Primary elections and the few who decide

Primary elections in the United States are the contests in which candidates are chosen to appear on the general election ballot, and in districts that reliably favor one party, the primary is often the election that actually decides who holds the seat. Primary turnout is far lower than general election turnout: the Bipartisan Policy Center estimated that about 21.3% of eligible voters took part in 2022 primaries,⁠¹⁴ while Pew Research Center counted about 45.1% turnout in the 2022 House general elections.⁠¹³ Researchers agree on that participation gap; they disagree on how much it explains polarization.

In plain words: in many places, the real choice happens months before November, and most people do not take part in it. The few who do may care more, know more, or lean further to one side than the average voter. Some scholars think this pushes politics toward the edges. Others have looked closely and found the effect is small. States have tried several changes to the rules, with uneven results.

This page defines the problem carefully, sets out the evidence, presents the main competing explanations, and states what an educational framework like My Vote Forward can and cannot contribute. Related pages: Polarization, Gerrymandering, Unequal participation, and Representation.

The problem, defined precisely

The primary election problem is a claim about who chooses representatives when one stage of a two-stage election carries almost all of the decision. It has three parts that should be kept apart: a fact about turnout, a fact about competition, and a disputed claim about consequences.

Defined terms.

• Primary election. A preliminary election that narrows the field of candidates before a general election. In most U.S. states it nominates one candidate per party.
• Closed primary. NCSL describes it this way: “a voter seeking to vote in a closed primary must be a registered party member”.⁠¹
• Open primary. A voter chooses which party’s primary to vote in, and in NCSL’s description “this decision is private and does not register the voter with that party”.⁠¹
• Multi-party or nonpartisan primary. A system in which, in NCSL’s words, “all candidates, regardless of party, are listed on a single ballot,” and the top finishers advance.⁠¹ California and Washington use top-two, Alaska uses top-four, and Louisiana uses what NCSL calls an “all-comers system (sometimes referred to as a ‘jungle primary’)”.⁠³
• Safe seat. A district in which one party’s nominee is very likely to win the general election, so the general election rarely changes the outcome.
• Decisive electorate. The set of people whose votes actually determine who takes office. In a safe seat, that is the winning party’s primary electorate, not the general electorate.

What the problem is not. It is not a claim that primaries are illegitimate, that primary voters are uninformed, or that one party’s primaries behave like the other’s. It is not a claim that low turnout is caused by any single rule. And it is not a claim that the general election is meaningless everywhere; in competitive districts it plainly decides the outcome.

Scale and evidence on primary turnout and safe seats

Primary turnout and the share of seats decided in primaries are measured with different denominators, and the numbers below should be read with those denominators in mind.

Primary turnout. The Bipartisan Policy Center’s report “2022 Primary Turnout: Trends and Lessons for Boosting Participation,” as summarized by Unite America on March 8, 2023, found that “Just 21.3% of all eligible voters participated in primary elections during the last cycle,” and that “The five states with the worst turnout in 2022” (Virginia, New York, Delaware, Connecticut, and Mississippi) “were all below 12%”.⁠¹⁴ (BPC’s own release could not be retrieved during this review, so the figure is cited through that summary.) For comparison, Pew Research Center’s count of official returns found “nearly 107.7 million valid votes were cast in the 2022 House elections,” representing “about 45.1% of the estimated voting-eligible population,” down from 48.1% in 2018 and up from 34.4% in 2014.⁠¹³ The primary electorate, in other words, is roughly half the size of the general electorate in a midterm year.

Who votes in primaries within a party. Pew matched 1,575 Republican and Republican-leaning registered voters on its panel to the national voter file and found that only 25% of Republicans verified as voting in the 2012 general election also had a record of voting in that year’s GOP primaries or caucuses.⁠⁵ The same analysis found that 36% of GOP primary voters were consistently conservative on a ten-item values scale, compared with 21% of Republicans who voted only in the general election.⁠⁵ It also found large areas of agreement: 84% of primary voters and 82% of general-only voters called government “wasteful and inefficient”.⁠⁵ This is one party, one year, and a panel sample, so it shows a difference in degree rather than a different species of voter.

Seats decided in primaries. Unite America, an organization that advocates for primary reform, states that “In 2022, 83% of the U.S. House was elected by just 8% of Americans,” counting 359 of 435 seats as “effectively decided months before the general election”.⁠⁴ Its denominator is the voting-age population (about 257 million), not registered or eligible voters, and the page does not publish its method for classifying a seat as decided in the primary.⁠⁴ The same page reports a different 2022 figure elsewhere (377 seats and 9%), which suggests the classification is sensitive to definitions.⁠⁴ The direction of the finding is consistent with Pew’s separate count that only 35 House seats in 2022 were noncompetitive or lightly competitive between parties in the narrow sense of having weak or no opposition, and that 16 districts listed only one candidate.⁠¹³ Those two numbers measure different things: Pew counted seats with almost no general election contest at all, while Unite America counted seats where the outcome was predictable. Readers should treat the 83% figure as an advocacy estimate with an undisclosed method, not as an official statistic.

Rules vary by state. NCSL’s table, updated May 19, 2026, sorts states into closed, partially closed, open to unaffiliated voters, partially open, open, and multi-party systems.⁠¹ By the page’s lists, 15 states use open primaries and 5 use a multi-party format, while the remainder restrict participation to some degree.⁠¹ Washington has used top-two since 2008, California since 2010, and Alaska has used top-four since 2020.⁠²

Causes and competing explanations for primary-driven polarization

The primary polarization hypothesis holds that because primary voters are fewer and more ideological, candidates who must first win a primary take more extreme positions. The evidence for that hypothesis is contested, and four explanations compete.

Explanation 1: the primary electorate pulls candidates outward. This is the reformers’ view. Grose summarizes the starting point: “Partisan polarization in Congress is at its highest levels in more than 100 years,” and “Political reformers advocate for changes to electoral institutions in order to reduce legislative ideological extremity”.⁠⁸ The Pew data above are consistent with a modest ideological gap between primary and general-only voters in one party.⁠⁵

Explanation 2: primary rules make little difference. McGhee, Masket, Shor, Rogers, and McCarty assembled a new dataset of state legislator ideal points and a detailed accounting of primary systems. Their finding: “the openness of a primary election has little, if any, effect on the extremism of the politicians it produces”.⁠⁷ If that is right, changing who may vote in a primary will not by itself change who wins.

Explanation 3: some rules do matter, especially top-two. Grose, using changes in primary systems “due to ballot proposition or in response to court cases” and legislator ideology from 2003 to 2018, reports that “The top-two primary has reduced ideological extremity among legislators, relative to those elected in closed primary systems,” and that legislators elected in open primaries “also show some evidence of attenuated extremity”.⁠⁸ This finding conflicts with McGhee and colleagues in part because it uses a different period, different legislatures (Congress rather than state legislatures), and a different identification strategy. The two studies are not necessarily contradictory; they may be measuring different institutions.

Explanation 4: general elections already filter extremists, where they are competitive. Hall studied U.S. House primaries from 1980 to 2010 with a regression discontinuity design. When an extremist narrowly wins a primary over a moderate, “the party’s general-election vote share decreases on average by approximately 9–13 percentage points and the probability that the party wins the seat decreases by 35–54 percentage points”.⁠⁶ Hall concludes that “general-election voters act as a moderating filter in response to primary nominations”.⁠⁶ The implication cuts both ways: the filter works where the general election is competitive, so the fewer competitive seats there are, the less the filter operates. This links the primary problem to Gerrymandering and to residential sorting.

Other causes outside the primary itself. Several factors that are not primary rules affect who runs and who wins: who chooses to run at all, donor networks (see Money in politics), national media attention, and geographic sorting. A complete causal account would need all of these, and no single study has one.

Who is harmed by low-turnout decisive primaries

The people most affected by low-turnout decisive primaries are those whose votes count only at the stage that no longer decides anything.

• Unaffiliated voters in closed primary states. NCSL notes that in some partially closed states, including Maryland, Oklahoma, and Oregon, both parties currently allow only party members to vote in their primaries.⁠¹ An unaffiliated voter in a safe district in such a state has no say in the decisive contest unless they register with a party.
• Voters of the minority party in a safe seat. Their general election vote is legally equal but rarely changes the result.
• Candidates with broad but shallow support. A candidate acceptable to most of the district can lose to one strongly preferred by a smaller, more active group.
• Parties themselves. Hall’s evidence suggests that where general elections are competitive, a party that nominates an extreme candidate often loses the seat.⁠⁶

Whether the general public is harmed in a deeper sense, through more polarized legislatures, depends on which of the explanations above is correct. That remains open.

What has been tried elsewhere: primary reforms and their results

Primary reform in the United States has been tried through ballot initiatives, legislation, and litigation, and the record includes court defeats, durable adoptions, and recent voter rejections.

California’s blanket primary (1996 to 2000), struck down. In 1996 California voters approved Proposition 198, which changed “the State’s partisan primary from a closed primary” to “a blanket primary, in which each voter’s ballot lists every candidate regardless of party affiliation”.⁠⁹ The U.S. Supreme Court held in 2000 that “California’s blanket primary violates a political party’s First Amendment right of association,” reasoning that it forced parties to “adulterate their candidate-selection process”.⁠⁹ The lesson: a primary that chooses a party’s nominee is constrained by the party’s own associational rights.

Washington’s top-two primary, upheld on its face. Washington’s Initiative 872 provided that “the two top votegetters for each office, regardless of party preference, advance to the general election”.⁠¹⁰ In 2008 the Supreme Court held it “facially constitutional” because, “Unlike California’s primary, however, the I–872 primary does not, by its terms, choose the parties’ nominees”.⁠¹⁰ Washington has used the system since 2008 and California adopted top-two in 2010.⁠²

Measured effects of top-two. Grose’s 2020 study is the strongest published evidence that top-two moderates legislators.⁠⁸ McGhee and colleagues’ broader finding that openness has little effect is the strongest caution.⁠⁷ Both are observational, and neither settles whether gains persist once candidates and parties adapt.

Alaska’s top-four with ranked-choice general elections. Alaska adopted top-four in 2020.⁠² In 2024 a ballot measure “that would repeal Alaska’s ranked choice voting and open primary system” failed by “664 out of 340,110 votes,” 50.1% to 49.9%, pending certification;⁠¹² after a recount concluded December 9, 2024, Ballotpedia reported the retention side ahead by 743 votes.⁠¹¹ The near tie is itself evidence that the reform has not produced settled public consent.

Rejections in 2024. Ballotpedia reported that in November 2024 voters defeated measures in Colorado (Proposition 131, top-four with ranked-choice voting), Idaho (Proposition 1, top-four with ranked-choice voting, with a 70.0% “no” vote), Nevada (Question 3, top-five with ranked-choice voting), Oregon (Measure 117), and Arizona (Proposition 140), and that Missouri voters approved Amendment 7, which includes a prohibition on ranked-choice voting, 68.4% to 31.6%.⁠¹¹ Some of these measures bundled primary reform with ranked-choice voting, so the results do not isolate voters’ views of either reform alone.

What the record suggests. Reforms that change who can vote in a primary have survived when they do not compel a party to accept outsiders in choosing its own nominee. Measured effects on polarization are modest and disputed. Voter support is uneven and in several states recently negative.

The strongest arguments across the political spectrum on primaries

Arguments about primary reform do not divide neatly by party. The strongest versions of the main positions are below, each stated as its proponents would state it.

The party-association argument. A political party is a voluntary association of people who share commitments. Its members, not outsiders, should choose who carries its name. The Supreme Court accepted a version of this argument in 2000.⁠⁹ On this view, the remedy for unrepresentative primaries is more members participating, not more nonmembers.

The open-access argument. Primaries are run by states, paid for with public money, and in safe districts they are the only meaningful election. Excluding unaffiliated voters from the decisive contest, on this view, is a public problem rather than a private party matter. Advocates point to turnout and seat counts such as Unite America’s.⁠⁴

The skeptical empiricist argument. Before changing rules, show that the rules cause the harm. McGhee and colleagues found little effect of primary openness on legislator extremity.⁠⁷ Rule changes have costs, including voter confusion and legal risk, and should be justified by evidence rather than intuition.

The accountability argument. Primary challenges are one of the few tools voters have to discipline incumbents in safe seats. Weakening party primaries could protect incumbents from their own voters. This argument is made by people across the spectrum who value strong accountability to engaged constituents.

The simplicity and stability argument. Voters in several states rejected new systems in 2024.⁠¹¹ Frequent changes in election rules can reduce trust and increase errors. A known system with understood flaws may be preferable to an untested one.

The competition argument. Hall’s evidence suggests the moderating filter is the competitive general election.⁠⁶ If so, the most effective reforms may be those that create more competitive districts, not those that change primary rules. See Gerrymandering.

Unresolved questions about primary elections

• Do top-two and top-four systems moderate legislators durably, or do parties and candidates adapt within a few cycles?
• How much of the ideological gap between primary and general electorates is about who votes, and how much is about who runs?
• Would higher primary turnout change outcomes, or would new primary voters resemble existing ones?
• What share of House seats are truly decided in primaries under a published, replicable method?
• How should the interests of party members in choosing their nominees be weighed against the interests of unaffiliated voters in safe districts?
• Do bundled reforms (primary changes combined with ranked-choice voting) fail at the ballot because of one component or both?