Can California’s New Primary Reduce Polarization? Maybe Not.

This is a guest post from Doug Ahler, Jack Citrin, and Gabriel Lenz.  For previous posts on the negligible relationship between primary rules and political polarization see here, here, and here.


According to many pundits and scholars, closed primary elections are a major contributor to the ideological polarization in Congress and state legislatures. By partitioning voters into two ideologically-sorted electorates, they argue, closed primaries incentivize candidates to adopt the positions of voters in their party rather than of their constituency as a whole. As a result, they elect representatives who consistently toe the party line and resist compromise. Advocates of reform, from academics like Morris Fiorina to practitioners like Arnold Schwarzenegger, therefore argue that replacing closed party primaries with a more open nominating process will reduce polarization and its offspring—gridlock and a noxious political atmosphere—by helping moderate candidates.

Are these claims about the consequences of reform valid? To shed further light on the consequences of this reform, we conducted a statewide experiment before California’s June 2012 primaries. As a result of a popular referendum, these elections replaced the closed party primaries with an open ballot that presented voters, regardless of party, with the same list of candidates. The top two vote getters advanced to the general election, thus allowing for a choice between two contenders from the same party. In a study sponsored by the Institute of Government Studies at UC Berkeley, we randomly assigned 2839 registered voters in U.S. House districts where moderate candidates faced more extreme candidates to one of two conditions for electoral choice: (1) the open ballot that would be used in the actual 2012 primary, or (2) the closed primary ballot that earlier elections employed.

For advocates of the reform, the results of this survey experiment are disappointing. If the open ballot did indeed help moderate candidates, they should have won more votes in the open-ballot condition than in the closed-ballot condition. But as shown in the scatterplot below, we find no such evidence: Moderate candidates for the House of Representatives fared no better under the top-two primary than they would have in closed party primaries. The vertical axis plots how much better (or worse) candidates performed among participants randomly assigned to the top-two ballot, while the horizontal axis plots candidate moderateness on a seven-point scale. The results fail to show the upward-sloping trend that advocates of primary reform argue we should see.


Why did the top-two reform fail to achieve its goals? While voters are generally quite moderate and were willing to cast crossover votes (roughly 12% of our participants who voted for a major party candidate did so), they largely failed to discern ideological differences between extreme and moderate candidates of the same party, particularly if they were challengers. In addition to asking about vote choices, we asked respondents to rate the candidates in their districts on a 7-point scale of political ideology. To obtain an objective measure of a candidate’s ideology, we visited her website and placed her on the same 7-point scale of liberalism-conservatism, and also hired 204 workers on Amazon’s Mechanical Turk to do the same rating task. The figures below plot the averages of respondents’ ratings for each candidate against the average of those generated by the “MTurkers.” The first graph includes Democratic candidates and the second Republican candidates.  The lack of correlation between two ratings is striking, as is the reluctance of respondents to place candidates at the extremes of the scale.


Of particular interest in the second graph—which includes only Republican candidates—are the respondent placements for District 24. Abel Maldonado is a well-known moderate politician in California, a former Lieutenant Governor and State Senator, and most importantly, the author and principal advocate of the Top-Two Primaries Act. His potential constituents rated him at roughly 5.25 on the 7-point scale. However, they gave almost the same rating to his fellow GOP challenger Chris Mitchum, a little-known actor and Tea Party candidate who the local GOP endorsed because of Maldonado’s violation of the pledge not to vote for raising taxes. So while Maldonado appears to have benefited from the open primary in the first graph in this post, this was not because voters in his district were attuned to the ideological differences between the two Republicans.

It is worth noting that the top-two primary was implemented in all statewide elections and not just the congressional races described above. We conducted a similar analysis for races for California’s State Senate, the upper house in the nation’s most polarized state legislature. Again, we find no evidence that the open primary helped moderate candidates. And while the open ballot failed to help moderate candidates in House races because voters struggled to identify moderate candidates, it failed in these down-ticket races because few voters even tried to locate the centrists: “Don’t know” accounted for a large majority of ideological ratings given to State Senate candidates. As a possible consequence of this lack of knowledge, voters appear to have relied much more heavily on partisanship in voting for State Senate. Just 5.6% of respondents cast a vote for a candidate from a party other than their own in these races.

While this research implies that open primaries are not the cure for polarization its advocates hoped for, its limitations should be mentioned. Open primaries may still moderate the behavior of elected officials even if voters fail to recognize or explicitly reward such moderation (Bullock and Clinton offer evidence that this generally is not the case.) Alternatively, experience with the new rules may cause both voters and candidates to adapt and gravitate toward the center. This may be especially true in California, where some believe that in the long-run primary reform in conjunction with nonpartisan redistricting will produce a less polarized legislature. Open primaries may also succeed in higher salience races, such as a gubernatorial contest, where voters have easier access to information about candidate ideology. Finally, there may be a search for moderation in the general election in cases where two candidates from the same party compete. Future work should address these possibilities. The consequences of primary reform are multifaceted and complex. But in California, simply changing the rules did not appear to change the likely outcomes.

19 Responses to Can California’s New Primary Reduce Polarization? Maybe Not.

  1. David desJardins March 27, 2013 at 5:02 pm #

    The goal seems off here. Primary reform is beneficial because it gives more voters a voice and a stake in the outcome, even if they don’t happen to belong to the dominant party. It’s not for politicians or reformers to decide whether we should have “more moderate” candidates; that’s up to the voters.

  2. Chad Peace March 27, 2013 at 5:09 pm #

    Agree with David. There is a fundamental misconception that the non-partisan primary is about moderation. It’s not. It’s about being accountable to the electorate. A “moderate” by statistical analysis has little to do with accountability.

    Voters want authenticity. And its hard to measure authenticity by statistical analysis. We should try though.

  3. Pepe March 27, 2013 at 5:37 pm #

    David and Chad,

    From the perspective of voters, I find it wrong for non partisans to vote in primaries. There is a fundamental contradiction between membership in an organization and having non members decide who should represent it. If you want accontability, use general elections; if you don’t like the candidates form your own party.

  4. David desJardins March 27, 2013 at 5:48 pm #

    Pepe, the California top-two primary isn’t about deciding whom should represent any party. It’s about advancing the two most popular candidates to the general election. Of those, there could be two, one, or zero of them from any given party. If you want to have an election to decide whom should represent your party, with voting only by party members, you’re free to do that. Just run it privately, yourself, using party funds. Then you can declare the winner to be the official representative of your party. But there’s no reason the state government should do that for you, at public expense.

  5. Pepe March 28, 2013 at 8:21 am #

    Of course there is. Electoral laws in the US and states make parties the agents of representation. Once they serve this public function, the state has an obligation to fund their activities. I don’t think fiscal conservatism should get in the way of democratic representation.

    • Dave Ely March 28, 2013 at 12:50 pm #

      Except, Pepe, that the point of the top two primary was to change those electoral laws, and remove the parties as direct agents of representation. That is why it is different from the earlier open primary laws.

      I think that another important goal is being overlooked. The choice is not just between moderation and partisanship in ideology. It is also between obstruction and governance in pursuit of whatever ideology might exist. The California Republican party had a policy supporting primary challenges to any legislator who decided to pursue their policy goals by making deals which included taboo tax changes. The partisan primary allowed them to enforce an obstructionist strategy. It is far too soon to say whether the non-partisan primary is effective in this regard, and it certainly isn’t measured by this study.

      In addition there is the problem of perverse effects of third party votes in a winner take all general election. A candidate who wins a third party primary and then competes in a winner take all general election suffers from a very rational fear among voters that voting for that candidate only results in the election of the less preferred major party candidate. The same candidate competing in a top two primary does not have this handicap, and has a much greater chance of winning in the general election with only one opponent. Over time this should allow the development of a much wider range of options, as opposed to picking a different point between two party extremes.

      • Drew March 28, 2013 at 1:33 pm #

        Third parties are much worse off under top two. First, in the preliminary round, the “spoiler” effect is still very real, creating incentives not to vote for them. Then, in the general election (which has more than double the turnout of the preliminary) the field is narrowed to only two, and it excluded every single third party candidate, which can have a deleterious effect on third party ballot access which may be tied to performance in general election races.

        Top two was billed as helping moderates and breaking up party polarization. This report suggests that it fails to do so. They also said it would promote competition, which it did not (even with independent redistricting the same year). Supporters can keep coming up with other things it might do and retreat to those, but on its stated goals it has been a failure.

        • Dave Ely March 28, 2013 at 2:18 pm #

          Third party ballot access is not really very relevant if there is no partisan primary at all. I am only aware of one minor party nominee to win a general election for any of the offices affected by the change in the last 30 years or so, and that person proceeded to join a major party. Let’s give it a few election cycles before deciding if the minor parties are worse off.

  6. Dave Ely March 28, 2013 at 2:03 pm #

    The discussion of Maldonado above gives weight to the idea that the top two primary reduced the ability of the local GOP to punish Maldonado for his “violation of the pledge”, even if he wasn’t perceived as more moderate ideologically.

    The first chart identifies a number of candidates where the voting structure made a large difference, and others where there is little or no difference. It might be instructive to give some analysis to the particulars of the contests with a large effect.

  7. David desJardins March 28, 2013 at 4:01 pm #

    Top-two is proving successful on the grounds on which I voted for it. It’s clear how it helps to enfranchise the voters who are in a district that’s overwhelmingly of a different party to their own. Whether it’s achieving some other goals that someone else might have wanted (like more moderate representatives, or helping third parties) is really immaterial to me. Third parties won zero seats in California legislature under the old system, and will continue to win zero seats under the new system; they aren’t any worse off.

    • John H March 28, 2013 at 5:18 pm #

      What part of the franchise is gained by being able to vote in a top-two rather than closed primary? The ability to influence the choice over the majority’s candidate?

      Frankly, the policy and partisan consequences of the reform matter a lot more than this dubious notion of enfranchisement.

      • David desJardins March 29, 2013 at 1:55 am #

        John, I think you are feigning incomprehension. In a district dominated by one party (like mine, or most of those in California), the winner of the primary of that party is guaranteed to win the general election. So only the members of that party have a voice in who represents them in the legislature. For all of the other voters, it’s irrelevant whom they vote for. In the top-two system, every voter in the district gets a vote in the final election that determines the elected representative (which often involves two members of the same party). That’s why it’s better.

        • John H March 29, 2013 at 10:18 pm #

          This point seriously misconstrues representation and democratic voice. Voters have voice in their capacity during the general election to vote for or against competing candidates. The mere fact that some voters frequently lose or do not participate in candidate selection does not diminish this fact. Your defense of the top-two rule seems to be based on the point that voters, and not parties, should decide who gets to run elections. Fair enough. But don’t argue to the straw man that the alternative (the party decides) disenfranchises voters.

          On the substance: The irony of the policy is that, rather than increasing ‘voice’ , it may very well do the opposite, by replacing lopsided districts with one-party enclaves (since credible candidates of the minority party will never get a chance to run head-to-head against the majority party candidate, and will adapt accordingly, i.e., change their party banner). And we know voters (esp. independent identifiers) have a hard time distinguishing the ideological differences amongst candidates of the same party, making it potentially easier for extremists to win. As a consequence, the top-two reform may both diminish democratic choice and increase polarization over time.

          • David desJardins April 4, 2013 at 2:07 am #

            I don’t care whether it increases or decreases “polarization”. That’s mostly up to the voters anyway. I do think it’s ludicrous to claim that a reform that increases the number of voters who have the opportunity to cast a meaningful vote is decreasing “democratic choice”. I am going to give up on taking you seriously. Probably shouldn’t have argued with anonymous poster in the first place.

  8. Chris March 29, 2013 at 3:44 am #

    The effect of “moderation” will probably only be felt in districts that are overwhelmingly Republican or overwhelmingly Democratic. In these districts, two candidates from one party will emerge and the more moderate of the two may be more likely to prevail, because the opposite party will play the role of electing the candidate. In races where the party split in the electorate is more even, then partisans on each side will come home during the general election. Thus, in the most Republican areas, the candidate may be more moderate, while in slightly less Republican areas, the candidate may be more extreme.

  9. Jim Riley March 30, 2013 at 1:30 pm #

    The study results don’t make sense.

    If I understand the methodology, if I were Joe Random voter from Vacaville (in CD 3), and Jane Random voter from Fairfield (also in CD 3) were selected for the poll we would be provided with one of three ballots, two partisan and one open primary ballot.

    If I were given a partisan ballot, I would receive a Democratic or Republican ballot based on my party affiliation. If I were a Democrat, the Democratic ballot would have the name of John Garamendi on it. If I were a Republican, I would be given a ballot with the names of the 4 Republican candidates.

    If I were given a Top 2 ballot, I would be given a ballot with all 5 candidates on it, including their party preference.

    Or am I completely misunderstanding what was done?

    The first chart shows that John Garamendi did 3% worse among voters given the open primary ballot. Reasonable enough – some Democrats would vote for one of the 4 Republicans if permitted to do so.

    But Republican Kim Vann did 14% worse, when in essence Garamendi was added to the Republican ballot.

    How could both Garamendi and Vann do worse?

  10. Dave Ely March 30, 2013 at 4:39 pm #

    You raise an interesting question, which might have a variety of answers, but it is impossible to answer with information included in the post. Here are some questions for the authors:
    What are the numerator and denominator for the percentages shown in the first chart? Given the fact that a number of candidates included were unopposed within their party, I’m guessing that the numerator is total votes for a given candidate and the denominator is the total votes cast for all candidates in the scenario, including those cast in a different party primary for the closed primary scenario. If this is correct, a number of other questions arise.
    How did you deal with non-partisan and minor party voters assigned to partisan primary scenario? In the most recent partisan primaries Democrats allowed non-partisans to vote in the Democratic primary, but Republicans did not allow non-partisan, and neither party allowed those registered with minor parties.
    Did you control for the partisan distribution of voters in the two pools for each election. In other words, did you randomly distribute each party to the two pools, or did you just randomly assign voters to the two pools and assume you would get a representative sample in each pool?
    You say districts were selected which included a moderate candidate and a more extreme candidate. I assume that you mean that at least one closed primary would include that matchup. Did you control in any way for the level of competitiveness or the perceived moderation of candidates from other parties?
    Why do you include other candidates, especially those unopposed in their own party on the first chart? Without a clear designation of the moderate and extreme opponents that you claim to be testing the others are just noise that could mask any effect. Why not chart how the candidates of interest performed relative to each other in the two scenarios?
    Did you control in any way for the partisan distribution of voters with the districts? In other words, what share of the total closed primary votes come from the party of the candidates of interest? If 80% of voters come from a their party, the open primary is not likely to have much effect, but if only 20% do then then effect could be much larger.
    You discuss the perceived ideology of the candidates, but you don’t say anything about the partisan distribution of perceptions. The most interesting question is if the average perception among the entire pool is the same as the perception among partisan. You say the perceptions of Maldonado and Mitchum were the same, but was the true for each partisan group, or just for the entire pool?
    I could go on. The number of alternatives here is large relative to the number of cases you are studying, so if you didn’t control for them your results are largely meaningless.
    Additionally I would say that by focusing on vote shares or numbers of votes you could be missing the boat. Why not look at what the resulting general election matchups would have been under the two scenarios. The relative performances of your pairs of moderate and extreme candidates are irrelevant if neither or both make it to the general election.

  11. Dave Ely March 30, 2013 at 6:48 pm #

    One additional comment on the subject of top-two that does not relate to this study, but is more on the order of “on the other hand”. Top-two also has the potential for perverse effects in a partisan electorate. Specifically when one party has two candidates and the other has 3 or more, there is a reasonable chance that the two candidates from one party will make it to the general, even if the other party has majority support.

  12. Doug Ahler March 31, 2013 at 1:43 pm #

    I’m glad to see our post has generated such interest and spurred discussion about the consequences of primary reform. Those who have questions about our methods, or anything else, may wish to read the entire paper. A draft recently prepared for the Western Political Science Association meeting can be found here:

    This paper should address questions about how the dependent variable was calculated, how participants were assigned to treatment/control, how districts were selected, etc. It also presents a number of alternative analyses and robustness checks. Finally, the paper should also convey a general sense of how we tried to balance clean identification of causal effects (internal validity) with being able to say something about the case of California (external validity).

    I’d like to emphasize that this was a randomized, controlled experiment. In expectation, our treatment and control samples should be balanced on all covariates. We did check for balance on a few key covariates, including ideology (because the difference between the closed primary rules and top-two rules could present a problem there). We found no significant differences between the groups. We do not include covariates in our regression models as a result. Following the argument of Freedman (2008), heavily modeling experimental data is not justified by randomization and may introduce bias.

    Finally, I’d like to engage the argument that moderate outcomes shouldn’t be the primary (bad pun fully intended) concern of the top-two reform. We aren’t trying to make a normative argument in this paper about the goal of electoral reform. Rather, we’re empirically addressing the oft-made claim that party primaries have been responsible for polarization, and that simply changing the electoral rules will decrease legislative polarization and its effects.

    Interestingly, though, we find little support for the more general claim that the top-two format should allow voters to choose candidates who are more ideologically proximate. In an analysis we recently conducted (noted in the draft, but will appear more prominently in the final version), we find that voters did not vote for more ideologically proximate candidates in the top-two condition. Voters actually chose slightly more proximate candidates in the closed condition, although results were nowhere close to statistically significant. We take these individual-level results as evidence that the non-effect of the top-two went beyond moderateness of outcomes. Of course, voters may be choosing on factors other than ideology — but the logic and framing of the top-two reform in both academic and political circles has always been very much about ideological proximity voting.

    We’re glad to see such interest in this topic. We ultimately believe that primary reform is a multifaceted topic with a number of separate inputs and outputs to be considered. Here, we’ve looked at the effect of electoral/ballot format on voters’ choices (and aggregated up to candidate performance), but there are many other things to think about. Knowledge has to be developed inductively. We look forward to continuing this dialogue with other scholars as California moves forward with the top-two and other states experiment with electoral reform.