Would Primary Voters Support Moderate Candidates?
Danielle M. Thomsen
Syracuse University
Abstract
This paper examines the conditions under which moderate candidates can gain support from
primary voters. I draw on an original survey of 1,000 Republican and Democratic primary
voters as well as primary election results from 1980 to 2012. First, I find that primary voters
strongly prefer ideologues to moderates. Second, the observational data similarly demonstrate
that moderates are less likely to win the primary than candidates at the extremes. Yet I also
show that moderates are more likely to win as the number of primary candidates increases, and
the effect of being a moderate diminishes when there are more candidates in the race. Thus,
although primary voters may favor ideologues over moderates, moderate candidates may, under
certain circumstances, have hope in primary elections after all.
Paper prepared for the 2015 American Political Science Association conference. Previous
versions were presented at the Parties as Micro-Polities workshop at Duke University, the 2015
Midwest Political Science Association conference, and the Political Institutions and Public
Choice (PIPC) 20th Anniversary conference at Duke University.
The dramatic rise in partisan polarization in Congress has been one of the most prominent
topics of academic and political debate for the past decade. In the current 114th Congress, there
is no ideological overlap between the two parties, and the distance between the Republican and
Democratic parties is now at a post-Reconstruction high (e.g., McCarty, Poole, and Rosenthal
2006). One of the most commonly cited explanations for polarization is the primary election
system. Senator Charles Schumer recently summarized this view in an editorial in the New York
Times: “The partisan primary system, which favors ideologically pure candidates, has
contributed to the election of more extreme officeholders and increased political polarization. It
has become a menace to governing.” As Schumer and many others have suggested, because
only a small faction of ideologically motivated voters turns out in primaries, they pull candidates
toward the extremes and “warp the national balance” of the electoral system. Indeed, this
argument has been so powerful that almost all who seek congressional reform advocate for
changes in the party primary system (i.e., Fiorina et al. 2006; Mann and Ornstein 2012).
Yet despite the logical appeal of the party primary argument, the impact of primaries on
congressional polarization is far from clear. On the one hand, a large body of research shows
that party activists have become increasingly extreme over the past few decades (e.g., Fiorina et
al. 2006; Layman and Carsey 2002; Layman et al. 2010; Theriault 2008). Abramowitz (2010)
provides one of the most comprehensive accounts of how the politically engaged subset of
Americans is deeply divided along ideological lines. And it is of course these individuals who
are the most likely to participate in primary elections. Similarly, the Pew Research Center
released a new report on mass polarization showing that Americans with more consistently
conservative and liberal views are much more likely to vote in the primary than those with a mix
of conservative and liberal views (Pew Research Center 2014).
1
At the same time, however, political scientists have for the most part struggled to find
linkages between partisan primaries and polarization. For one, the evidence that ideologues even
fare better in primaries is mixed (Brady et al. 2007; Hall and Snyder 2015; Hirano et al. 2010),
and across studies, the magnitude of the effect of candidate ideology on primary outcomes is
small (see Hall and Snyder 2015 for a review). Furthermore, Hirano et al. (2010) show that the
introduction of primary elections, the level of primary turnout, and the threat of primary
competition are not associated with partisan polarization in roll call voting. Differences in
primary rules also seem to provide few answers. Closed primaries, or those in which only party
members can vote, do not produce more extreme candidates than open primaries (McGhee et al.
2014; Rogowski and Langella 2014; but see Gerber and Morton 1998). Sides and Vavreck
(2013) attribute these collective dead ends to the fact that primary voters look similar on many
measures to other voters within their party (see also Geer 1988; Norrander 1989). They
conclude, “Polarization does not seem to emanate from voters at any stage of the electoral
process” (Sides and Vavreck 2013, 11).
Additional evidence on the limited effect of primaries comes from recent reforms. Most
notably, the implementation of the top-two primary in 2012 in California was expected to
increase turnout and thereby diminish the impact of extreme voters on candidate selection.
Subsequent studies suggest that this was perhaps too optimistic. Moderate candidates fared no
better under the top-two primary than they would have in closed primaries (Ahler, Citrin, and
Lenz 2015), and if anything, California lawmakers took more extreme positions after 2012
(Kousser, Phillips, and Shor 2013). Ahler et al. (2014) suggest the reform’s failure is because
voters were largely unaware of the ideological orientation of candidates. Hirano et al. (2015)
2
show that voters do learn about candidate ideology in gubernatorial and senate races, but there is
little indication that they do so in races with limited media coverage and resources.
Thus, it is first not clear whether and how much primary voters even favor extremists. It
is also not clear whether primary voters learn about candidate ideology in U.S. House races and
are able to make a choice based on ideology. These questions are difficult to examine with
observational data because moderates are much less likely to run for Congress than those at the
extremes (Thomsen 2014a). Regardless of whether voters have complete information on
candidate positions, a liberal Republican or conservative Democrat is unlikely to even be on the
ballot. More generally, research on partisan polarization has paid less attention to the choices
that voters face. Yet political candidates are not selected in isolation, and they appear alongside
many others. It is possible that voters are more likely to select moderates under different
configurations of candidates. It is certainly the case that in order for a moderate to win, there
must be a moderate candidate from which voters can choose. But if we can learn about when
moderates are more likely to win, we can better evaluate when moderates could and should run.
The goal of this paper is to examine the conditions under which moderates can gain
support from primary voters. I draw on survey and observational data to analyze this question.
First, the survey data show that both Republican and Democratic primary voters strongly prefer
ideologues to moderates. This result conforms to the conventional wisdom about the preferences
of primary voters. Second, I use observational data to explore the relationship between candidate
ideology and primary election outcomes between 1980 and 2010. Moderates are less likely to
win the primary than those at the extremes, but I also find that moderates are more likely to win
as the number of primary candidates increases. The effect of being a moderate diminishes when
there is a greater number of candidates in the race. I do not sort through the specific mechanisms
3
here, but it is possible that, in these circumstances, the signal is too muddled or the information is
too costly to obtain. Thus, although primary voters may prefer extremists to moderates,
moderate candidates may, under certain circumstances, have hope in primary elections after all.
When Might Moderates Win?
We know very little about the actual preferences of primary voters and how central
ideology is to candidate evaluations and vote choice. The data that has emerged suggests that a
greater proportion of consistently conservative Republicans and consistently liberal Democrats
participate in party primaries than those who hold a mix of liberal and conservative positions
(Pew Research Center 2014). The conventional wisdom is certainly that primary voters today
prefer ideological extremists. Indeed, the most politically engaged public has become more
ideologically polarized over time (Abramowitz 2010), and when given full information on
candidate ideology, Republican voters will choose conservative Republicans and Democratic
voters will choose liberal Democrats. The idea that voters favor ideological proximity is not
new, and a vast literature on candidate behavior is rooted in the logic that candidates take
positions based on the preferences of the electorate (e.g., Downs 1957; Mayhew 1974).
Yet candidate ideology is not likely to matter for electoral outcomes in all circumstances,
and there are several ways in which moderate candidates, or at least relatively moderate
candidates, could win the primary. As noted above, if voters do not have full information on
candidate positions, it is difficult to see how they can reward or punish based on ideology. The
fact that primary voters have a hard time distinguishing between same-party candidates provides
support for this possibility (Ahler et al. 2015). Voter-candidate congruence may be an especially
tall order in a primary depending on the number of candidates on the ballot. Voters have to
4
invest different amounts of energy into learning about candidate ideology depending on the
configuration of candidates, and additional candidates on the ballot will require more effort from
voters to learn about their various positions. The main hypothesis here is that candidate ideology
will have a minimal effect on primary outcomes as the number of candidates in the race
increases. Put differently, even if primary voters prefer ideologues to moderates, the effect of
being a moderate may diminish when there are more candidates on the ballot. In these cases, the
ideological signal may be too muddled or the information may be too costly to obtain.
It is also possible that the ideological makeup of the primary competition matters for the
electoral success of moderates. For example, if there are multiple ideologues in the race, primary
voters may split their votes among these candidates. One potential outcome is that the moderate
candidate could win—again, despite the fact that primary voters generally prefer ideologues to
moderates. For the same reason, we might expect moderates to be even less likely to win as the
number of moderate candidates increases. The goal here is not to identify the specific
mechanism for why moderates could win under various circumstances, but rather to learn more
about the relationship between candidate ideology and primary outcomes and the conditions
under which moderates may be more likely to win the primary than we would otherwise expect.
As a result, the main contribution of the paper is largely empirical, but the broader
theoretical point is that candidates are not chosen in isolation. They are embedded among many
others, and it is the array of choices that matters for electoral outcomes and the persistence of
polarization in Congress. To be sure, it is often the case that candidates, particularly incumbents,
are unopposed in a primary, but it is rarely the case where open seat races do not attract an ample
field of primary candidates. And it is primarily through replacement processes that the
ideological gulf between the two parties has continued to widen (Theriault 2006). In sum, the
5
composition of candidates has an important impact on which individuals are ultimately elected,
but research on polarization has for the most part overlooked how the makeup of choices affects
the makeup of Congress as an institution. In the sections below, I draw on both survey and
observational data to analyze the preferences of primary voters and the conditions under which
moderates may be more likely to attract support from primary voters.
Preferences of Primary Voters
As noted at the outset, primary voters are widely believed to be the driving force behind
partisan polarization in Congress. Yet there is a dearth of data on how primary voters perceive
and evaluate ideologues compared to moderates, and this problem is exacerbated by the fact that
so few moderates are now running for Congress (Thomsen 2014a). To examine the preferences
of this small but influential subset of the American public, I draw on data from an original survey
of 1,000 Republican and Democratic primary voters. The survey was conducted by YouGov in
June and July of 2014, and YouGov verified that all respondents had participated in the 2012
primary elections. Each respondent received information about two U.S. House candidates who
shared their party affiliation.1 One candidate was described as more moderate than most of
his/her colleagues. The other candidate was described as more conservative or more liberal than
most of his/her colleagues (for Republican and Democratic primary voters, respectively). Both
candidates were portrayed as having previous political experience and being knowledgeable
about the district. The moderate and liberal vignettes for the Democratic male candidates are
provided below (the Republican vignettes are in the Appendix). Republican primary voters
received Republican candidates, and candidate ideology was manipulated in the same way for
1
The combination of candidates included two women, two men, or one woman and one man, but
candidate gender is beyond the focus of this paper.
6
Democratic and Republican primary voters. Primary voters were asked to evaluate each
candidate on their favorability, effectiveness as a U.S. representative, policy positions, and
character traits, and they also rated the ideology of the candidates.2 They were then asked which
of the two candidates they would vote for in the party primary.
Figure 1: Example of Candidate Descriptions
Figure 2 shows the mean favorability and effectiveness ratings of the moderate and
extreme candidates, broken down by party. The favorability and effectiveness scales range from
1 to 7, with higher values indicating more positive evaluations. Both Democratic and
Republican primary voters rate candidates who are described as more liberal and conservative
than their colleagues, respectively, higher than those who are described as more moderate than
2
Primary voters rated the conservative and liberal candidates as significantly more extreme than the
moderate candidate, so it is clear they were aware of the candidates’ ideological differences.
7
their colleagues. Among Democrats, the mean favorability rating for a moderate candidate is
4.99, compared to 5.72 for a liberal candidate, and their perceived effectiveness as a U.S.
representative is 4.93 for moderates and 5.36 for liberals. Similarly, among Republicans, the
mean favorability rating for moderate and conservative candidates is 4.65 and 5.74, respectively.
Moderate candidates receive an average effectiveness score of 4.54, and conservatives receive an
average score of 5.46. All of these differences are statistically significant at p<0.01.
6
Figure 2: Favorability and Job Performance Ratings Among Democratic and Republican
Primary Voters, For Extreme and Moderate Candidates
5.74
5.72
5.46
5.36
4.99
4.93
4.54
Favorability
Effectiveness
0
2
4
4.65
Favorability
Effectiveness
Democrats
Moderate Candidate
Republicans
Extreme Candidate
Primary voters were also asked which candidate they would vote for in the party primary.
Figure 3 illustrates the expressed preference between the extreme and moderate candidate. The
numbers look even worse for moderates when we examine expected vote choice. On the
Democratic side, 71% of primary voters favor the more liberal candidate, and on the Republican
side, 82% of primary voters favor the more conservative candidate (p<0.01). It is clear that
8
primary voters from both parties overwhelmingly prefer ideologues to moderates, though
Republican voters seem to have an even stronger preference for ideologues than do Democratic
voters. Republican primary voters are significantly less likely to support the moderate candidate
than their Democratic counterparts (p<0.01; differences not shown).
Figure 3: Expressed Vote Choice Between Extreme and Moderate Candidate, Among
Democratic and Republican Primary Voters
.8
0.82
.4
.6
0.71
0.29
0
.2
0.18
Democrats
Moderate Candidate
Republicans
Extreme Candidate
On the one hand, these patterns conform to the general narrative regarding the impact of
partisan primaries on polarization. Although the findings are not surprising, they support the
popular suspicion that primary voters strongly prefer ideologues to moderates when presented
with such a choice. But at the same time, it is never the case where the ideology of the
candidates is printed on the ballot, and it is often the case where the array of choices does not
simply include an ideologue versus a moderate. Furthermore, while the liberal Republicans and
conservative Democrats of yesteryear have largely vanished from office, we can nevertheless
9
analyze the electoral fortunes of relatively moderate congressional candidates that seek office
today. More generally, we are interested in when primary voters are more likely to select a
moderate candidate. The next section draws on observational data to delve into this question
further.
Candidate Configurations and Primary Outcomes
This section examines the conditions under which moderate candidates can win the
primary, particularly when there are different configurations of candidates on the ballot. The
analyses are based on primary election results for the U.S. House of Representatives from 1980
to 2012. Primary election results were obtained from the Federal Election Commission (20082012) and the America Votes series (Scammon, McGillivray, and Cook 1980-2006). These data
were merged with Bonica’s (2014) ideology estimates of candidates who ran for congressional
office during this time. Bonica (2014) uses campaign finance data to place the vast majority of
congressional candidates on a common ideological scale.3 What is particularly advantageous
about these data is that they allow us to analyze the ideology of the vast majority of candidates
who won as well as lost their primary. The full dataset includes a total of 23,653 Republican and
Democratic primary candidates; of these 23,000 candidates who appeared on the primary ballot,
17,395 (74%) have Bonica ideology scores.4
3
See Bonica (2014) for a full description of the data and validation.
The Bonica dataset includes candidates who filed with the Federal Election Commission. Candidates
who do not exceed the $5,000 threshold of campaign fundraising are not required to file. Those who are
excluded are thus more likely to be long-shot candidates, but it is not clear that they are more likely to be
extremists. Even so, these excluded candidates comprised only 7% of primary winners and 0.03% of
general election winners, so they have virtually no effect on party change in Congress. Furthermore, the
Bonica data provide the best publicly available measures of the ideological positions of primary winners
and losers over time. 4
10
The number of ideological moderates as a percentage of all primary winners during this
time is displayed in Figure 4 below. There are a few noteworthy patterns that emerge. First, the
percentage of Republican and Democratic primary winners who are at least as liberal as Olympia
Snowe (R-ME) and at least as conservative as Bart Gordon (D-TN) has declined over time. In
1980, nearly 19% of primary winners were liberal Republicans and conservative Democrats, but
this figure dropped to just 4% by 2012. However, we can also see that candidates who are not
quite as moderate as Snowe and Gordon but nevertheless more moderate than the party mean in
Congress still make up a sizeable proportion of primary winners. Furthermore, these numbers
have remained relatively stable over time, with candidates on the moderate side of the party in
Congress constituting 32% of primary winners in 1980 and 35% of winners in 2012. The higher
values in part reflect the victories of incumbents who are on the moderate side of the party mean,
but we can also see similar rates among non-incumbent candidates in open seat races. To be
sure, the actual value of the party mean has changed in important ways over time—with both
parties moving away from the center and toward the extremes—but it does not seem to be the
case that candidates who are on the moderate side of the party in Congress comprised a much
larger share of non-incumbent winners thirty years ago than they do today.
11
0
Percentage of Primary Winners
.1
.2
.3
.4
Figure 4: Ideological Moderates as a Proportion of House Primary Winners, 1980-2012
1980
1984
1988
1992
1996
2000
2004
2008
2012
Olympia Snowes and Bart Gordons
More Moderate than Party in Congress
More Moderate than Party in Congress, Open Seats
Source: America Votes; Federal Election Commission; Bonica (2014).
Thus, while ideological moderates have consistently made up a minority of primary
winners, these data provide us with ample opportunity for an analysis of electoral success across
relatively moderate U.S. House candidates. Of the 12,770 candidates who won their primary
from 1980 to 2012, 4,242, or 33%, were on the moderate side of the party in Congress. Only
1,311 primary winners (10%) were as moderate as Snowe and Gordon during this time period,
but this is still a sizeable number of candidates. Lastly, there were 1,418 non-incumbent
candidates who won in open seats during this time period, and 428 (30%) were more moderate
than the party in Congress.
I use a logistic regression to examine the relationship between candidate ideology and
primary election outcomes. The dependent variable is coded one if the candidate won the
12
primary and zero if she lost. The main independent variable—ideological moderate—is trickier
to measure. On the one hand, we are interested in the relationship between ideological centrism
and electoral outcomes, but what we are really interested in is how party primaries exacerbate the
gap between the two parties in Congress. Ideological centrism can be measured in a variety of
ways, such as the distance from the most extreme candidate in the primary (Hall and Snyder
2015) or simply as the left-right positions of candidates (e.g., Rogowski and Langella 2014). I
use a different measurement here in light of the broader concern with party change in Congress.
Candidates are simply coded as one if they are more moderate than the party mean in Congress
and zero if they are more extreme than the party mean.5 The strength of this measurement of
ideology is its direct connection to the congressional party and its implication for party change.
If candidates on the moderate side of their party win, the ideological gulf between the parties will
diminish; if candidates on the extreme side of their party win, the gulf will widen.
I include an interaction between ideological moderate and the total number of candidates
in the primary to examine how candidate ideology matters as the number of candidates on the
ballot increases.6 Again, the main expectation is that moderates are more likely to win the
primary as the number of candidates increases. Primary voters may have a harder time
distinguishing among candidates, and the information may be too costly to obtain. I do not sort
through the various mechanisms here. As discussed above, it is also possible that the number of
other candidates who are moderates or ideologues influences whether or not a moderate
5
I also used a continuous measure of ideology, with higher values indicating increasing distance from the
party mean in the moderate direction and lower values indicating increasing distance from the party in the
extreme direction. The findings are the same as those presented here. While the degree of moderatism
and extremism certainly matters for party change, I use the dichotomous measure because electing any
candidate on the moderate side of the party is a first step toward minimizing polarization.
6
I ran the models with both the number of total candidates in the primary as well as the number of
primary candidates who have Bonica ideology scores. The results are similar, but I include the number of
candidates with Bonica scores here because it is more taxing for voters to choose among numerous
quality candidates than it is to choose among numerous long-shot candidates.
13
candidate prevails. It may be that an increasing number of extreme competitors helps moderates
because primary voters will divide their votes among ideologues. Similarly, we might expect
moderates to be even less likely to win as the number of fellow moderate candidates increases. I
explore both of these possibilities in more detail below.
Several control variables are included in the models. First and foremost, sitting members
of Congress generally sail to primary election victory, and when applicable, I include a dummy
variable for incumbents. In addition, the number of primary candidates is expected to be
negatively associated with the probability of winning the primary. House candidates who raise
more money are also expected to be more likely to win, and contributions are obtained from
Bonica’s (2014) dataset and measured as logged values of total campaign receipts. I account for
candidates’ own party presidential vote share in the congressional district, and the models
include an interaction between own party presidential vote share and ideological moderate. We
might expect moderates to be less likely to win in districts that would be more favorable to
candidates at the extremes. Lastly, I include a dummy variable for Republican candidates, and
all of the models include state and year fixed effects.
The results are presented in Tables 1 and 2 below; each will be discussed in turn. I use a
few models to examine the relationship between candidate ideology and electoral success. The
baseline model is provided in Column 1. With respect to the control variables, the results are
largely consistent with the expectations above. Incumbency has a huge effect on primary and
general election outcomes, and it is very rare for incumbents to lose in the primary. The
disproportionate retirement of moderates is therefore especially detrimental for patterns of
congressional representation given that incumbents of all stripes are so much more likely to win
than non-incumbents (Thomsen 2014b). Unsurprisingly, candidates who raise more money are
14
also more likely to win the primary. In addition, as the number of primary candidates increases
and as the district becomes more favorable to their party, the probability of winning is expected
to decrease. However, the interaction between candidate ideology and the presidential vote share
of the district is not significant, which is consistent with the notion that congressional elections
have become increasingly nationalized.
Table 1: The Effect of Candidate Ideology on Primary Election Victory, 1980-2012
(1)
Ideological Moderate
(2)
(3)
-1.22**
(0.22)
-0.87**
(0.03)
___
-1.41**
-1.52**
(0.23)
(0.35)
Number of Primary Candidates
-0.90**
-0.39**
(0.03)
(0.03)
Ideological Moderate x
0.12**
0.10*
Number of Primary Candidates
(0.04)
(0.05)
Incumbent
3.22**
3.25**
___
(0.14)
(0.13)
Log of Mean Receipts Raised
0.38**
0.38**
1.02**
(0.02)
(0.01)
(0.03)
Own Party Presidential Vote Share
-4.24**
-4.19**
-4.88**
(0.27)
(0.26)
(0.37)
Ideological Moderate x
0.76
0.56
0.67
Own Party Presidential Vote Share
(0.43)
(0.43)
(0.66)
Republican
0.08
0.09
-0.01
(0.05)
(0.05)
(0.07)
Constant
-0.66
-0.63
-9.82**
(0.46)
(0.51)
(0.69)
Number of Observations
17,338
17,338
6,639
Log-Likelihood
-5254.13
-5250.37
-2960.04
Note: Entries are logistic regression coefficients with robust standard errors in parentheses.
All models include state and year fixed effects. **=p<0.01, *=p<0.05.
In terms of the main variables of interest, across models, moderates are less likely to win
the primary than those at the extremes, which is consistent with the conventional wisdom as well
as the survey results in the previous section (see Hall and Snyder 2015 for a similar finding). In
the full sample of candidates in Column 1, the predicted probability of winning the primary is
15
0.76 for ideologues, compared to 0.58 for moderates.7 In Column 2, however, we can see that
the interaction between the number of candidates and ideological moderate is positive and
significant, indicating that moderates are more likely to win as the number of primary candidates
increases. In fact, moderates and ideologues are statistically indistinguishable in their probability
of victory when the number of primary candidates is greater than five. Yet perhaps more
importantly, we know that primary victories are dramatically unequal in how much they
contribute to party change in Congress. Incumbents rarely face primary challenges, and even
when they do, members of Congress do not alter their behavior much as a result (Boatright 2013;
see also Poole and Rosenthal 2007). Increases in polarization are thus mainly occurring through
member replacement processes (Theriault 2006). In short, it is the election of the new guard, not
the re-election of the old guard, that is spurring party change in Congress.
In Column 3, the sample is restricted to non-incumbents in races where there is more than
one candidate; the rest of the section focuses on these results given the disproportionate impact
of non-incumbents on party change in Congress. The predicted probability of primary victory is
again lower, on average, for moderates than it is for ideologues (0.28 and 0.14, respectively), but
the positive and significant interaction term is of greater concern here. Figure 5 presents the
marginal effect of being an ideological moderate on the probability of winning as the number of
primary candidates increases. When there are two candidates, being an ideological moderate is
expected to decrease the probability of winning by nearly 20 percentage points (from 39.7% to
20.0%).8 As the number of candidates increases, however, the negative effect of being a
moderate diminishes. When there are four and six primary candidates, being a moderate is
7
All variables are set at their mean or mode so these values are for non-incumbents. The values of course
vary dramatically depending on candidate type. For example, the probability that moderate incumbents
win the primary is 0.97, compared to 0.99 for ideologues.
8
These values are calculated for Republican candidates, but because the results are virtually the same for
Republicans and Democrats, party-specific results are not provided here.
16
predicted to decrease the probability of winning the primary by 11 and 5 percentage points,
respectively. In fact, candidate ideology does not matter much at all when there are very large
numbers of candidates in the primary, though this is certainly the exception in terms of
frequency. (There are 214 non-incumbent candidates in this sample, or 3% of the total, who ran
in races with at least 8 candidates.) Yet the main point is that while being a moderate is for the
most part a barrier to electoral success in contemporary party primaries, it appears to be less of a
barrier as the number of primary competitors increases.
Figure 5: The Marginal Effect of Being a Moderate Across Increasing Numbers of Primary
Candidates, 1980-2012
Note: Values are estimated from the model in Column 3 in Table 1.
As noted above, we are also interested in how the ideology of other primary candidates
influences the electoral prospects of moderates, particularly with respect to whether they are
17
moderates or ideologues themselves. Table 2 presents the results with the measures of the
number of primary candidates that are more moderate and more extreme than the party in
Congress. The findings resemble those in Table 1. Across models, moderates are less likely to
win the primary, and as the number of primary competitors increases—moderate and extreme
alike—the probability of winning of course decreases as well.
Table 2: The Effect of Other Candidates’ Ideology on Primary Election Victory, 1980-2012
(1)
Ideological Moderate
(2)
(3)
-1.48**
(0.23)
___
-1.61**
-1.62**
(0.23)
(0.35)
Ideological Moderate x
-0.13
-0.03
Number of Extreme Candidates
(0.07)
(0.07)
Ideological Moderate x
___
0.27**
0.21**
Number of Moderate Candidates
(0.06)
(0.07)
Number of Extreme Candidates
-0.95**
-0.91**
-0.38**
(0.03)
(0.03)
(0.03)
Number of Moderate Candidates
-0.73**
-0.87**
-0.42**
(0.03)
(0.05)
(0.05)
Incumbent
3.22**
3.24**
___
(0.13)
(0.13)
Log of Mean Receipts Raised
0.38**
0.38**
1.02**
(0.01)
(0.01)
(0.03)
Own Party Presidential Vote Share
-4.16**
-4.13**
-4.86**
(0.26)
(0.26)
(0.37)
Ideological Moderate x
0.74
0.71
0.80
Own Party Presidential Vote Share
(0.43)
(0.44)
(0.66)
Republican
0.10*
0.12*
0.02
(0.05)
(0.05)
(0.07)
Constant
-0.56
-0.56
-9.79**
(0.51)
(0.51)
(0.69)
Number of Observations
17,338
17,338
6,639
Log-Likelihood
-5242.40
-5231.03
-2955.38
Note: Entries are logistic regression coefficients with robust standard errors in parentheses.
All models include state and year fixed effects. **=p<0.01, *=p<0.05.
Yet somewhat surprisingly, when we look at the interaction terms in Columns 2 and 3, it
appears that moderate candidates are actually more likely to win the primary as the number of
18
moderate candidates increases. Moderates are less likely to win when there are fewer numbers
of moderate candidates, but this effect diminishes as the number of moderate candidates
increases. In the sample of non-incumbents in Column 3, a one standard deviation increase in
the number of moderate candidates results in a 2.6 percentage point decrease in the probability of
winning the primary for ideological moderates (from 12.0% to 9.4%). By comparison, for
ideologues, a similar increase in the number of moderates leads to a 9.8 percentage point
decrease (from 27.8% to 18%). However, when there are four or more moderate candidates, the
probability of winning the primary is not statistically different for moderates and ideologues
(these values are 10.4% and 7.1%, respectively, and the confidence intervals overlap). While it
makes sense that moderates cannot win if they do not run, it is not clear why a greater number of
moderates would be beneficial to the electoral prospects of moderates. This result does not seem
to conform to the vote-splitting logic outlined above. Similarly, we might have expected
moderates to be more likely to win as the number of ideologues increases, but this relationship is
neither positive nor statistically significant. Regardless of the specific mechanism, one key
problem for the future of polarization is that there are so few races with sizeable numbers of
moderates; of the 6,600 candidates in this sample of non-incumbents, only 371, or 5.6%, ran in
races with at least four or more moderate candidates.
In sum, moderates do appear to have a harder time winning the primary than those at the
extremes. While a variety of factors such as incumbency, campaign receipts, and district
partisanship affect the probability of primary victory, it is clear that candidate ideology plays a
role as well. Yet what is perhaps most interesting is that the relationship between candidate
ideology and primary victory also depends on the particular configuration of candidates on the
ballot. In particular, I find that moderates are more likely to win as the number of candidates in
19
the primary increases. In other words, moderates may suffer less than many observers would
expect, at least when there is a greater number of candidates on the ballot. I do not sort through
the various mechanisms here, but it is possible that in these cases, it becomes too costly for
primary voters to obtain information on the various competitors.
Conclusion
The survey and observational data collectively suggest that primary voters are more
likely to favor ideologues over moderates. The preference for ideologues is particularly
pronounced in the survey data when primary voters were given full information on candidate
positions. It is clear that very liberal Republicans and very conservative Democrats are unlikely
to prevail in the current polarized environment, and the findings support the general narrative
about the attitudes and beliefs of primary voters. Nevertheless, we can leverage newly available
ideology data to examine the conditions under which candidate ideology may matter more or less
for electoral outcomes and gain insight into when moderates might be more likely to win than
they would otherwise. This paper is a first step toward that goal.
The broader message of the paper is that although primary voters may prefer ideologues
to moderates, an ideologue may not always win the primary. Candidates are not chosen in
isolation, and the collective arrangement of candidates matters for who is ultimately elected to
office. More generally, scholars have paid less attention to the makeup of choices that voters
face, but the composition of candidates has clear implications for the composition of Congress.
For those who bemoan the rise of partisan polarization in Congress, these findings should be
seen as good news. Candidate ideology does not always outweigh all other variables, and
ideological centrism does not influence electoral outcomes in all circumstances. Primary voters
20
may be more likely to select moderates or ideologues depending on the options that are presented
to them. Yet for reformers who wish to diminish the ideological gulf between the two parties,
the very first step is to encourage more moderates to run for office. Because regardless of the
configuration of choices on the ballot, in order for a moderate candidate to get elected, there
must be a moderate for voters to choose.
21
References
Abramowitz, Alan I. 2010. The Disappearing Center: Engaged Citizens, Polarization, and
American Democracy. New Haven, CT: Yale University Press.
Ahler, Douglas J., Jack Citrin, and Gabriel S. Lenz. 2014. “Do Top-Two Primaries Improve
Congruence? An Experimental Test of California’s Top-Two Primary.” Working Paper,
University of California, Berkeley.
Aranson, Peter H., and Peter C. Ordeshook. 1972. “Spatial Strategies for Sequential Elections.”
In Probability Models of Collective Decision Making, eds. Richard Niemi and Herbert
Weisberg. Columbus, OH: Merrill.
Bafumi, Joseph, and Michael C. Herron. 2010. “Leapfrog Representation and Extremism: A
Study of American Voters and Their Members in Congress.” American Political Science
Review 104 (3): 519–42.
Clinton, Joshua D. “Representation in Congress: Constituents and Roll Calls in the 106th
House.” Journal of Politics 68(2): 397-409.
Coleman, James. 1971. “Internal Processes Governing Party Positions in Elections.” Public
Choice 11(1): 35-60.
Coleman, James. 1972. “The Positions of Political Parties in Elections.” In Probability Models of
Collective Decision Making, eds. Richard Niemi and Herbert Weisberg. Columbus, OH:
Merrill.
Bonica, Adam. 2014. “Mapping the Ideological Marketplace.” American Journal of Political
Science 58(2): 367-387.
Brady, David W., Hahrie Han, and Jeremy C. Pope. 2007. “Primary Elections and Candidate
Ideology: Out of Step with the Primary Electorate?” Legislative Studies Quarterly 32(1):
79-105.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Fiorina, Morris P., Samuel J. Abrams, and Jeremy C. Pope. 2006. Culture War? The Myth of a
Polarized America. New York: Pearson/Longman.
Geer, John. 1988. “Assessing the Representativeness of Electorates in Presidential Primaries.”
American Journal of Political Science 32(4) 929-945.
Gerber, Elisabeth R. and Rebecca B. Morton. 1998. “Primary Election Systems and
Representation.” Journal of Law, Economics, and Organization 14(2): 304-324.
22
Hall, Andrew B. and James Snyder. 2015. “Candidate Ideology and Electoral Success.” Working
Paper, Harvard University.
Hirano, Shigeo, Gabriel S. Lenz, Maksim Pinkovskiy, and James M. Snyder, Jr. 2015. “Voter
Learning in State Primary Elections.” American Journal of Political Science 59(1):
91-108.
Hirano, Shigeo, James M. Snyder, Jr., Stephen Ansolabehere, and John Mark Hansen. 2010.
Primary Elections and Partisan Polarization in Congress. Quarterly Journal of Political
Science 5: 169-191.
Kousser, Thad, Justin Phillips, and Boris Shor. 2013. “Reform and Representation: Assessing
California’s Top-Two Primary and Redistricting Commission.” Presented at the annual
meeting of the Midwest Political Science Association.
Jacobson, Gary C. 2004. The Politics of Congressional Elections. New York: Longman.
Layman, Geoffrey C. and Thomas M. Carsey. 2002. “Party Polarization and ‘Conflict Extension’
in the American Electorate.” American Journal of Political Science 46(4): 786-802.
Layman, Geoffrey C., Thomas M. Carsey, John C. Green, Richard Herrera, and Rosalyn
Cooperman. 2010. “Activists and Conflict Extension in American Party Politics.”
American Political Science Review 104(2): 324-346.
Mann, Thomas E., and Norman J. Ornstein. 2012. It’s Even Worse Than It Looks: How the
American Constitutional System Collided With the New Politics of Extremism. New
York: Basic Books.
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance
of Ideology and Unequal Riches. Cambridge: Massachusetts Institute of Technology
Press.
McGhee, Eric, Seth Masket, Boris Shor, Steve Rogers, and Nolan McCarty. 2014. “A Primary
Cause of Partisanship? Nomination Systems and Legislator Ideology.” American Journal
of Political Science 58(2): 337-351.
Norrander, Barbara. 1989. “Ideological Representativeness of Presidential Primary Voters.”
American Journal of Political Science 33: 570-587.
Owen, Guillermo and Bernard Grofman. 2006. “Two-Stage Electoral Competition in Two-Party
Contests: Persistent Divergence of Party Positions.” Social Choice and Welfare 26:
547-569.
Rogowski, Jon C. and Stephanie Langella. Forthcoming. “Primary Systems and Candidate
Ideology: Evidence from Federal and State Legislative Elections.” American Politics
Research.
23
Scammon, Richard M., Alice V. McGillivray, and Rhodes Cook. 1990-2006. America Votes 19
27: A Handbook of Contemporary American Election Statistics. Washington, DC: CQ
Press.
Sides, John and Lynn Vavreck. 2013. “On the Representativeness of Primary Electorates.” Paper
presented at the conference “Political Representation: Fifty Years after Miller and
Stokes,” Vanderbilt University, March 2013.
Theriault, Sean M. 2008. Party Polarization in Congress. New York: Cambridge University
Press.
Thomsen, Danielle M. 2014a. “Ideological Moderates Won’t Run: How Party Fit Matters for
Partisan Polarization in Congress.” Journal of Politics 76(3): 786-797.
Thomsen, Danielle M. 2014b. Party Fit in the U.S. Congress: The Intersection of Ideology,
Political Parties, and Gender. Doctoral Dissertation, Cornell University.
24
Appendix A: Survey of Primary Voters
Liberal and Moderate Democrat
Conservative and Moderate Republican
25
Appendix A: Survey of Primary Voters
Vote Choice Question
Note: When primary voters placed the mouse over the candidate names (highlighted in blue),
they received the same information about the ideology of each candidate as in the earlier slides.
26