Political-Party Nominations and the Sense That the “Real Choice” Has Already Been Made

Full report

On this page

Research memo — United States focus, with comparative evidence

The central finding is that the intuition behind the question is partly correct but too simple if stated as “the parties choose the candidates.” In the modern United States, candidate selection is better understood as a multi-stage filtration process. Before most citizens see a general-election ballot, potential candidates have decided whether they fit a party, raised or failed to raise money, obtained or failed to obtain organizational support, qualified for the ballot, survived a primary electorate that is usually much smaller and more politically engaged than the general electorate, and—in presidential contests—survived a sequential process in which early results affect later viability. Empirical research finds evidence of meaningful selection at several of these stages.

That does not mean that party leaders simply dictate nominees. American primaries transfer unusually large amounts of formal nomination authority to voters compared with many democracies; party-backed candidates can lose; outsiders can win; and general-election voters sometimes impose a large penalty on nominees who emerged from a primary. A regression-discontinuity study of U.S. House elections found that when a relatively extreme candidate barely defeated a more moderate candidate in a primary, the party subsequently lost roughly 8–12 percentage points of general-election vote share and 35–53 percentage points in its probability of winning. In competitive districts, therefore, the general election is demonstrably not merely ceremonial.

The most defensible conclusion is consequently conditional: the perception that important choices have already been made is most grounded where the pre-primary candidate pool is narrow, primary participation is low or selective, and the constituency is so partisan that winning the dominant party's nomination is close to winning office. It is least grounded where candidate entry is broad, the nomination electorate is inclusive and well attended, and the general election is genuinely competitive.

Bottom line

Three different phenomena are often conflated under the label “party gatekeeping.”

Entry filtration happens before voters see a ballot. Danielle Thomsen's research on state legislators found that liberal Republicans and conservative Democrats—potential congressional candidates who fit their party's ideological reputation less well—were less likely to run for Congress. Her mechanism is partly self-selection rather than an order from party headquarters: potential candidates estimate whether they belong in the party and can win its nomination, and some opt out. Hans Hassell's study of U.S. Senate primaries identifies a more organizational mechanism: candidates less connected to the party were less likely both to remain in the primary and to win it, leading him to conclude that party networks can help remove alternatives before primary voters make their choice.

Electorate filtration occurs at the primary or caucus. In the 2022 midterms, primary turnout amounted to 21.3% of the voting-eligible population according to the Bipartisan Policy Center; that figure combines participation across parties, meaning that the voters actually selecting any one party's nominee were necessarily a smaller share of the total eligible electorate. Survey evidence from congressional primaries also depicts primary voters as unusually attentive, strongly partisan, older and more educated, with relatively few self-described independents participating even where rules allow them and with ideological distributions that are not identical to those of general-election voters.

Electoral filtration occurs because the consequences of a nomination depend on district competitiveness. Where one party is overwhelmingly favored, its primary can become the election with the greatest practical consequence; where the district is competitive, the general electorate can and does reject nominees. Hall's causal evidence is particularly important here: more extreme primary winners did substantially worse in subsequent general elections, with the electoral penalty concentrated where general-election competition could matter.

This distinction matters politically. Saying that “a small group selected the winner” can be substantially accurate in a safe seat without implying that a central party committee selected the candidate. The decisive group may instead be a relatively small body of primary voters, after an even earlier process influenced by candidates themselves, donors, activists, elected officials, organizations and media attention.

There is also important counterevidence to the proposition that simply opening primaries solves ideological selection. McGhee, Masket, Shor, Rogers and McCarty examined state legislative ideology across differing nomination systems and found that the openness of a primary had “little, if any” relationship to the ideological extremism of the politicians it produced. Thus, closed-primary rules are not established as a singular cause of polarization, nor is an open primary an established universal cure.

The candidate-selection pipeline

A useful way to locate power is to follow it chronologically rather than asking whether “voters” or “parties” select candidates in the abstract.

The table makes an important conceptual point: influence is cumulative rather than exclusive. A donor does not cast ten thousand primary votes by giving money. A party committee does not necessarily possess a legal veto. A newspaper does not put a candidate on the ballot. Yet each can affect whether the alternatives among which voters later choose ever become credible alternatives.

Fundraising, incumbency and visibility

The funding stage is especially easy to misdescribe. The strongest evidence does not support the literal proposition that donors routinely “appoint” nominees. Rather, fundraising affects whether a campaign can hire staff, communicate, organize and remain viable. The distinction is important because money acts mostly before and through voter choice, not instead of it. Brookings' congressional-primary research describes a long-standing pattern in which incumbents are frequently unopposed or opposed by candidates who are underfinanced and little known.

Incumbency therefore operates partly as an entry deterrent. A prospective challenger anticipating an expensive race against a visible officeholder may never enter. The resulting uncontested or weakly contested primary can look overwhelmingly supportive of the incumbent even though one important causal event—the failure of competitive candidates to emerge—occurred before any votes were cast. This interpretation is consistent with the candidate-emergence evidence in Thomsen rather than requiring a theory of direct party command.

Media attention can similarly affect the informational environment without constituting a formal veto. Hall found that the general-election penalty associated with more extreme nominees was larger where voters had better media information about them, suggesting that information conditions affect how effectively the general electorate can respond to primary selection. The literature is much less successful at assigning a stable numerical “share of nomination power” to media than it is at documenting turnout, party support or candidate-entry effects.

Primary rules, primary electorates and effective decisiveness

The term primary covers meaningfully different institutions. NCSL's May 2026 classification underscores that there is no single American primary system.

In a closed primary, participation in a party's nomination is generally restricted to voters registered or otherwise affiliated with that party. Unaffiliated voters are excluded from that party's nomination contest. The democratic-participation criticism is straightforward: citizens who do not join a party may be excluded from an election that, in a very safe constituency, is effectively more consequential than November. The association-based answer is equally straightforward: someone who declines to affiliate with an organization does not necessarily have an entitlement to determine that organization's representative.

Semi-closed or partially closed systems loosen the boundary, usually by allowing parties to admit unaffiliated voters while continuing to exclude voters registered with a rival party. This preserves a meaningful membership boundary while giving independents some access.

Systems open to unaffiliated voters let independents select a party ballot under specified rules, while registered party members may remain tied to their own party's contest. Partially open systems can permit crossing party lines but may treat the voter's ballot choice as an affiliation or registration action; NCSL identifies Illinois, Indiana, Iowa, Nevada and Ohio in this category in its 2026 classification.

In a conventionally open primary, a voter can select which party's primary to participate in without a prior party-registration requirement, although the voter generally cannot participate in multiple parties' nomination contests in the same election. NCSL's 2026 classification lists a substantial group of states—including Alabama, Georgia, Michigan, Minnesota, Mississippi, Missouri, Montana, South Carolina, Texas, Vermont, Virginia and Wisconsin—under its open category.

These formal access rules matter, but formal eligibility and actual participation are different. Opening the door to independents does not ensure that large numbers walk through it. Brookings' 2018 primary-voter study found an electorate heavily composed of strong partisans, with relatively few self-identified independents even in places where they could participate; primary voters also tended to be older, better educated and more politically attentive.

That is why the apparently intuitive chain

closed primary → extreme electorate → extreme nominee

is too deterministic. The electorate can be selective because of turnout and political motivation, not merely registration law. Consistent with this, McGhee and colleagues found little evidence that primary openness itself systematically produced more moderate state legislators.

Caucuses and conventions

Caucuses intensify the distinction between permission to participate and practical participation. The Election Assistance Commission describes caucuses as party-run meetings governed by party rules, often occurring at a particular physical place and time, with rules differing across parties and states. Requiring synchronous meeting participation imposes a different burden from casting a conventional ballot; as an institutional inference, this gives highly motivated activists greater potential leverage, although the size of that effect should not be generalized from structure alone.

Conventions and delegate processes move power another step toward organized participants. In presidential nominations, primary and caucus results usually allocate or select delegates, with the national convention formally completing the nomination. In contemporary uncontested conventions the result is generally settled before delegates assemble, but the important point is that the presidential nomination is not formally a single nationwide popular primary.

Top-two and top-four systems

Top-two reforms change the structure more fundamentally. California's law places candidates for voter-nominated offices on a common primary ballot and advances the top two regardless of party. California explicitly distinguishes these offices from traditional party-nominated offices: the winners are not formally nominees chosen by a political party. Washington likewise describes its top-two primary as a winnowing election in which voters may choose among all candidates and the top two advance.

The participation advantage is clear conceptually: voters are not barred from considering a candidate because they belong to a different registration category. But top-two does not eliminate filtration; it relocates it into a common first-round electorate. Two candidates from the same party can advance, and candidates outside the leading parties can still be eliminated before November. Early analysis of California and Washington also cautioned against assuming that top-two would automatically depolarize legislatures: Brookings' review of the 2014 cycle found limited early evidence of the promised moderating effect. This is broadly compatible with McGhee and colleagues' finding that nomination-system openness has much less ideological effect than reform rhetoric sometimes presumes.

A top-four system retains more alternatives after the preliminary round. Alaska is the principal recent U.S. example in NCSL's classification: four candidates rather than two survive the all-candidate primary. That design reduces the chance that one early winnowing contest leaves November voters with only two candidates, but it does not abolish upstream barriers such as candidate emergence, fundraising or ballot qualification.

How representative is the primary electorate?

The strongest answer is less representative on participation and political engagement, variably different on ideology, and not reducible to one demographic stereotype.

The 21.3% eligible-voter turnout observed across 2022 primaries demonstrates how small the aggregate participating population can be. The Brookings primary surveys provide evidence that participants are disproportionately attentive, educated, older and strongly partisan. Those differences create the possibility of demographic and ideological selection.

But two qualifications matter. First, the voters who participate in a Democratic primary and those who participate in a Republican primary are not trying to reproduce the median of the entire country; they are selecting representatives of distinct political coalitions. Second, finding that primary voters are unusually ideological does not prove that primary rules cause nominee extremism. McGhee et al.'s cross-state evidence is an important counterweight to precisely that causal inference.

Does a small subset sometimes make the practically decisive choice?

Yes. But “how often?” does not have a defensible single national percentage.

A national figure would require a definition of when an election is “effectively decided” in the primary—perhaps a threshold based on general-election win probability, partisan margin or absence of serious opposition—and would have to be calculated separately across offices and cycles. The 21.3% turnout statistic is evidence about participation, not a direct estimate that 78.7% of offices are “decided” by primary voters.

The causal evidence nevertheless shows the two ends of the continuum. In electorally competitive districts, a primary electorate can choose a nominee that the general electorate subsequently punishes severely: Hall estimates an 8–12 point general-election vote penalty from barely nominating the more extreme candidate in close House primaries. In strongly partisan constituencies, by contrast, the general-election constraint is weaker; in such seats the nomination is much closer to the effective selection of the officeholder. The phenomenon is therefore structurally common enough to matter, especially in safe seats, but not equivalent to saying that American general elections as a class are merely ratification exercises.

Presidential nomination dynamics

Presidential nominations accentuate pre-election filtering because the process is sequential rather than simultaneous.

Most presidential primaries and caucuses occur months before the November election, and voters' choices translate into convention delegates under party and state-specific rules. A citizen voting late in the calendar therefore does not necessarily face the candidate field that an early-state voter faced.

Knight and Schiff's Journal of Political Economy study treats presidential primaries explicitly as a sequential voting system and identifies momentum and social-learning effects: information generated by one electorate affects subsequent evaluations and electoral performance. This creates a democratic asymmetry even without any conspiracy or centralized party control. Early voters provide information, donors and activists update their expectations, media attention reallocates, and candidates whose perceived path to the nomination collapses may withdraw.

Consequently, later voters can encounter a field that has already been narrowed by early electorates and by the viability reactions to those results. The appropriate inference is not that early states literally “cast the votes” of later states. It is that voting order gives some electorates influence over what choices remain available to later electorates. Knight and Schiff's evidence is consistent with that proposition.

Presidential politics also illustrates why “party power” should not be treated as a single institution. The influential book The Party Decides argues that networks of elected officials, activists and other party actors historically coordinated around candidates through endorsements and pre-primary support rather than simply waiting for mass voting to begin (Cohen, Karol, Noel & Zaller 2008). Yet the record of outsider candidates and failed elite coordination demonstrates that these signals are not a reliable veto. The better model is elite coordination plus voter choice, with the balance varying sharply by election.

This also clarifies candidate withdrawal. A late-state voter may technically participate fully while having no opportunity to vote for a candidate who withdrew after early contests. That is a genuine reduction in the voter's choice set, but it can arise from primary votes, fundraising decisions, polling, candidate judgment and media expectations simultaneously—not solely from party officials.

Participation versus association

The central normative conflict cannot be resolved merely by maximizing the number of people permitted to cast a nomination ballot. Two democratic values are genuinely in tension.

Comparative candidate-selection scholarship describes essentially the same dilemma. Rahat and Hazan characterize candidate-selection methods by how inclusive the selectorate is—from a single leader or small party agency through party members to the entire electorate—and emphasize that increasing inclusiveness can trade off against organizational control and other representational goals (Rahat & Hazan 2001).

This produces a crucial distinction:

Party gatekeeping can be democratic when parties are themselves voluntary associations, internally accountable, transparent and electorally contestable, and when their screening helps citizens organize political responsibility around meaningful programs.

The same gatekeeping can be democratically problematic when it concentrates effective selection in unrepresentative networks, deters plausible alternatives before citizens participate, or operates in a district where victory in the party's nomination virtually assures public office.

Neither conclusion follows merely from the existence of a party endorsement or a primary.

The U.S. evidence also warns against a false institutional choice between “party bosses” and “the people.” Hassell finds organizational influence inside a primary system, while Thomsen finds self-selection before any formal party decision. Modern nomination politics is therefore a hybrid in which voters exercise real authority over a choice set partly constructed by other actors.

Legitimacy consequences

Here the evidence is considerably stronger for representation and political efficacy than it is for the specific causal claim that believing parties pre-filter nominees causes democratic alienation.

Pew's 2023 survey found that 25% of U.S. adults said neither major party represented the interests of people like them even somewhat well. Among independents and people identifying with other parties, the figure was 41%.

That unrepresented group also reported systematically more negative political attitudes. Only 21% had at least some confidence in the future of the political system, compared with 32% among those represented by the Republican Party, 49% among those represented by the Democratic Party and 59% among those who said both represented them well. Only 35% of the unrepresented group believed voting by people like them could affect the country's direction at least somewhat, compared with 56% among those represented by Republicans and 72% among those represented by Democrats. Only 26% of the unrepresented said there was usually a candidate sharing their views.

The same survey found that people who felt unrepresented were less attentive to government and more interested in additional political parties: half of the unrepresented group wished for more parties, compared with four in ten or fewer among the represented groups. A later Pew survey conducted in November 2025 found that 37% of U.S. adults overall said wishing for more political parties described their views extremely or very well, rising to 44% among adults under 50.

Those results map onto several of the requested legitimacy outcomes, but their limitations are essential:

Public beliefs about power are nevertheless consistent with a wider legitimacy problem. Pew found that about 70% of adults thought ordinary people in lawmakers' districts had too little influence, while roughly eight in ten members of both partisan coalitions thought large campaign donors had too much influence on congressional decision-making. These questions concern governing influence, not candidate nomination, so they should not be presented as direct evidence that voters believe donors choose nominees. They do show that the broader perception of unequal political influence is widespread.

What can actually be said causally?

The evidence is asymmetric.

There is reasonably strong causal evidence about electoral consequences of nomination choice. Hall's regression-discontinuity design exploits extremely close primaries and shows that barely nominating a more extreme candidate substantially worsens the party's general-election performance.

There is good observational evidence about pre-primary filtering. Thomsen links party fit to candidate emergence, and Hassell links party connectedness to persistence and nomination success. Both identify important mechanisms, but neither licenses the claim that every candidate who does not run was “blocked by the party.”

There is counterevidence against a simple institutional causal story. McGhee and colleagues find little if any relationship between primary openness and the ideological extremity of resulting legislators.

There is strong correlational but weak causal evidence on legitimacy. Pew demonstrates that feeling unrepresented is associated with lower system confidence, lower efficacy, fewer perceived candidate choices and greater desire for additional parties. It does not show that party nomination procedures created those attitudes.

Thus, a claim such as “party pre-filtering causes political distrust and abstention” goes beyond the evidence reviewed here. A more defensible claim is: pre-filtering is empirically real; feeling politically unrepresented is correlated with distrust and low efficacy; but the causal bridge between the two remains insufficiently identified.

Comparative perspective and major uncertainties

The United States is unusual not because other democracies lack candidate gatekeeping, but because it combines candidate-centered primary participation with single-member elections and a highly durable two-party system. Comparative systems redistribute the same selection power to different stages.

Closed-list proportional representation

Under a fully closed-list PR system, the party presents an ordered candidate list, voters choose among parties, and the voter cannot reorder candidates at the general election. Candidate-level association power is consequently stronger: whoever controls the party's nomination and ranking procedures has substantial influence over which individuals receive seats. The internal selector may nevertheless be a party membership ballot, local organization, nomination committee, leadership body or some combination; “closed list” does not itself imply rule by a single leader (Rahat & Hazan 2001; Reynolds, Reilly & Ellis 2005).

From the participation perspective, this may look less democratic than an American primary because citizens cannot select among individual candidates within the party at the general election. From another perspective, proportional allocation can make the party choice in the general election more meaningful: voters can support a smaller party without requiring that party first defeat one of two dominant coalitions in a single-member district (Reynolds, Reilly & Ellis 2005).

The tradeoff therefore differs from the American one. A closed-list country can have more party-level choice but less candidate-level choice.

Open-list proportional representation

In open-list PR, parties still screen who may appear on their lists, but voters can influence which of those candidates receive seats through preference votes. It therefore divides nomination power: the party controls admission to the list while ordinary voters have more influence over intraparty ranking at the general election (Rahat & Hazan 2001; Reynolds, Reilly & Ellis 2005).

This addresses one aspect of the “decision already made” concern. A voter can influence not just how many seats a party receives but which of its candidates occupy them. The countervailing risk is that candidates from the same party must compete against one another for personal votes, potentially increasing candidate-centered campaigning and weakening collective party responsibility (Rahat & Hazan 2001).

Single transferable vote and other ranked systems

Systems using the single transferable vote, most prominently Ireland for national parliamentary elections, permit voters to rank individual candidates and transfer preferences under the counting rule. Candidate selection consequently remains partly with parties—because parties decide whom to nominate—but more of the final intraparty selection occurs in the public election itself (Reynolds, Reilly & Ellis 2005).

More generally, ranked ballots do not automatically solve pre-selection. A ranked general election containing only candidates who survived restrictive entry procedures can still be heavily prefiltered. Ranked choice primarily increases the information voters can express among the candidates actually on the ballot.

This is an important lesson for U.S. reform arguments: changing the counting rule and changing the candidate-entry or nomination rule are separate interventions.

Top-two and top-four as hybrid models

California and Washington's top-two systems move intraparty competition into an electorate open to every voter, eliminating the traditional formal party nomination for covered offices. Alaska's top-four model preserves four finalists rather than two after the initial winnowing stage. These designs prioritize participation over a party's exclusive control of a nomination ballot.

Yet neither removes candidate emergence, money, endorsements, incumbent advantage or voter attentiveness as sources of filtration. And the empirical literature does not justify assuming that broader first-round access will by itself create ideologically moderate legislators.

Overall assessment

The research supports five propositions with different confidence levels.

High confidence: meaningful political selection occurs before the U.S. general-election ballot. Some potential candidates self-select out; party networks affect candidacy and nomination success; primary participation is much smaller than general-election participation; and in safe constituencies the primary has unusually high practical importance.

High confidence: this is not equivalent to centralized party control. Primary voters possess genuine authority, party-supported candidates can be defeated, and general-election voters can severely punish a primary nominee.

Moderate confidence: ideological and demographic selection occurs within primary electorates, but the magnitude varies by place, party, office and year. Primary voters differ systematically from the broader public, yet changing closed to open rules alone has not been shown to reliably change the ideology of officeholders.

High confidence about correlation, low confidence about causation: Americans who feel unrepresented by both parties exhibit much lower confidence and efficacy and are more attracted to additional party choices. There is not comparably strong evidence that perceiving party nomination gatekeeping causes those attitudes.

Low-to-moderate confidence on a single national “how often?” estimate: there is no canonical statistic measuring the percentage of American offices for which the general election is merely a ratification of a small-subset primary decision. Any responsible estimate would need to define practical decisiveness, incorporate uncontested primaries, candidate withdrawals, seat competitiveness and offices below the federal level, and distinguish primary turnout from the smaller number of voters who actually supported the winning nominee. The evidence is sufficient to establish that the phenomenon is important; it is insufficient to attach one universal percentage to it.

The democratic conclusion is consequently not that party gatekeeping is inherently democratic or inherently anti-democratic. Association and participation solve different problems. Parties need enough autonomy to form coherent organizations, recruit candidates and offer intelligible alternatives; citizens need enough access to prevent those organizations or their surrounding donor-activist networks from turning public elections into choices among alternatives that the electorate had little meaningful role in shaping. The institutional question is where along the pipeline each democracy wishes to locate that balance.

Selected high-quality sources

National Conference of State Legislatures. “State Primary Election Types,” updated May 19, 2026. The best current overview of U.S. closed, partially closed, open-to-unaffiliated, partially open, open and multiparty primary rules.

U.S. Election Assistance Commission. “Primary Election Types” and “Presidential Elections.” Federal election-administration explanations of party primaries, nonpartisan primaries, caucuses and presidential delegate selection.

**Bipartisan Policy Center. 2022 Primary Turnout: Trends and Lessons for Boosting Participation, 2023.** Reports nationwide midterm-primary participation of 21.3% of eligible voters.

Brookings Institution, Primaries Project. “Political Polarization and Voters in the 2018 Congressional Primaries” and “The Ideology of Primary Voters,” 2018. Large primary-voter survey covering partisanship, ideology, political attention and demographics.

**Thomsen, Danielle M. “Ideological Moderates Won't Run: How Party Fit Matters for Partisan Polarization in Congress.” Journal of Politics 76(3), 2014.** Central evidence on candidate self-selection before the primary ballot.

**Hassell, Hans J. G. “Party Control of Party Primaries: Party Influence in Nominations for the US Senate.” Journal of Politics 78(1), 2016, pp. 75–87.** Evidence that party connectedness predicts both persistence in Senate primaries and nomination success.

**Hall, Andrew B. “What Happens When Extremists Win Primaries?” American Political Science Review, 2015.** Regression-discontinuity evidence identifying large general-election penalties from narrowly nominating relatively extreme House candidates.

**McGhee, Eric, Seth Masket, Boris Shor, Steven Rogers and Nolan McCarty. “A Primary Cause of Partisanship? Nomination Systems and Legislator Ideology.” American Journal of Political Science 58(2), 2014, pp. 337–351.** Important counterevidence: primary openness has little if any effect on resulting legislators' ideological extremity.

**Knight, Brian, and Nathan Schiff. “Momentum and Social Learning in Presidential Primaries.” Journal of Political Economy 118(6), 2010, pp. 1110–1150.** Major analysis of the consequences of sequential presidential-primary voting.

**Cohen, Marty, David Karol, Hans Noel and John Zaller. The Party Decides: Presidential Nominations Before and After Reform. University of Chicago Press, 2008.** Foundational account of presidential party networks, endorsements and elite coordination; best treated as an influential theory rather than a claim that party elites invariably prevail.

**Pew Research Center. Americans' Dismal Views of the Nation's Politics, especially “How Well the Major Parties Represent Americans,” 2023.** The strongest public-opinion evidence here linking felt nonrepresentation to low system confidence, low efficacy and desire for additional parties.

Pew Research Center. “Nearly 4 in 10 Americans Wish They Had Options Beyond the Republican and Democratic Parties,” January 2026. Current evidence on demand for additional party alternatives.

California Secretary of State. “Primary Elections in California.” Authoritative explanation of voter-nominated offices and California's top-two system.

**Rahat, Gideon, and Reuven Y. Hazan. “Candidate Selection Methods: An Analytical Framework.” Party Politics 7(3), 2001, pp. 297–322.** Foundational comparative framework for analyzing who controls candidate selection and the inclusiveness-versus-control tradeoff.

**Reynolds, Andrew, Ben Reilly and Andrew Ellis, eds. Electoral System Design: The New International IDEA Handbook. International IDEA, 2005.** Standard comparative reference for list PR, preferential systems, STV and the distribution of voter choice across electoral systems.