Participation in Presidential Primaries

• Author: Lawrence S. Rothenberg,Richard A. Brody
• DOI: 10.1177/106591298804100203
• Published date: 01 June 1988
• Date: 01 June 1988
• Subject Matter: Articles

PARTICIPATION IN PRESIDENTIAL PRIMARIES
LAWRENCE S. ROTHENBERG, California Institute of Technology
and
RICHARD A. BRODY, Stanford University
INCE
1968 the number of presidential primaries has exploded. In
1984 Democratic voters participated in 26 primary elections; Repub-
lican electors were involved in 29. They selected 56.4 percent and
68.2 percent of the delegates to their party’s respective nominating con-
ventions. The proliferation of primary contests is one of the most sig-
nificant changes in electoral politics in recent years. It has become
increasingly important to understand what motivates citizens to vote in
these elections.
Participation in presidential primaries contrasts sharply with the highly
publicized decline in general election turnout since 1960; over this same
period, rates for primary elections have not decreased noticeably. The
analysis that follows attempts to explain why citizens participate in pri-
mary elections. We examine voting patterns in primaries from 1952 to
1980 at the aggregate level and find that election-specific characteristics
-
particularly those affecting electoral stakes -
are important deter-
minants. We
also suggest that a number of countervailing processes have
kept primary participation from declining along with general election
turnout. Primaries are heavily influenced by context; changes in it have
had a notable impact on participation rates and have prevented turnout
from spiraling downward.
PREVIOUS RESEARCH ON PRIMARY TURNOUT
Primary turnout is not well understood. Factors identified as signifi-
cant determinants of participation in primary elections vary from one
analysis to the next. For the most part, students have focused on two
broad sets of explanatory factors: changes in the rules and scheduling
of primaries and differences in particular aspects of each primary cam-
paign, such as the levels of competition among candidates (for overviews
of previous research, see Davis 1980; Moran and Fenster 1982). The ba-
sic problem, however, is that no analysis has incorporated these factors
into a general model of the turnout process. Resulting specification er-
rors, as well as estimation problems, have led to widespread disagree-
ment over what determines primary turnout.
A
number of studies, for example, have focused on whether the rules
under which a primary is conducted significantly affect turnout rates.

---

254
Ranney (1977) found that participation tends to be higher in binding
primaries than in advisory contests. But others discovered no significant
variation among states employing different rules (Morris and Davis 1975).
Similarly, some scholars emphasized the importance of competition
among candidates (Zeidenstein 1970; Morris and Davis 1975; Morris,
Sample and Davis 1976; Moran and Fenster 1982). Yet others found that
competition is insignificant in determining turnout (Ranney 1977). Its im-
pact on primary participation -
like that of differing rules -
remains
uncertain.
Perhaps the best attempt to discern which factors influence primary
turnout is Moran and Fenster’s (1982) multivariate model of changes in
primary turnout levels between 1968 and 1980 (see also Kenney and Rice
1985). However, even their analysis has a number of drawbacks. First,
they examined only change: They do not explain why mean levels of turn-
out vary consistently across states over time. Second and more impor-
tant, they did not control for possible state-specific or election-specific
effects, raising the possibility of autocorrelated and heteroskedastic er-
ror components (e.g., Kmenta 1971).’ As we will demonstate below,
election-specific effects in particular are quite significant in the years Mo-
ran and Fenster studied. Finally, despite their claim that they were con-
ducting a diachronic analysis, they could say little about trends in primary
turnout.
It is time to move beyond haphazardly specified models that contain
one or more
of the features believed to be important for primary partici-
pation. Studies need to be informed by both a theoretical understanding
of the turnout process and a thoughtful analysis of the details distinguish-
ing primary elections.
THEORIES OF PARTICIPATION AND PRIMARY VOTING
For much of the last several decades, the study of the participation
process has revolved around citizens’ decision-making calculi. Potential
participants are claimed to behave as if they maximize their expected
value. The decision to vote has largely been analyzed using one variant
or another of Downs’s (1957) expected utility formulation.
As is by now common knowledge, Downs hypothesized that citizens
vote if the benefits stemming directly from their own participation ex-
ceed the costs of involvement. While the efficacy of any single elector
is problematic (Riker and Ordeshook 1968; Meehl 1977), it is true that
voters’ perceptions of both their stake in an election and the costs of par-
ticipating are important predictors of who goes to the polls.
A considerable body of empirical analysis purports to demonstrate
that greater perceived benefits result in higher electoral turnout. Wol-
finger and Rosenstone (1980) discovered that groups with a substantial
1
This criticism also applies to the recent piece by Kenney and Rice (1985). However, a
probably more severe problem with this piece is the almost certain presence of en-
dogeneity between one of the independent variables, percentage registered, and turnout.

---

255
stake in government programs, such as farmers, vote in greater numbers;
so too do government employees in states with extensive patronage sys-
tems. More recently, Wolfinger and his associates (Squire, Wolfinger and
Glass 1985) found that citizens who have resided in their communities
for a long time - and who, it stands to reason, believe that they have
a larger stake in the political process -
participate more.2 Similarly, Brody
and Page (1973; but cf. Weisberg 1977; Brody and Grofman 1982) dis-
covered that individuals perceiving differences between candidates are
more
likely to vote. Strong party identifiers, who presumably think that
they have more riding on electoral outcomes, have also long voted in
greater numbers in general elections (Campbell et al. 1960; Abramson and
Aldrich 1982).
Costs and individuals’ capacity to reduce them are associated with
turnout as well: Education, experience, familiarity with the details of
qualifying to vote, and a micro- or macrosocial environment that facili-
tates participation are all related to higher rates of participation (Verba
and Nie 1972; Brody 1978; Rosenstone and Wolfinger 1978; Wolfinger
and Rosenstone 1980; Katosh and Traugott 1982). Conversely, residen-
tial mobility and physical immobility, complex registration requirements,
and personal circumstances that divert one’s attention from the election
are all barriers to voting.
But even the most broadly conceptualized expected utility models
have been inadequate for explaining turnout. The electorate appears sus-
ceptible to influences beyond the costs and benefits of voting. For ex-
ample, both psychic rewards (Campbell et al. 1960) and habituation to
the act of voting (Brody and Sniderman 1977) are important factors in
stimulating participation. Particularly in light of the limited interest and
lack of knowledge among members of the electorate, these findings sug-
gest that voters may not be quite as calculating as is frequently inferred.
Is there a viable alternative model of participation? The answer is yes,
if attention is directed toward elite decision-making. It may be that can-
didates, not voters, maximize their expected utility. Political aspirants
decide whether or not to run; and then candidates direct their time,
money, and other resources to serve their electoral ends. It is their ac-
tivities -
and those of the media that follow and reinforce the candi-
dates’ leads -
that stimulate or dissuade voters from going to the polls.
Mass participation may be a by-product of elite activities. The &dquo; ’surge
and decline’ phenomenon&dquo; (Campbell 1966) that typifies participation
in presidential and nonpresidential election years, and presumably charac-
terizes high- and low-stimulus presidential primaries, reflects elite
processes.
From this perspective, the electorate is quite malleable. Of course,
some
citizens will almost always vote, propelled by a variety of long-term
factors -
habit, feelings of obligation, party loyalty, etc. Others are so
apathetic or disaffected that they will never bother to cast a ballot. But
2

Squire, Wolfinger and Glass (1985) may be measuring voting costs as well as electoral stakes.

---

256
most of the voting-age population falls somewhere in between. They re-
spond to short-term stimuli, and their probability of voting increases sig-
nificantly in highly charged electoral environments. Therefore, the finding
that large stakes (e.g., greater candidate competition) make a difference
reflects voters getting caught up in high-stimulus electoral contexts. The
hypotheses that elite decision-making is crucial in elections and that cam-
paign resources can be employed to shape voters’ opinions have been
convincingly demonstrated in recent years (Jacobson 1980; Jacobson and
Kernell 1981; Calderia and Patterson 1982a). There is no reason to be-
lieve that they are not important for turnout as well (Calderia and Patter-
son 1982b, 1983). Voters may behave as if they are acting rationally when
they are simply responding to environmental stimuli.3
3
In short, there are two...