---

Summary

Key Takeaways from the Research

• The evidence that the top-two system reduces polarization is mixed.
• Theoretically, one reason to expect that the top-two system might reduce polarization is that in second-round races involving two Democrats or two Republicans, voters who favor the party without a candidate (i.e., out-partisans) should tend to vote disproportionately for the more moderate of two competing candidates.
• Conditional on voting, out-partisans appear to vote for the more moderate candidate. Relatively moderate candidates appear more likely to win second-round elections when competing against a co-partisan.
• However, a large number of out-partisans do not vote in these second-round elections (i.e., they roll off).
• Existing evidence suggests that the modest overall effect of the top-two system might be due to the low number of second-round races in which both candidates are from the same party, the high degree of roll-off among out-partisans in second-round races, low levels of information among voters about candidates’ positions, and the electoral success of incumbents.

Important Questions We Need Better Answers To

• How does the top-two system affect the strategic behavior of elites, including candidates, campaign donors, activists, and party leaders? For example, do more moderate candidates run?
• What accounts for the high roll-off in second-round elections? Is it due to lack of information, affective polarization, and/or indifference?
• How does the top-two system affect outcomes, such as candidate quality, incumbent performance and accountability, issue-by-issue congruence, and descriptive representation?
• Theoretically, should we expect the “top-four” and “top-five” systems to operate similarly to the top-two system?

---

Introduction

Political observers have pointed to party primaries as a potential contributor to the increasing ideological polarization and deepening conflict between the two major political parties.

Since the introduction of the Australian ballot in the U.S., elections for major offices have almost all been partisan, with each party having a nominee on the general election ballot, but no more than one per elected position.

Parties have chosen their nominees in a variety of different ways: party committees, caucuses, conventions, or direct primary elections. Although the use of direct primaries in U.S. elections traces back to the 19th century, the widespread adoption of mandatory direct primaries occurred in the first half of the 20th century.¹ Progressive reformers promoted the increased participation of the electorate as a way to weaken the control of political party elites. Some scholars argue that direct primaries produced “more democracy” since they at least offer voters some choice in regions and contexts that otherwise lack significant general election competition.² There is some evidence that primaries give voters the opportunity to evaluate candidates based on characteristics that differ from the main dimensions of conflict between the political parties.³

Recently, political observers have pointed to party primaries as a potential contributor to the increasing ideological polarization and deepening conflict between the two major political parties. Some reformers propose replacing party primaries with “top-two primaries” or related systems. Under the top-two system, candidates from all parties compete against one another in the first-round election, and the top two vote-getters proceed to the second round regardless of their party affiliations. Thus, unlike the traditional electoral system, it is possible for both of the second-round candidates to be from the same party. The Democratic and Republican parties are not guaranteed to have nominees in the “general” election. Some reformers argue that the top-two system will lead to less polarization. This primer evaluates the logic and evidence surrounding this claim.

The top-two system is sometimes referred to as a “non-partisan” primary. “Non-partisan” is something of misnomer since party labels appear on the ballots. Election officials and many scholars reserve the term non-partisan for elections where party labels are not on the ballot. The system is also sometimes referred to as a “top-two primary.” This too is something of a misnomer because the system changes the structure of general elections as well as primaries. As discussed below, the possibility for two co-partisans to compete in the second-round “general” election is one of the most intuitive reasons to expect the system to reduce polarization.

Party Primaries and Polarization

To assess whether or not party primaries cause polarization, we need to compare the system to alternative nomination systems–after all, candidates must be nominated by some method. In fact, the evidence suggests that party primaries have had at most a small effect on polarization, especially when compared to the systems that these primaries replaced.

First, almost all states were using partisan primaries to nominate major party candidates for federal and state offices by 1920, but partisan polarization did not begin sharply increasing until about 1980. So it seems unlikely that the direct primary per se is the main cause of partisan polarization. Second, there is evidence that conventions and party committees are unlikely to reduce polarization relative to party primaries and might even make the situation worse.⁴ Since convention delegates and party activists tend to more ideologically extreme than even partisan voters, any system that gives these actors more influence over nominations is unlikely to favor moderates. Unless strategic considerations weigh especially heavily on their decisions–which is not necessary in safe constituencies, and based on the evidence does not seem to be an overriding concern even in competitive constituencies–these delegates and party activists can choose nominees who match their own ideological extremism.

Theoretically, we might expect the effect of primaries to vary depending upon who votes in them.⁵ In fact, there is considerable variation across states in who is allowed to vote in primaries. In some states, primaries are “closed,” that is only voters registered with a party may vote in that party’s primaries. In other states, primaries are open to unaffiliated voters, or parties may choose to open their primaries to unaffiliated voters. In states with “open” primaries, all voters may vote in either party’s primary, but in any given election they must choose one party. In “blanket” primaries, voters can even choose candidates from different parties across offices within an election. There is mixed evidence that allowing greater freedom for voters to participate in different partisan primaries reduces polarization. Some scholars find that moderates are somewhat less likely to be elected under closed primaries compared to more open primaries.⁶ Others find little evidence that the openness of the primary system has a significant effect on policy or ideological polarization.⁷ The evidence from surveys and polls is also mixed on the question of whether open primaries lead to more moderate primary electorates.⁸

Finally, some research has explored the relationship between turnout in party primaries and the election of ideologically extreme representatives. There is mixed evidence that turnout may be correlated with the ideological extremism of parties’ primary electorates– i.e., lower turnout is associated with a larger gap between the parties’ voters.⁹ In any case, primary election turnout does not appear to be correlated with the ideological extremism of those who are nominated and elected.¹⁰

---

Current discussions about reducing polarization through nomination reforms focus on eliminating party primaries altogether, in favor of nomination systems where all candidates compete together, and “winners” go on to the next round regardless of partisan affiliations.

The most prominent example is the top-two system, used in California since 2012 and Washington state since 2008. In this system, the two candidates with the highest vote shares in the initial round compete in the second round. Louisiana uses a closely related system in which the second-round election is held only if no candidates get majority support in the first-round. (This system is sometimes referred to as the Louisiana System or Jungle primary; some states, such as Georgia and Texas, employ this system for special elections.) As in California and Washington, the candidates in the second round can both be from the same party. Recently Alaska has begun experimenting with a variation, the top-four system with ranked-choice voting in the second round. Also, Nevada passed an initiative for a top-five system in 2022. Voters must pass it again in 2024 in order to change the constitution and allow the system to go into effect.

The intuition for why the top-two system favors moderates more than party primaries, and should therefore reduce polarization, has not been fully developed. Only one theoretical paper directly compares top-two systems with traditional partisan two-round systems.¹¹ The model in that paper suggests that the top-two system will produce moderate winners more often than the party primary. However, the model is highly stylized-for example, it restricts attention to the case in which there are exactly two potential candidates from each party. Much of the literature focuses on how voters will respond to the choices presented in the first- and second-round elections. Scholars have paid less attention to how candidates and potential candidates, and other elites such as campaign donors, might respond. One recent exception investigates the behavior of campaign donors and argues that these donors might interfere with the moderating forces of the top-two system.¹²

One argument is similar to the justification given for adopting open party primaries: that is, allowing all voters to participate in the first round might increase the share of voters who are moderate and therefore likely to favor moderate candidates. Also, similar to blanket primaries, where voters are not constrained to vote only for one party’s candidates, moderate partisan voters are free to choose a moderate candidate from an opposing political party in the first round. That is, a moderate Democratic voter could support the nomination of moderate Republican candidate for some offices while still supporting the nomination of Democratic candidates for other offices.

Other arguments follow from the unique potential for the top-two primary to produce situations where candidates from the same party to compete in the second-round “general” election.¹³ In a simple case when a moderate Democrat competes against a more extreme Democrat in the second round, the moderate candidate’s position should be relatively more attractive to the Republican voters than the relatively extreme candidate’s position. An analogous logic holds for second-round races between two Republicans. Thus, if partisan voters support moderate candidates from opposing parties, then this should increase the likelihood that the moderate wins the election.

This logic merits close attention especially because second-round elections involving two candidates from the same party are most likely to occur in districts where one party has a significant electoral advantage. Under party primaries, candidates from an advantaged party in “safe” districts do not have the same incentives to take moderate positions to compete in the general election. Similarly, in these districts, voters do not need to engage in strategic moderation when choosing among candidates. On the other hand, in relatively competitive districts, forward-looking voters have an incentive to choose moderates over extremists even under party primaries, in order to have a better chance of winning in the general election. Thus, in competitive districts, it is difficult to say which system provides the strongest incentives for moderation.

In districts where one party has a clear electoral advantage, the top-two system should provide stronger incentives for moderates to enter in the first round compared to traditional party primaries. Thus, the top-two system might reduce polarization either because it significantly increases the probability that any given moderate candidate wins and/or because it increases the number of moderate candidates present on the ballot.

The intuition above focuses on candidates’ ideological positions and relies heavily on a simplified model of voters’ decision-making. However, previous research has shown that when voters are choosing among candidates from the same party in primaries, or in non-partisan elections, they appear to weigh other factors – e.g., candidates’ performance in office, scandals, previous experience, specific policy issues, race and ethnicity and gender.¹⁴ In fact, voters appear to weigh these other factors more in primaries and non-partisan elections than they do in partisan general elections. Some scholars see this as a virtue of non-partisan electoral competition.¹⁵ Therefore, a more complete understanding of voter and candidate behavior under the top-two system probably requires a more comprehensive model that incorporates factors in addition to ideology.

Regarding turnout, there is at least one situation where the top-two system should have a positive effect. Consider a first round with two Democrats and one Republican on the ballot for the top office–e.g., governor. Under party primaries, we would expect turnout to be relatively high in the Democratic primary, but possibly quite low in the Republican primary (since there is no contest). Under the top-two system, however, Republicans have a strong incentive to turn out, to ensure that the Republican candidate receives enough votes to move on to the second round and avoid a second round with two Democrats.

Potential candidates face complex decisions about whether or not to run in the top-two system, which warrants further theoretical study.

Finally, existing theoretical discussions tend to ignore the strategic voting and coordination problems that voters face in the first round when more than two candidates run, with a few exceptions.¹⁶ Vote-splitting can lead to “perverse” outcomes. Suppose for example that four Democrats and two Republicans run in a Democratic-leaning constituency. If Democratic voters split their votes too evenly, then it is possible for the Republican candidates to place first and second, leading to a second round election with two Republicans and no Democratic candidates. For example, suppose all voters in a constituency are either Democrats or Republicans and let D be the share of voters in the district who are Democrats. Let d ≥ 2 be the number of Democratic candidates and r ≥ 2 be the number of Republican candidates. Assuming voters in both parties divide their votes equally across candidates from their respective parties, then the Republican candidates will defeat the Democratic candidates (and therefore place first and second) if and only if D < d/(d+r). This means that potential candidates face complex decisions about whether or not to run in the top-two system, which warrants further theoretical study.

---

Recent empirical studies argue that the top-two system in California and Washington reduced the overall level of partisan polarization among state and federal legislators elected in those states.

To define partisan polarization within a given legislature, researchers use roll-call based “ideological” scores, and compute the difference between the average (or median) position of Democratic and Republican legislators. One review of the literature on the effects of the California top-two system concluded that there is evidence of a moderate effect on reducing polarization.¹⁷ However, a similar review found little evidence for an effect on polarization.¹⁸

One study found no evidence that polarization decreased among California members of Congress between 2010 and 2012.¹⁹ However, another study of roll-call scores from 2003 through 2018 found evidence for moderation among members of Congress.²⁰ A third study found some evidence that Democrats in the California state legislature may have taken more moderate positions after the introduction of the top-two system, but no significant evidence of a similar pattern among California Republicans or legislators in Washington state.²¹

Assessing the top-two system is challenging in part because the estimated effects vary depending on the comparison group. Scatterplots of partisan polarization among state legislators in both California and Washington do not demonstrate a dramatic decline in polarization following the introduction of the top-two system. However, under the top-two system, polarization in the California State Assembly and the Washington State Senate have declined relative to the average change across other states. The patterns are less clear for the California State Senate and the Washington House of Representatives. It is still the case that the legislative chambers in California and Washington are among the most polarized in the country. Regarding Congress, the gap between Republicans and Democrats in California appears to have narrowed especially when compared to other states.

Louisiana’s top-two-like systems provides some additional evidence that these types of systems do not meaningfully reduce polarization. In the years prior to the introduction of the Louisiana System, the Louisiana congressional delegation was less extreme than the U.S. House as a whole (and even relative to other southern delegations). After adopting the top-two type system, the Louisiana delegation, like other congressional delegations from the South, moved in the extreme direction faster than the non-southern delegations. In recent years, Louisiana representatives appear more extreme than the average U.S. House member and about as extreme as other southern delegations.

So far, only three states have adopted a top-two, or similar, system. The evidence for significant change in polarization in these states is not overwhelming–sample sizes are small, and the estimates are sensitive to specification, comparison group, and time period. Therefore, we should hesitate to draw any strong conclusions about the hypothesis that top-two systems reduce polarization.

Second-round elections with both candidates from the same party should produce the strongest incentive to favor relatively moderate candidates. And there is some evidence in the existing literature that the system is working in these situations.²² However, voter partisanship and other non-ideological factors may limit the extent to which the system favors moderates even in these cases. In addition, most first-round elections do not lead to second-round elections with both candidates from the same party, even when the underlying partisanship of the voters is so lopsided that it could easily support such an outcome.

According to intuition above, moderates should be favored when voters are forced to choose among candidates from parties from the opposing end of the ideological spectrum. Evidence from individual level surveys through 2016 suggest that voters in congressional elections do behave in this manner.²³ The same is true for 2020. However, voters who do not identify with the party of the two candidates in the second round roll-off at higher rates than those who identify with that party.²⁴ This roll-off likely hurts moderates more than other candidates.

Similar to voting behavior in traditional primary elections, voters in the top-two system appear to give substantial weight to non-ideological factors, such as incumbency, when considering candidates from the same party. Among the first-round U.S. House elections in California and Washington up to 2022, incumbents came in first place overall 98% of the time and first place among co-partisans 99% of the time. In the second-round elections between candidates from the same party, incumbent U.S. House members defeated non-incumbents about 90% of the time; and no incumbents have lost to non-incumbents since 2016. Even if the second-round winners tend to be more moderate in incumbent-contested races, which appears to be the case, it is difficult to identify what is more salient to voters, incumbency or moderate positions. In two of the three cases where a challenger defeated an incumbent, the estimates of relative positions of the challenger and incumbent depend on the measure used. In the remaining case, the challenger was identified as relatively extreme.

In the small number of open-seat second-round races in California and Washington with U.S. House candidates from the same party, the relatively moderate candidates do appear to win more often–at least in the handful of cases where it was possible to classify the candidates’ relative ideological positions.²⁵ This is consistent with the logic for where the top-two system is most likely to favor moderate candidates. However, these open-seat second-round races are only a small fraction of the second-round races contested under the top-two system. With such a limited number of cases in the U.S. House, additional work at the state legislative level or with future elections needs to be done to check the robustness of these patterns.

Some empirical evidence suggests that voters have difficulty differentiating the ideological position of candidates who share the same party label, especially in lower profile races.

Finally, even if voters give substantial weight to candidates’ ideological positions in their voting decisions, some empirical evidence suggests that voters have difficulty differentiating the ideological position of candidates who share the same party label, especially in lower profile races. One study of the 2012 California congressional elections found that survey respondents’ placements of candidates correlate only weakly with measures of candidates’ actual locations within parties. Using a survey experiment, this study also finds that voters are no more likely at choosing ideologically closer candidates under the top-two system compared to the party primaries.²⁶ Other evidence suggests that this may be less of problem in high profile races–e.g., for governor and U.S. senator.²⁷

---

Conclusion

Does the top-two system lead to substantially different outcomes than the traditional primary and general election? Does it lead to the election of ideologically less extreme politicians and less polarization?

The evidence so far does not point to large effects that are easily discernible and consistently identified in the literature. Perhaps the best evidence comes from the relatively small number of cases in which two candidates from the same party compete in the second round. In these situations, moderates tend garner more votes, especially from voters who must cross party lines. Although many of these voters roll-off, those who vote, disproportionately support the more moderate candidates.

There is still a lot that we do not know about the functioning of this system. One has to do with understanding the reasons why some voters roll-off and others do not when a candidate from their party is not on the second-round ballot. Is this due to lack of information, or affective partisanship, or indifference? In the absence of cross-over voting in the second round, it is unclear how a top-two or even a top-four or top-five system would favor the election of moderate candidates over party primaries.

In general, more theoretical and empirical work needs to be done on how the top-two system affects elite behavior. How does it change the number and types of candidates who run? Do more moderates run? How does it affect the behavior of campaign donors, activists, and party leaders? One study found evidence that campaign donations increased significantly under the top-two system, but left open whether or not this significantly affected polarization.²⁸

More theoretical and empirical work needs to be done on how the top-two system affects elite behavior.

Another underexplored consideration is the effect of the top-two system on minor parties. These parties have overwhelmingly opposed the system, arguing that it hurts them because their candidates almost never move on to the second-round elections. Under the top-two system, independent and minor-party candidates have appeared in only about 4% of the second-round elections for Congress. Some evidence suggests that fewer minor party candidates compete even in the first round.²⁹

Finally, since so much of the recent literature has focused on the problem of polarization, we know relatively little–theoretically or empirically–about how the top-two system affects other things we care about. These include candidate quality, incumbent performance in office and electoral accountability, issue-by-issue congruence, descriptive representation, and inter-party competition.

---

References

---