late-voting states still had competitive state-level elections (2005:11). Using data from
all state primary elections from 1972-2004, Atkeson and Maestas find turnout to increase by
5.5 percent when presidential primaries are held simultaneously as state-level elections
(2008).
This examination of voting levels in back-loaded contests suggests that, as opposed to our
first hypothesis, the compressed delegate-selection calendar does not significantly lower
turnout. On the one hand, we notice that for some electoral cycles, participation rates seem to
fluctuate according to the percentage of delegates awarded. Yet, there is no evidence for the
1996 election. This points out that, on the other hand, determinants such as the presence of a
favorite-son candidate as well as the presence of additional races on the ballot seem to matter
equally.
b. Democratic Primaries: Highest State Turnout in the 2008 Nominating Process
Similar to Republican contests, participation rates in Democratic primaries were not
disproportionately low. Turnout fluctuated to a larger degree than for GOP primaries. The
lowest voting level was registered in 2012, with barely one in ten voters casting a ballot. Yet,
the percentage of delegates selected was reasonably low (58.6%) compared to the other
electoral cycles when it approximates 80 percent. This finding does not correspond with our
hypothesis. In contrast, 2008 experienced the highest electoral level with more than five in ten
voters participating. In line with our hypothesis, this may be explained by the competitive
nature of the Democratic race. In fact, in contrast to previous elections when the presidential
nomination was a foregone conclusion by the time voters in Indiana got to the polls, the race
was not settled in early May 2008. Indeed, Barack Obama and Hillary Clinton came to
Indiana with their party nomination still up for grabs. This explains why Hoosiers responded
23
in record numbers with the 2008 primary experiencing a historic turnout rate of 54.71 percent
– almost 20 percentage points higher than in all other primaries. In a study on the 2008
election, Norrander finds participation to equal 35 percent of the partisan voting-age
population in Democratic primary events (2010:62). When compared with the total votes in
the previous Democratic primary, the 2008 figure represents an improvement of roughly one
million. Indiana Democrats were thus among the 27 of the 40 Democratic primaries to break
their old record and to set a record-high turnout (Patterson, 2009:54). Reporting on the
incredible turnout increase, voter-turnout specialist Patterson remarks that “[u]nlike in other
recent primary elections, turnout did not fall after Super Tuesday” (2009:57) owing to the
intensity of the Obama-Clinton race.
But this voting record needs to be explained by considering other factors. Indirectly, this
uniquely prolonged nomination battle was possible because Democratic proportional
representation rules, according to Schaffner, “prolong the divisiveness of the nomination
battle and make achieving party unity more difficult” (2012:167). Besides, the intensive
campaigning by both contenders in the state as well as massive spending on campaign ads
help to influence participation. In addition to campaign variables, participation in Indiana
Democratic contest was influenced by aspects of the legal structure, i.e. cross-over voting. An
observer argues that “given the choice of choosing a presidential candidate for the first time,
enough of those moderates and independents voted in the Democratic primary” (quoted in The
Statehouse File, 2012).
In this last examination of turnout in presidential primaries, we find strong evidence
supportive of our first hypothesis. If the nomination race is already settled by the time
Hoosiers went to the polls, turnout ranged from slightly more than 11 percent to 25 percent.
But, as in 2008, if the race is not yet determined, participation may be considerably higher.
Yet, additional factors shape turnout such as legal and campaign determinants.
V.Conclusion
This study has focused on the impact of the front-loaded calendar on primary voter turnout, a
factor that has been somewhat overlooked in the literature. One of the main criticisms of
front-loading is the early effective end of the nomination campaign (Atkeson and Maestas
2009). Indeed, as opposed to the pre-1968 period where candidates were able to remain in the
race until the end of the primary season, contemporary White House aspirants withdraw more
quickly. This, consequently, leaves fewer choices for subsequent primary voters. As
candidates step down from the race, citizens have fewer incentives to turn out (Atkeson and
24
Maestas 2008). If the major party nomination races are still unresolved by the time voters turn
out to vote, we hypothesize, it may encourage more citizens to participate in selecting the
Democratic and Republican nominees. In other words, we may expect voter turnout to be
significantly higher in elections scheduled before the nomination race is all but settled. If, on
the contrary, the winner of the nomination has been declared by the time voters get to the
polls, citizen participation is expected to collapse since millions of voters are left without the
chance of casting a meaningful vote. In addition, we hypothesize that the electoral calendar
may have a different impact on Democratic and GOP primary participation levels.
To test our hypotheses, we analyze turnout rates in both major parties’ delegate-selection
contests in Michigan and Indiana from 1992 to 2012. Two intriguing patterns called for
explanation: the fact that Republican and Democratic participation levels are not influenced
equally by the electoral calendar even though both major parties schedule their events on the
same day and the absence of dramatically low voting levels in late-scheduled primaries.
With regard to the first pattern, we find strong evidence supportive of the first hypothesis
about the importance of timing. Casting an early ballot is identified as a strong predictor of
voting. This is important, since presidential races are today settled in the early spring and
voters in late-scheduled primaries are left without a meaningful voice. But, in line with our
second hypothesis, partisan elections seem to be influenced differently. While turnout in some
Michigan Republican contests and Indiana Democratic seems to be fostered by the sequential
nomination system, we find no evidence of this effect in Michigan Democratic and Indiana
Republicans primaries. The puzzle of why voters in Michigan Democratic contests turned out
in lower rates could be explained by campaign variables. Part of the difference is due, as in
the 2008 event, to candidate strategies when all major contenders refrained from campaigning
in the Midwestern state.
Yet the second lesson from this comparison is the absence of a steep decline in voting
participation in Indiana after front-loading emerged in the late 1980s. Indiana does not serve
as a good case to show the negative effect of front-loading on citizen involvement in latescheduled primaries. In both Republican and Democratic elections we find other factors
explaining participation. Combining balloting for president with state-level contests has
prevented voter participation rates to plummet. Campaign and legal factors play also an
important role in influencing turnout. Our findings suggest that intensive campaigning by
presidential aspirants affects turnout in primaries as campaign activities foster electoral
involvement by reducing the costs of voting, and particularly, the costs of information
25
(Ranney 1977; Rosenstone and Hansen 1993). Moreover, legal factors and above all, the open
nature of contests in Indiana, is an important variable influencing participation as emphasized
in the 2008 Democratic primary.
Although the study adds to the literature on front-loading, it also has severe limitations. First,
the conclusions are derived from relatively few contests in two individual states in a limited
time frame. We readily concede that additional research is needed to assess whether our
results generalize to other populations and other elections, and that our findings must be
viewed as preliminary until additional statistical research is conducted. Moreover, it is based
exclusively on primary data. Data from caucus election is not included. Yet, Michigan
Democrats used these mechanisms on several occasions which prevents extensive
comparison.
Acknowledgment
An earlier version of this article was presented at the UK Annual Conference of the American
Politics Group of the Political Studies Association, January 3-5, 2013, Leicester. I would like
to thank panel participants as well as my PhD supervisor Emilie van Haute for their
comments, advice and support.
VI. References
 Scientific Literature
Aldrich, John H., (1980), Before the Convention: Strategies and Choices in Presidential
Nomination Campaigns, Chicago: University of Chicago Press.
Aldrich, John H., (1993), “Rational Choice and Turnout”, American Journal of Political
Science, 37(1):246-278.
Atkeson, Lonna Rae, (2010), “Voter Decision Making on the Heels of Iowa and New
Hampshire”, in David W. Patton and Jennifer L. Robinson (eds.), The Rise of the West in
Presidential Elections, Salt Lake City: University of Utah Press.
Atkeson, Lonna Rae and Cherie D. Maestas, (2009), “Meaningful Participation and the
Evolution of the Reformed Presidential Nominating System”, PS: Political Science &
Politics, 42:59-64.
Blais, André, (2000), To Vote or Not to Vote. The Merits and Limits of Rational Choice
Theory, Pittsburgh: University of Pittsburgh Press.
Boyd, Richard D., (1986), “Election Calendars and Voter Turnout”, American Politics
Quarterly, 14(1-2):89-104.
Buell Jr., Emmett H. and William G. Mayer (eds.), (2004), Enduring Controversies in
Presidential Nominating Politics, Pittsburgh: University of Pittsburgh Press.
26
Burden, Barry C. and Philip Edward Jones, (2009) “Strategic Voting in the USA”, in Bernard
Grofman, André Blais, and Shaun Bowler (eds.), Duverger’s Law of Plurality Voting. The
Logic of Party Competition in Canada, India, the United Kingdom and the United States,
New York: Springer.
Chen, Kong-Pin and Sheng-Zhang Yang, (2002), “Strategic Voting in Open Primaries”,
Public Choice, 112(1-2):1-30.
Cook, Rhodes, (2004) The Presidential Nominating Process: A Place for Us?, Lanham:
Rowman & Littlefield Publishers.
Cook, Rhodes, (2008), Race for the Presidency: Winning the 2008 Nomination, Washington,
D.C.: CQ Press.
Davis, James W., (1980), Presidential Primaries. Road to the White House, Westport:
Greenwood Press.
Downs, Anthony, (1985), An Economic Theory of Democracy, New York: Addison-Wesley.
Galderisi, Peter F. and Marni Ezra, (2001) “Congressional Primaries in Historical and
Theoretical Context”, in Peter F. Galderisi, Marni Ezra and Michael Lyons (eds.),
Congressional Primaries and the Politics of Representation, Lanham: Rowman & Littlefield
Publishers.
Geer, John G., (1989), Nominating Presidents: an Evaluation of Voters and Primaries,
Westport: Greenwood Press.
Hagen, Michael G. and William G. Mayer, (2000), “The Modern Politics of Presidential
Selection: How Changing the Rules Really Did Change the Game”, in William G. Mayer
(ed.), In Pursuit of the White House 2000: How We Choose Our Presidential Nominees, New
York: Chatham House Publishers.
Jewell, Malcolm E., (1977), “Voting Turnout in State Gubernatorial primaries”, Western
Political Quarterly, 30:236-254.
Jewell, Malcolm E. and Sarah M. Morehouse, (2001), “The Nominating Process”, in Malcolm
E. Jewell and Sarah M. Morehouse, Political Parties and Elections in American States,
Washington D.C.: CQ Press.
Kenney, Patrick J. and Tom W. Rice, (1985), “Voter Turnout in Presidential Primaries: A
Cross-Sectional Examination”, Political Behavior, 7(1):101-112.
Maisel, Sandy L. and Mark D. Brewer (2012), Parties and Elections in America. The
Electoral Process, Lanham: Rowman & Littlefield Publishers.
Mayer, William G., and Andrew E. Busch, (2004), The Front-Loading Problem in
Presidential Nominations, Washington, D.C.: Brookings Institution Press.
Milbrath, Lester W. and Lal M. Goel, (1977), Political Participation: How and Why Do
People Get Involved in Politics?, Chicago: Rand McNally.
Moran, Jack and Mark Fenster, (1982), “Voter Turnout in Presidential Primaries: A
Diachronic Analysis”, American Politics Research, 10(4):453-476.
27
Norrander, Barbara and Gregg W. Smith, (1985), “Type of Contest, Candidate Strategy, and
Turnout in Presidential Primaries”, American Politics Quarterly, 13(1):28-50.
Norrander, Barbara, (1986a), “Correlates of Vote Choice in the 1980 Presidential Primaries”,
The Journal of Politics, 48(1):156-166.
Norrander, Barbara, (1986b) “Measuring Primary Turnout in Aggregate Analysis”, Political
Behavior, 8(4):356-373.
Norrander, Barbara, (1986c), “Selective Participation: Presidential Primary Voters as a Subset
of General Election Voters”, American Politics Quarterly, 14(1-2):35-53.
Norrander, Barbara, (1992), Super Tuesday: Regional Politics and Presidential Primaries,
Lexington: University of Kentucky Press.
Norrander, Barbara, (1996), “Presidential Nomination Politics in the Post-Reform Era”,
Political Research Quarterly, 49(4):875-915.
Norrander, Barbara, (2010), The Imperfect Primary. Oddities, Biases, and Strengths of U.S.
Presidential Nomination Politics, New York: Routledge.
Patterson, Thomas E., (2002), The Vanishing Voter. Public Involvement in an Age of
Uncertainty, New York: Alfred A. Knopf.
Patterson, Thomas E., (2009), “Voter Participation: Records Galore This Time, but What
about Next Time?”, in Steven S. Smith and Melanie J. Springer (eds.), Reforming the
Presidential Nomination Process, Washington, D.C.: The Brookings Institution.
Patton, David W. and Jennifer L. Robinson (eds.), The Rise of the West in Presidential
Elections, Salt Lake City: University of Utah Press.
Ranney, Austin, (1972), “Turnout and Representation in Presidential Primary Elections”,
American Political Science Review, 66(1):21-37.
Ranney, Austin, (1977), Participation in American Presidential Nominations, 1976.
Washington, D.C.: American Entreprise Institute for Public Policy Research.
Riker, William H. and Peter C. Ordeshook, (1982), “A Theory of the Calculus of Voting”,
The American Political Science Review, 62(1):25-42.
Rosenstone, Steven J. and John Mark Hansen, (2003), Mobilization, Participation, and
Democracy in America, New York: Longman Classics Edition.
Rothenberg, Lawrence S. and Richard A Brody, (1988), “Participation in Presidential
Primaries”, The Western Political Quarterly, 41(2):253-271.
Rubin, R.L., (1980), “Presidential Primaries: Continuities, Dimensions of Change, and
Political Implications.” in William J. Crotty (ed.), The Party Symbol, San Francisco: W. H.
Freeman.
Schaffner, Brian F., (2012), Politics, Parties and Elections in America, Belmont: Wadsworth.
Schier, Steven E., (1982), “Turnout Choice in Presidential Nominations: A Case Study”,
American Politics Research, 10(2):231-245.
28
Smith, Steven S. and Melanie J. Springer (eds.), (2009), Reforming the Presidential
Nomination Process, Washington, D.C.: The Brookings Institution.
Verba, Sidney and Norman H. Nie, (1972), Participation in America: Political Democracy
and Social Equality, New York: Harper & Row.
Webb, Paul, (2004), “Campaigns, Elections and Electoral Systems”, in Mary Hawkesworth
and Maurice Kogan (eds.), Encyclopedia of Government and Politics, New York: Routledge.
Wekkin, Gary D., (1988), “The Conceptualization and Measurement of Crossover Voting”,
The Western Political Quarterly, 41(1):105-114.
Wolfinger, Raymond E. and Steven J. Rosenstone, (1980), Who Votes?, New Haven: Yale
University Press.
 Working Papers
Alvarez, Michael R. and Jonathan Nagler, (1997), “Analysis of Crossover and Strategic
Voting”, expert witness report prepared for California Party v. Bill Jones.
Atkeson, Lonna Rae, and Cherie D. Maestas, (2008), “Racing to the Front: The Effect of
Frontloading on Presidential Primary Turnout”, Paper prepared for the “2008 Reforming the
Presidential Nomination Process Conference”, at the University of Iowa, Iowa City, January
3-4, 2008.
Mayer, William G., (2012), “The Presidential Nomination Process in the 2012 Republican
Race: An Early Assessment”, Paper sent by the author.
Wattier, Mark J., (2005), “Presidential Primaries and Frontloading: An Empirical Polemic”,
Paper prepared for “State of the Party: 2004 and Beyond,” A Conference Sponsored by the
Ray C. Bliss Institute of Applied Politics, University of Akron, Akron, OH, October 5-7,
2005.
 Newspaper Articles
APPLE R. W. Jr., “The 1992 Campaign: Primaries and Caucuses; As Field Narrows, 2
Democrats Spar on Economic Plans”, The New York Times, March 5, 1992.
BLAKE Aaron, “Michigan Democratic Party Encourages Crossover Voting in GOP
Presidential Primary”, Washington Post, February 22, 2012.
BRADSHER Keith, “The 2000 Campaign: Michigan; McCain Could Singe Fire Wall for
Bush in Michigan Primary”, The New York Times, February 9, 2000.
CHRISTOFF Chris and SCHMITT Ben, “Romney, Clinton Lead Michigan Poll”, USA
Today, January 14, 2008.
HOFFMAN Kathy B., “Democratic Crossover in Mich. Didn’t Stop Romney”, Real Clear
Politics, February 29, 2012.
JACOBS Ben, “Michigan’s Primary Party Crashers: Democrats Crossing Over to Thwart
Romney”, The Daily Beast, February 28, 2012.
29
KRULL John, “Presidential Primary Could Affect Indiana’s Senate Race”, The Statehouse
File, February 10, 2012.
MEJIA Mercedes and WHITE Jennifer, “Presidential Primary, Democrats casting “crossover”
vote?”, Michigan Radio, February 14, 2012.
TONER Robin, “The 1992 Campaign; Clinton Wins Big Victories in 2 Midwestern
Primaries; Bush Increases G.O.P. Edge”, The New York Times, March 18, 1992.
TROY Tom, “Indiana Get A Shot at Political Spotlight”, Toledo Blade, April 13, 2008.
WORTHINGTON Rogers, “Some GOP Voters Are Thinking Clinton”, Chicago Tribune,
March 18, 1996.
Notes
1
Exploring additional election cycles is prevented by the interruption of the conduct of presidential primaries in
Michigan from 1976 to 1988. And because front-loading became evident in the late 1980s, it does not make
sense, in our opinion, to extend our analysis to electoral cycles prior to 1992.
30