Content uploaded by Isaac Hale
Author content
All content in this area was uploaded by Isaac Hale on Aug 16, 2018
Content may be subject to copyright.
UC Berkeley
California Journal of Politics and Policy
Title
Some General Consequences of California's Top-Two Primary System
Permalink
https://escholarship.org/uc/item/5zt6b4r1
Journal
California Journal of Politics and Policy, 8(2)
ISSN
1944-4370
Authors
Highton, Benjamin
Huckfeldt, Robert
Hale, Isaac
Publication Date
2016-01-01
DOI
10.5070/P2cjpp8230564
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
1
Some General Election Consequences of
California’s Top-Two Primary System
Benjamin Highton
University of California, Davis
Robert Huckfeldt
University of California, Davis
Isaac Hale
University of California, Davis
Approved by the electorate in 2010, California’s Proposition 14 amended the state constitu-
tion and adopted the top-two primary system for congressional, statewide, and state legislative
elections. Instead of separate party primary elections for a given office, the amendment mandat-
ed that a single primary election open to all registered voters be held. The top two vote getters,
irrespective of party affiliation, would advance to the general election.
Advocates for the reform argued that the new system would produce more competitive con-
tests and result in more ideologically moderate elected officials. Yet, initial research suggests
that these outcomes have mostly not taken place (Ahler et al. 2014, 2015; Kousser et al. 2014;
Kousser 2015; Masket 2013; McGhee 2015). In this paper we focus on one clear consequence of
the top-two primary system, general elections that are contested between candidates of the same
political party.
Under the old system, same party general elections could not occur. Under the new system
they can, and do, take place. About one in six state legislative and congressional general elec-
tions in the 2012 and 2014 election cycles have pitted a Democrat against a Democrat or a Re-
publican against a Republican. The purpose of this paper is to analyze this phenomenon and its
consequences for electoral competitiveness and polarized voting.
Background
It is not difficult to develop intuitions about why the adoption of the top-two primary might
have important electoral and institutional effects. For example, under the previous, “closed” sys-
tem, a moderate Democrat candidate might have had a more difficult time advancing to the gen-
eral election against a more traditional or even liberal Democrat in a primary election limited to
registered Democrats. A single all-inclusive primary offers a moderate Democrat more potential
supporters, especially because registrants with no party preference are permitted to vote. The
same is true for a moderate Republican candidate.
Whatever one’s intuitions, previous theoretical research on primary rules suggests that their
consequences are hard to predict and are unlikely to have consistent and uniform effects. For in-
stance, open party primaries not restricted to party registrants can produce more ideologically
2
extreme outcomes than closed primary systems (Chen and Yang 2002; Cho and Kang 2015;
Cooper and Munger 2000; Oak 2006). The reasons have to do with the complicated decisions of
potential candidates under different institutional rules and the interplay between voters and can-
didates that takes place.
Empirically, previous research reports uneven, and often modest, effects of primary rules.
The most thorough analysis is McGhee et al. (2014), which analyzes the relationship between the
degree of primary openness and the ideological extremism of legislators. Relying on more exten-
sive data than previous studies and addressing a variety of important measurement issues,
McGhee et al. (2014) reports minimal effects of primary rules on ideological extremism. To be
sure, it is possible that general patterns do not apply to California as its short-lived experience
with a version of an open primary system—a “blanket primary”—appears to have been more
consequential (Alvarez and Sinclair 2012; Bullock and Clinton 2011).
The top-two primary system was used to select general election candidates in 2012 and 2014.
Keeping in mind that campaigns, candidates, and voters have had little time to adjust to the re-
form (Sinclair 2015a, 2) and there may be a long period of adjustment before the state arrives at
a new, potentially more moderate equilibrium (McGhee 2010, 12), a handful of initial studies
report modest, at best, effects. Ahler et al. (2014) reports that in an experiment comparing the old
and new ballot forms, moderate candidates were not advantaged with the top-two primary ballot.
Kousser et al. (2014) compares legislators elected under the new rules in 2012 to those elected
under the old rules in 2010 and finds virtually no change in the congruence between the ideolog-
ical locations of legislators and voters in 2012. McGhee (2015) analyzes the legislative score-
cards issued by the California Chamber of Commerce and finds “fairly mixed” evidence for an
effect of the top-two primary system.
Among studies that do find effects, the significance appears modest. Kousser (2015) analyzes
statewide contests in 2014 and finds some effects on campaign strategies and candidate entry de-
cisions but no ultimate effect on general election outcomes. Grose (2014) finds more ideological
moderation in state legislative roll call voting among Democrats (but not Republicans) in 2013
compared to 2011, but the biggest shift—among Democrats in the state senate—may have been
due to factors unrelated to the top-two.1
The one area where the effects may have been more substantial has to do with general elec-
tions between candidates of the same party. In an analysis of an assembly general election be-
tween two Republicans, Sinclair (2015b) explains how the more moderate of the two (Frank Bi-
gelow) beat the conservative (Rico Oller). Sinclair and Wray (2015) analyze Google Trends data
for state legislators seeking reelection and find that the “single greatest predictor of search vol-
ume is whether the legislator faced a co-partisan challenger” (10). Nagler (2015) analyzes turn-
out in same party general elections and finds that partisans of the excluded party had a higher
rolloff (abstention) rate, which could limit the electoral advantage of the more moderate candi-
date of the other party.
1 “In 2013, Republican senators almost universally represent districts with very low percentages of
Democratic voters—and most Republicans in the senate in 2013 were not newly elected in 2012. In con-
trast, Democratic senators (post-2013) represent a wider range of districts and a significant number were
newly elected in 2012. Some of the newly elected Democratic senators represent districts in which the
Democrats do not have a substantial electoral edge. This means that the mean Democratic senator ideolo-
gy may have shifted due to a few key pickups in districts where constituents are not particularly partisan.
Thus, Democratic senators moderated while Republican senators became more conservative” (Grose 2014,
14).
3
General elections with candidates of the same party are a clear consequence of the adoption
of the top-two primary system. Under the old system, they could not take place. Our purpose in
the remainder of this paper is to analyze their occurrence and a variety of electoral consequences.
Our findings offer a host of new insights.
Data
We focus on California’s U.S. House and state legislative (assembly and senate) elections
from 2002 through 2014. Elections in the 2002–2010 period were held under the old primary
system and serve as a useful baseline. With 173 districts (53 house, 80 assembly, and 40 senate)
and seven elections years, there are 1,071 elections for analysis.2 Much of the data we rely on are
publicly available from the California Secretary of State’s website (http://www.sos.ca.gov
/elections). Most of our analyses consider the party registration advantage in a district. To create
this measure, we relied on the last report of registration before each election and computed the
Democratic and Republican percentages of all registrants. The party registration advantage is
simply the absolute value of the difference. Higher values indicate one of the parties has a larger
advantage.
We also use the 2012 and 2014 Cooperative Congressional Election Study (CCES) common
content surveys, made publicly available through Dataverse (http://projects.iq.harvard.edu/cces
/data). Because of their large national sample sizes there are sufficient observations to compute
district-level estimates of voting behavior in congressional (U.S. House) elections. Across the
two surveys for California congressional districts, there is an average of about 100 respondents
per district per election year and 60 voters. Specifically, we focus on ideological and partisan
voting polarization. Relying on the traditional questions tapping ideological and partisan identifi-
cation we compute and analyze the differences in preferences between liberals and conservatives
(ideological polarized voting) and Democrats and Republicans (partisan polarized voting).
Results
Table 1 shows the distribution of party registration advantage for the two electoral periods. In
both the first (2002–2010) and second (2012–2014) periods there are only a small number of dis-
tricts that would be judged highly competitive based on the balance of party registration. In the
first period just five percent of districts had a party advantage of five percentage points or less,
and in the second period the figure was nine percent. In contrast, in almost half the districts in
both periods (44 percent and 47 percent, respectively), one of the major parties enjoyed a regis-
tration advantage of at least 20 percentage points. The possibility for the top-two to bring elec-
toral competition where it otherwise would not take place follows from the observation that there
are so many districts in California dominated by one party. If primary elections in these districts
produce candidates of the same party, instead of observing a general election with the candidate
from the party favored by the registration advantage winning easily, candidates of the same party
may find themselves in close, competitive elections.
2 Although there are 40 senate seats, there are 140 (not 280) senate elections to analyze over the seven
election cycles because only half the seats are up each year due to the four-year term length for senators
compared to the two-year terms for the house and assembly.
4
Table 1. The Distribution of Party Registration Advantage
Party registration advantage (%) 2002–2010 2012–2014 Total
0–5 5 9 6
5–10 12 16 13
10–20 40 29 37
20–30 21 25 22
30–40 10 9 9
40+ 13 13 13
N 765 306 1,071
Note: Party registration advantage is the absolute value of the difference between the Democratic and
Republican shares of all registrants in a district based on the final report of registration in advance of the
general election. Cell entries report the percentage of U.S. House, state Assembly, and state Senate elec-
tions with the indicated levels of party advantage for each period and overall.
Overall, 17 percent of congressional and state legislative general elections in the 2012 and
2014 election cycles have been between candidates of the same party. This type of election is
much more likely to occur in districts where one party has a sizable registration advantage. When
party registration advantage was less than five percent, only one of 26 elections involved two
candidates from the same party.3 In contrast, when party registration advantage was more than 40
percent, 16 of 40 elections involved candidates from the same party.
Table 2 shows the full relationship between party registration advantage and type of general
election. As one party gains a registration advantage, general elections between a Democrat and
Republican decline from 96 percent (registration advantage of less than five percentage points)
to 43 percent (registration advantage over 40 points). A greater party registration advantage is
associated with more same party general elections, though the increase is not linear. Where the
party advantage ranged from five to 20 percentage points, 12 percent were same party elections.
In the party advantage rage of 20 to 40 points, 20 percent of elections were between candidates
of the same party.
Uncontested general elections (those where only one candidate entered the primary) and
those between a major party candidate and a minor party candidate were more common in dis-
tricts with a greater party registration advantage, increasing from zero percent where the registra-
tion advantage was less than 5 points to 18 percent where the advantage exceeded 40 points.
For comparison, Table 2 shows the relationship between general election type and party reg-
istration in the 2002–2010 period when the top-two primary system was not in effect. Consistent
with the more recent data, the rate at which major party candidates faced no competition or com-
petition from a third-party candidate were higher in districts with more lopsided party balances.
About one in four elections (26 percent) were uncontested or between a major party candidate
3 In CD-25 in 2014, the Republican Steven Knight beat the Republican Tony Strickland. In terms of
party registration, there were nearly identical numbers of Democrats and Republicans (37.24 percent and
37.19 percent, respectively), producing a party registration advantage score of 0.05 percentage points.
5
Table 2. Party Registration Advantage and the Same-Party General Election Candidates
Party registration 2012–2014 2002–2010
advantage (%) D-R D-D or R-R Other D-R Other
0–5 96 4 0 97 3
5–10 84 14 2 96 4
10–20 84 10 6 92 8
20–30 75 20 5 93 7
30–40 67 19 15 84 16
40+ 43 40 18 74 26
Total 76 17 7 90 10
Note: Party registration advantage is the absolute value of the difference between the Democratic and
Republican share of all registrants in a district based on the final report of registration in advance of the
general election. Cell entries report the percentage of each type of election for the given periods. D-R is a
general election between a Democrat and Republican. D-D/R-R is a general election featuring two candi-
dates of the same party. Other includes all remaining elections, typically a major party candidate against a
minor party candidate or a major party candidate in an uncontested election.
and a minor party candidate in districts with a party registration advantage over 40 points com-
pared to just three percent in districts where the partisan balance was nearly even.
Because same party general elections are more likely to occur where one party has a strong
registration advantage, it is likely—or at least plausible—that same party general elections pro-
duce or heighten electoral competition where there otherwise would be none or less. To investi-
gate this possibility, we use a common measure of electoral competitiveness—whether the win-
ning candidate received less than 60 percent of the vote (a competitive election) or more than 60
percent of the vote (an uncompetitive election). Based on this measure, 62 percent of the same
party general elections were competitive compared to just 38 percent of the elections between a
Democrat and Republican.4
The higher rate of competitive elections with candidates from the same party is noteworthy in
part because those elections are more likely to take place in districts traditionally associated with
low rates of competitiveness. Consider the 2002–2010 period. The first column of entries in Ta-
4 There was only one race between a major party candidate and a nonmajor party candidate that was
competitive. In CD 33 in 2012 Democrat Henry Waxman beat Bill Bloomfield who ran as an independent
with 54 percent of the vote. In the 2002–2010 period only 18 percent of the elections between a Democrat
and Republican were competitive.
6
Table 3. Election Competitiveness by Party Registration Advantage and Election Type
Party Registration 2002–2010 2012–2014 2012–2014
Advantage (%) D-R D-R D-D or R-R
0–5 77 96 100
(35) (25) (1)
5–10 44 61 86
(90) (41) (7)
10–20 20 40 56
(282) (75) (9)
20–30 3 16 60
(148) (56) (15)
30–40 0 0 100
(61) (18) (5)
40+ 0 0 44
(72) (17) (16)
Total 18 38 62
(688) (232) (53)
Notes: Cell entries report the percentage of elections that were competitive—the winner received less
than 60 percent of the vote—in the designated categories. The numbers of elections on which the percent-
ages are based are in parentheses.
ble 3 shows the relationship between party registration advantage and electoral competitiveness
in races where a Democrat faced a Republican.5 A strong negative relationship is evident. In the
districts most closely divided between the parties, 77 percent of elections were competitive.
Where the party advantage was 20–30 points just 3 percent were competitive and there were no
competitive elections out of the 133 elections held where the party registration advantage ex-
ceeded 30 points.
As shown in the second column of entries in Table 3, a similar, though stronger, pattern is
evident for the 2012–2014 period. Where the party registration advantage was smaller, there
were even more competitive elections in the 2012–2014 period compared to the 2002–2010 peri-
od. For example, where the party advantage was less than 5 points, 96 percent (24 of 25) of elec-
tions were competitive in 2012–2014 compared to 77 percent in 2002–2010. But, in the districts
where the party advantage was greatest there were no competitive elections in 2012–2014, as
was the case in 2002–2010.
The clear exceptions are same party elections. At every level of party registration advantage
same-party elections are more competitive than two-party elections. Given the smaller number of
elections, the percentage differences can be misleading, but even so it is clear that where there
was little or no competition in the 2002–2010 and 2012–2014 periods with two-party elections,
same-party elections were more competitive. Twenty-one of the 36 same-party elections (58 per-
5 Uncontested elections and those between a major party candidate and a minor party candidate are
excluded from Table 3.
7
cent) held in districts where the party advantage was 20 points or more, were competitive, com-
pared to just 10 and 1 percent competitive elections in two-party contests in the 2012–2014 and
2002–2010 periods, respectively.
More competitive elections are typically associated with higher turnout. On the one hand, be-
cause same-party elections are more likely to be close and have higher spending levels, they
might also have higher turnout. On the other hand, given that Nagler (2015) found that partisans
of the excluded party were more likely to abstain in same-party elections, it is possible that
same-party elections might be a unique instance where greater competitiveness brings lower
turnout.
To investigate the turnout effects of same-party elections, we focus on “voter rolloff.”
Rolloff is the difference in turnout at the top of ballot and turnout for a lower ballot contest. For
example, in 2014 the gubernatorial election was the top of the ticket election. In the 47th assem-
bly district there was a same-party general election between two Democrats (Cheryl Brown and
Gil Navarro). In the assembly election 23.5 percent of registered voters in the district voted while
25.6 percent of registered voters in the district voted in the gubernatorial election. Rolloff, then,
was -2.1 percentage points (23.5 - 25.6 = -2.1) in that election.
We focus on rolloff rates rather than actual turnout rates for a variety of reasons. First, by
comparing turnout within districts (but across elections) we control for all factors that systemati-
cally vary across districts. Second, because gubernatorial elections are held in midterm election
years, there are never coinciding presidential and gubernatorial elections, which means there is
always a high profile (presidential or gubernatorial) election at the top of the ticket that drives
turnout decisions for many registrants.
Consistent with Nagler’s (2015) individual-level results, we find substantial rolloff in same-
party elections. In the 53 same-party elections the average rolloff was 7.2 percentage points
compared to 2.8 percentage points in two-party elections, for a rolloff difference of 4.4 percent-
age points. Because previous research shows that rolloff tends to be greater in presidential years
and in elections that are less competitive, we estimated a model of rolloff that includes these fac-
tors. The first set of entries in Table 4 are estimates from a simple OLS regression based on the
2012-2014 elections that includes dummy variables for same-party elections and other elections
(major party/minor party and uncontested elections), treating two-party elections as the baseline.
As shown in the table, rolloff is 4.4 points greater (more negative) in same-party elections, and
larger than the rolloff differential in other elections.
The estimates in the second column of Table 4 are based on a model that also includes a
dummy variable distinguishing presidential (2012) from gubernatorial (2014) election years, the
degree of party registration advantage in districts, and a set of indicators for how competitive the
final general election outcome was. As shown in the table, rolloff was 4.2 percentage points
greater in the presidential year of 2012 compared to the gubernatorial year of 2014 and there was
also notably more rolloff (6.1 percentage points) in elections where the winner received more
than 90 percent of the vote. At the same time, the estimated effect of a same-party election on
rolloff remains at 4.4 points. Thus, while same-party elections are associated with closer elec-
tions, they also appear associated with greater rates of turnout rolloff.
The last variable we consider is voting polarization. Given the large and growing ideological
differences between the parties and the fact that the California Legislature is the most polarized
of all state legislatures (Shor and McCarty 2011), we expect polarized voting along partisan and
ideological lines in general elections between a Democrat and a Republican. But, in same-party
elections, the relationship between partisanship, ideology, and voting may be much weaker. With
8
Table 4. Rolloff in General Elections
Independent variable 2012–2014 2012–2014 2002–2010
Election type
D-R (baseline) ‒ ‒ ‒
D-D or R-R -4.4** -4.4** N/A
(0.4) (0.3)
Other -3.2** -1.6** -3.1**
(0.7) (0.5) (0.4)
Presidential year -4.2** -4.1**
(0.2) (0.2)
Party registration advantage .004 -.01
(.01) (.01)
Electoral competitiveness (winner’s vote)
< 60% (baseline) ‒ ‒
60-90% -0.3 -0.3
(0.3) (0.2)
>90% -6.1** -7.6**
(0.9) (0.6)
Constant -2.8** -0.6** -1.9**
(0.2) (0.2) (0.2)
Adjusted R2 .27 .69 .70
S.E.E. 2.8 1.9 2.2
N 306 306 765
Notes: In presidential years, rolloff is defined as the difference between turnout in a district election
(U.S. House and state assembly) and presidential turnout in a district. In gubernatorial election years
rolloff is the difference between turnout in a district election and gubernatorial turnout in a district. High-
er rates of rolloff are indicated by negative values of greater magnitude. See text for details.
* indicates p<.10; ** indicates p<.05.
candidates of the same party, there is no obvious role that voters’ partisan loyalties could play.
And to the extent that candidates of the same party are ideologically similar, voters’ ideological
preferences may not be much related to voting either.
To investigate the question of voting polarization we turn to the 2012 and 2014 Cooperative
Congressional Election Study common content surveys, described earlier. These surveys include
congressional district identifiers (but not state legislative district identifiers), and there is an av-
erage of about 60 voters per district per election year. For each district in each year we compute
ideological and partisan voting polarization. Ideological voting polarization is the difference in
voting between self-identified liberals and conservatives. Partisan voting polarization is the dif-
ference between Democratic and Republican party identifiers. If everyone in one group votes for
one candidate and everyone in another group votes for the other candidate, then the voting polar-
ization score is 100. If the two groups vote for the candidates at identical rates, the polarization
score is 0.
9
Table 5. Voting Polarization in California Elections for the House of Representatives
Type of General Election
D-R D-D or R-R
Ideological voting polarization 76 59
Partisan voting polarization 86 55
Notes: Cell entries report the median levels of voting polarization in different and same-party general
elections for California congressional elections in 2012 and 2014. Ideological voting polarization is the
difference in vote percentages between self-identified liberals and conservatives. Partisan voting polariza-
tion in the difference in the vote percentages between self-identified Democrats and Republicans.
Table 5 reports median polarization scores for the two types of elections. First, consider Cali-
fornia congressional elections where a Democrat runs against a Republican. In those races there
is substantial ideological and partisan voting polarization. The median level of ideological voting
polarization is 76 percentage points while the median level of partisan voting polarization is even
higher, 86 percentage points. In the same-party general elections, the levels of polarized voting
are notably smaller, but still substantial. Ideological voting polarization is reduced to 59 percent-
age points while partisan voting polarization is reduced to 55 percentage points. Thus, even when
candidates of the same party face off in general elections, voters’ ideological and partisan prefer-
ences remain related to their ballot choices, a point we return to below.
Conclusion
Most people agree that the adoption of the top-two primary system in California represented
a potentially significant reform with possibly far-reaching consequences. Yet, analyses of its ef-
fects based on the first two election cycles during which it has been used to select general elec-
tion candidates have typically not reported substantial effects. To be sure, minimal initial effects
may not imply lack of longer-term effects, as “California’s elected officials, party elites, prospec-
tive candidates, and voters are in the process of charting out the contours of a new electoral re-
gime, and it may take several years to reach some new equilibrium” (Masket 2013, 188). At the
same time, what is notable about the first two election cycles is the occurrence of a nontrivial
number of general elections with candidates of the same party.
Because same-party elections occur disproportionately in districts where one party has a sub-
stantial party registration advantage, an effect of the top-two primary is to enhance the im-
portance of the general election relative to the primary. In a typical and only marginally competi-
tive California legislative district, where one party is clearly dominant, the traditional party pri-
mary system selects two candidates—one representing the dominant party and the other repre-
senting the minority party, with third parties depending on petition drives to place a candidate on
the ballot. In this context, the general election outcome is nearly certain. The dominant party
candidate typically wins, and the minority party typically loses. By default, the main purpose of
the minority party primary is to choose the loser at the general election. The top-two primary of-
fers a solution to this problem by allowing for two candidates of the same party to advance to the
10
general election. As we have shown, this is a common phenomenon and is associated with
heightened electoral competition where ballot choices remain strongly associated with voters’
ideological and partisan preferences.
Going forward, there are a variety of issues to be addressed. First, while occurring more
commonly in districts where one party is dominant, some notable same-party elections have tak-
en place elsewhere. For example, in a district with a modest Democratic Party registration ad-
vantage of six percentage points, two Republicans advanced to the general election in the 31st
House District in 2012 because too many Democrats entered the primary election and therefore
deprived the party of even having a chance of winning the seat. One focus of future research
should be to understand better the tactics the state parties employ to try and avoid scenarios like
this and to control the process, perhaps with their endorsement powers (Kousser et al. 2015).
Second, in districts where one party has a substantial party registration advantage, why do
some primaries produce two candidates of the same party, but others do not? In this paper, we
have not investigated the correlates of same-party general elections beyond the degree of party
registration advantage. But it seems likely that other factors are at play, too, and the possibilities
should be investigated.
Third, our finding that substantial ideological and partisan voting polarization is evident in
same-party general elections is intriguing. It is possible that the appearance of partisan voting
polarization is merely the result of the fact that party and ideological identification are closely
related in contemporary American politics. But even if that is the case, the question remains as to
why we observe ideological voting polarization in same-party general elections. One possible
explanation is that voters perceive and act on ideological differences between the candidates,
even when of the same party. Moreover, given that same-party elections occur most commonly
in districts that strongly favor one of the two parties, it is also possible that in these elections the
more ideologically extreme candidate may be advantaged, which could help explain the general
lack of a moderating influence of the top-two primary reported in previous research. At this point,
these are conjectures, but they strike us worthy propositions for future empirical tests.
In conclusion, the parties, candidates, potential candidates, and political strategists are clearly
still adapting to the new system, and adaptation requires time. Indeed, after initially resisting the
introduction of the direct primary in the early 20th century, political parties successfully adjusted
to its place in American electoral machinery. Hence it is too early to tell what the ultimate con-
sequences of the top-two primary system will be. At the moment, though, it can be said that Cali-
fornia is serving the classic function assigned to every state as a laboratory of democracy, exper-
imenting with an alternative mechanism of democratic control.
11
References
Ahler, Douglas J., Jack Citrin, and Gabriel S. Lenz. 2014. “Do Open Primaries Help Moderate
Candidates? An Experimental Test of California’s 2012 Top-Two Primary.” Department of
Political Science, University of California, Berkeley.
———. 2015. “Why Voters May Have Failed to Reward Proximate Candidates in the 2012 Top
Two Primary.” California Journal of Politics and Policy 7 (1): 1–14.
Alvarez, R. Michael, and Betsy Sinclair. 2012. “Electoral Institutions and Legislative Behavior:
The Effects of Primary Processes.” Political Research Quarterly 65 (3): 544–57.
Bullock, Will, and Joshua D. Clinton. 2011. “More a Molehill than a Mountain: The Effects of
the Blanket Primary on Elected Officials’ Behavior from California.” Journal of Politics 73
(1): 1–16.
Chen, Kong-Pin, and Sheng-Zhang Yang. 2002. “Strategic Voting in Open Primaries.” Public
Choice 112 (1/2): 1–30.
Cho, Seok-ju, and Insun Kang. 2015. “Open Primaries and Crossover Voting.” Journal of Theo-
retical Politics 27 (3): 351–79.
Cooper, Alexandra, and Michael C. Munger. 2000. “The (Un)Predictability of Primaries with
Many Candidates: Simulation Evidence.” Public Choice 103 (3/4): 337–55.
Grose, Christian R. 2014. The Adoption of Electoral Reforms and Ideological Change in the Cal-
ifornia State Legislature. Los Angeles, CA: USC Schwarzenegger Institute.
Kousser, Thad. 2015. “The Top-Two, Take Two: Did Changing the Rules Change the Game in
Statewide Contests?” California Journal of Politics and Policy 7 (1): 1–17.
Kousser, Thad, Scott Lucas, Seth Masket, and Eric McGhee. 2015. “Kingmakers or Cheerlead-
ers? Party Power and the Causal Effects of Endorsements.” Political Research Quarterly 68
(3): 443–56.
Kousser, Thad, Justin Phillips, and Boris Shor. 2014. “Reform and Representation: A New
Method Applied to Recent Electoral Changes.” Social Science Research Network. <http://
ssrn.com/abstract=2260083>.
Masket, Seth. 2013. “Polarization Interrupted? California’s Experiment with the Top-Two Pri-
mary.” In Governing California: Politics, Government, and Public Policy in the Golden State,
ed. E. Rarick. Berkeley, CA: Berkeley Public Policy Press.
McGhee, Eric. 2010. Open Primaries. San Francisco, CA: Public Policy Institute of California.
———. 2015. “California’s Top Two Primary and the Business Agenda.” California Journal of
Politics and Policy 7 (1): 1–16.
McGhee, Eric, Seth Masket, Boris Shor, Steven Rogers, and Nolan McCarty. 2014. “A Primary
Cause of Partisanship? Nomination Systems and Legislator Ideology.” American Journal of
Political Science 58 (2): 337–51.
Nagler, Jonathan. 2015. “Voter Behavior in California’s Top Two Primary.” California Journal
of Politics and Policy 7 (1): 1–14.
Oak, Mandar P. 2006. “On the Role of Primary Systems in Candidate Selection.” Economics and
Politics 18 (2): 169–90.
Shor, Boris, and Nolan McCarty. 2011. “The Ideological Mapping of American Legislatures.”
American Political Science Review 105 (3): 530–51.
Sinclair, Betsy. 2015a. “Introduction: The California Top Two Primary.” California Journal of
Politics and Policy 7 (1): 1–6.
Sinclair, Betsy, and Michael Wray. 2015. “Googling the Top Two: Information Search in Cali-
12
fornia’s Top Two Primary.” California Journal of Politics and Policy 7 (1): 1–12.
Sinclair, J. Andrew. 2015b. “Winning from the Center: Frank Bigelow and California’s Nonpar-
tisan Primary.” California Journal of Politics and Policy 7 (1): 1–33.
