Democracy, Agenda Control, and Political Agency in the United States

Full report

On this page

Concepts that must be kept separate

The proposition under investigation is best treated as a dimensional hypothesis, not as a claim that a country either “is” or “is not” democratic. Modern empirical work on democracy increasingly separates electoral, liberal, participatory, deliberative, and egalitarian dimensions rather than reducing political systems to one binary label; V-Dem’s current research program is explicitly organized around such multidimensional measurement. The question here is narrower: can relatively strong electoral procedures coexist with weak citizen control over the construction of politically viable alternatives?

For this investigation, the key concepts should be defined as follows.

This vocabulary reveals why the proposition is coherent rather than contradictory. A voter may have a fully effective vote at the final electoral stage—the vote is counted accurately and can help determine which candidate wins—while having very little influence over the upstream processes that determined who ran, who obtained organizational backing, what policies candidates considered feasible, and which issue dimensions structured the contest.

That difference can be represented schematically as:

choice within an agenda ≠ control over the agenda

The first is essential to electoral democracy. The second is much closer to the “political agency” hypothesized here.

The proposition should therefore not be stated as “elections are meaningless.” That would be empirically indefensible: election outcomes plainly alter officeholding and policy, and winning or losing elections has measurable effects on citizens' political satisfaction. The stronger and more testable claim is:

The democratic value of voting may depend partly on whether citizens believe the menu from which they vote contains politically consequential alternatives sufficiently close to their preferences.

That formulation is strongly compatible with existing theory and receives meaningful—but incomplete—empirical support.

Why agenda control is analytically different from voting power

The central theoretical insight predates contemporary arguments about polarization or campaign finance. The “two faces of power” tradition associated with Peter Bachrach and Morton Baratz distinguished visible victories in authoritative decisions from the power to prevent certain conflicts from reaching authoritative decision in the first place. Later empirical work explicitly operationalizing the “two faces” distinction has found circumstances in which interest-group agenda-setting capacity is more closely related to policy outcomes than conventional measures of lobbying legislators. For example, Alexander Finger’s study of teachers' unions found stronger relationships for agenda-setting measures than for measures focused solely on direct legislative engagement.

This matters because an election normally occurs late in a long political production process:

social condition → perceived problem → salient public issue → proposed remedies → “serious” remedies → potential candidates → actual candidates → resourced candidates → nominees → viable general-election alternatives → winner → governing coalition → implementation

Citizen voting has its most direct leverage near the right-hand side. Political agency, as defined here, concerns influence across much more of the chain.

Media agenda-setting research supplies one mechanism. McCombs and Shaw's classic 1972 Chapel Hill study found a close correspondence between the issues receiving media emphasis and those undecided voters considered important, helping establish the empirical research program now called media agenda-setting. This does not show that media organizations unilaterally manufacture public priorities. Real-world events, audiences, parties, social movements, officials, interest groups, and competing media influence one another. The robust theoretical point is narrower: salience itself is politically consequential, and the issues about which voters form electoral judgments are not generated solely by voters acting individually.

Agenda power also appears inside legislatures. Hall and Wayman's influential study of political action committee contributions and congressional committees found that money could operate through the mobilization of legislative effort—encouraging sympathetic legislators to devote time and activity to issues—rather than simply purchasing roll-call votes. That distinction is crucial. Influence over whether and how intensely something is pursued can be politically important even where final votes cannot easily be shown to have been “bought.”

The same point cautions against a simplistic model of elite conspiracy. Agenda control in the United States is highly distributed. Journalists choose stories; party organizations recruit and endorse; activists generate primary pressure; donors choose recipients; candidates self-select into races; voters nominate candidates; committee chairs schedule legislation; presidents prioritize issues; courts alter feasible policy space; organized groups supply expertise and pressure; and electoral institutions determine which coalitions are strategically sustainable. Different actors can therefore exercise agenda power without any actor controlling the whole system.

The empirical question is consequently not:

“Who secretly chooses everything?”

It is:

“At each filtering stage, which actors have the greatest probability of moving an alternative forward, and how much effective leverage do ordinary citizens possess at that stage?”

That version is both theoretically sharper and empirically tractable.

Where ordinary citizens influence the American political menu

A stage-by-stage assessment of the United States produces a mixed picture. At some stages ordinary citizens possess powerful mechanisms; at others their influence is indirect, conditional, unequal, or difficult to measure.

The primary-election stage deserves particular emphasis because it complicates a simple elite-gatekeeping story. Andrew Hall's quasi-experimental research on U.S. House primaries from 1980–2010 exploited very close primary contests and estimated that when the more ideologically extreme candidate narrowly won the nomination, the party subsequently lost roughly 8–12 percentage points of general-election vote share and experienced a very large decline in its probability of winning the seat. Primary voters therefore do not merely ratify a predetermined general-election candidate: nomination choices can substantially alter what everyone else is offered.

That is evidence for citizen agency—but also evidence that different stages empower different electorates. The general-election voter cannot choose a primary loser. Once nomination has occurred, an earlier selectorate has transformed the later electorate's feasible set.

Party viability creates another filter. Strategic-voting research is instructive because “viability” can alter voter behavior without anyone formally excluding a candidate. Kawai and Watanabe's structural analysis of Japanese plurality elections estimated that strategic considerations were widespread even though only a much smaller share of voters ultimately cast what the authors call a “misaligned” vote for someone other than their most-preferred candidate. The deeper lesson is that observed votes can conceal latent preferences: the absence of votes for a weak party does not necessarily mean absence of support for that party's program.

American public opinion supplies suggestive evidence of such a supply-side dissatisfaction. In October 2024, Gallup found 58% of U.S. adults saying a third major party was needed because Republicans and Democrats did a poor job representing the American people. It was the twelfth consecutive Gallup measurement in which a majority favored the proposition, although support had declined from 63% in 2023.

That finding is important but easy to misuse. It does not prove that 58% would vote for the same third party, that a stable third-party coalition exists, or that plurality electoral rules alone prevent one from emerging. “Neither party represents me well” can aggregate people whose desired alternatives are mutually incompatible. It nevertheless demonstrates a large difference between acceptance of the existing party supply and mere willingness to participate in elections.

The last stage—policy responsiveness—produces one of the most consequential scholarly disputes. Gilens and Page analyzed 1,779 U.S. policy cases and reported that economic elites and business interest groups had substantial independent associations with policy change while average citizens had little or no estimated independent effect once preferences of other groups were included. This is often summarized as evidence that ordinary citizens “do not matter,” but that interpretation is too strong.

Peter Enns reanalyzed the conceptual problem and showed why. Preferences of middle- and high-income Americans tend to move together across issues: in his analysis of Gilens's data, relative policy support among the 50th and 90th income percentiles correlated at 0.94 overall and still 0.68 among issues where their support levels differed by more than ten percentage points. Consequently, a government can appear statistically responsive only to affluent preferences while often adopting policies that are also ranked similarly by middle-income citizens—what Enns calls coincidental representation.

The appropriate conclusion is therefore neither “American policy perfectly reflects voters” nor “elections have no policy influence.” The literature establishes a serious problem of identification: because social groups frequently want the same policy changes, it can be difficult to determine whose preferences caused policy adoption. Where preferences genuinely diverge, unequal responsiveness becomes easier to test.

For the democracy-versus-agency hypothesis, that distinction is central. A citizen may receive policy she wants without having caused it, or fail to receive policy she wants even though elections are genuinely competitive. Outcome congruence and political agency are related but not identical variables.

“My side lost” versus “my side was never viable”

This is the most important empirical distinction in the hypothesis, and existing research provides substantially more support for it than a purely theoretical critique of American elections would suggest.

First, there is overwhelming reason not to dismiss the ordinary winner–loser mechanism. Elections themselves alter political satisfaction. Blais and Gélineau's panel analysis of the 1997 Canadian federal election found significant effects of electoral winning and losing on satisfaction with democracy even after taking pre-election attitudes into account; both national and local outcomes mattered. A particularly useful 2026 quasi-experimental study by Miroslav Nemčok, Jean-François Daoust, and Piret Ehin examined an unexpected midterm government change in Estonia that altered some voters' winner/loser status without an election. It found suggestive, though fragile, evidence that becoming a loser reduced democratic satisfaction; importantly, effects appeared with a lag around a major policy change, which is more consistent with policy instrumentality than with a momentary emotional reaction to losing.

Contemporary U.S. survey movements reinforce the point. Pew's 2025 cross-national study found that American evaluations of democratic functioning shifted substantially by party after the 2024 election, with Republicans becoming markedly more satisfied and Democrats less so following the change in national political control. That is exactly what a winner–loser account predicts.

So the agency hypothesis cannot plausibly replace the winner–loser model.

But it can add an additional mechanism.

Hannah Ridge's study of 31 post-election surveys in 25 countries asked how perceived differences among parties condition the winner–loser gap. The gap in satisfaction with democracy was significantly greater when respondents perceived large differences among parties. This is an important corrective to one intuitive version of the agency hypothesis: people do not simply become most dissatisfied whenever parties look alike. When citizens see meaningful differences, winning and losing become more consequential because government choice matters more.

Yet perceived differentiation is not the same thing as having a viable representative alternative.

A particularly direct recent test comes from Sara Hobolt, Julian Hoerner, and Toni Rodon's 2026 study, Having a say or getting your way? Political choice and satisfaction with democracy. Combining cross-national observational evidence with a conjoint experiment, they report that having an ideologically congruent party is associated with greater democratic satisfaction when that party has a realistic possibility of entering government; ideological congruence without political instrumentality does not produce the same benefit.

That result gets unusually close to the proposed distinction:

“A party like me exists” is not necessarily enough. What matters is whether supporting it can plausibly affect who governs or what government does.

This means three states should be distinguished empirically rather than collapsed together:

The third state is where the democracy-versus-agency hypothesis has its strongest empirical plausibility.

Gallup's third-party findings are therefore more interesting when interpreted as a representation-supply measure than as a forecast of actual third-party voting. A majority saying the two major parties represent the public poorly is consistent with widespread dissatisfaction with the menu of alternatives. It does not establish why that dissatisfaction exists, whether respondents themselves lack representation, or whether they agree on an alternative.

The existing evidence thus supports a refined proposition:

Citizens care both about getting their preferred side into power and about possessing a politically consequential vehicle through which their preferences can plausibly enter the governing contest.

Those are different mechanisms. Winner–loser studies establish the first very strongly. The second has growing direct evidence but is much less extensively tested, especially in longitudinal U.S. designs capable of separating perceived nonrepresentation from generalized distrust, ideological extremity, economic grievance, and partisan losing.

How much of the proposed causal pathway is actually established?

The proposed sequence is:

perceived absence of representation → low external political efficacy → belief that elections offer constrained alternatives → loss of faith in electoral correction → anti-system attitudes → openness to fundamental political change

Taken as one causal chain, the evidence is not yet sufficient. Some links are strong, some are almost definitional, several are probably reciprocal, and the final transition is substantially underdetermined.

Two links deserve especially careful interpretation.

First, representation and efficacy are partially endogenous. A low-efficacy person may perceive every institution more negatively; a badly represented person may rationally develop low efficacy. Cross-sectional surveys typically cannot tell which happened first.

Second, “anti-system” is too broad to be a dependent variable unless decomposed. Someone who says “replace plurality elections with proportional representation” is questioning institutional design but not necessarily electoral democracy. Someone who supports an independent outsider is challenging established parties but not necessarily constitutional government. Someone who stops voting is withdrawing. Someone willing to cancel an election or ignore adverse results is challenging a very different democratic norm. Treating all four as one phenomenon would artificially strengthen the hypothesis.

The evidence also points toward multiple behavioral responses rather than one radicalizing trajectory.

Withdrawal is theoretically straightforward: if the perceived probability that one's participation can affect an outcome approaches zero, the expected instrumental value of voting or campaigning falls. But political alienation does not always produce apathy.

Insurgent participation can produce the opposite response. Primary elections allow dissatisfied citizens to try to alter a major party internally; Hall's research demonstrates that primary electorates can indeed impose substantially different nominees on a party and materially alter general-election outcomes. The 2016 Sanders and Trump insurgencies likewise motivated research specifically connecting political distrust with insurgent-primary support.

Third-party advocacy is another outlet. Gallup's 2024 figure of 58% support for the proposition that a third major party is needed shows that dissatisfaction with party representation can coexist with a preference for additional electoral competition.

Institutional reform is still another. A citizen can conclude that the problem lies in nomination rules, district structure, campaign finance, ballot access, or electoral formula rather than in elections themselves. That is precisely why dissatisfaction with political agency should not automatically be coded as hostility toward democracy.

The most defensible causal interpretation is therefore a branching model, not a deterministic chain:

perceived nonrepresentation / nonviability → reduced external efficacy or confidence in correction → political discontent → conditional response

with the response branching toward withdrawal, intraparty insurgency, third-party activity, institutional reform, or—in some circumstances—more genuinely anti-democratic attitudes.

Which branch occurs likely depends on internal efficacy, partisan attachment, ideological extremity, available organizations, perceptions of procedural fairness, elite messages, and whether citizens believe an institutional remedy remains accessible. The evidence assembled here does not justify treating authoritarian or anti-electoral change as the default endpoint.

What alternative electoral arrangements would change—and what they would not

Electoral institutions matter because they alter the relationship between latent preference diversity and politically viable alternatives. They do not abolish agenda control. They move it.

A useful comparison is therefore not “democratic versus undemocratic systems,” but where each system places the bottlenecks.

The comparative literature therefore supports a general proposition but not a one-directional ranking of systems.

Greater proportionality usually moves the bottleneck from party viability before the election toward coalition formation after the election. A voter may gain a party very close to her preferences but lose the ability to determine exactly which governing coalition that party subsequently joins. That is more party choice but potentially less direct government choice.

Closed-list proportional representation can substantially expand the number of viable parties while giving party organizations more control over individual candidates. That increases one dimension of agency and can reduce another.

Single-member ranked-choice voting can broaden expressive candidate choice while leaving the one-seat structure intact. It may therefore address the specific problem “I am afraid to rank my favorite first,” but it does much less to address “my political tendency cannot translate 15% dispersed national support into 15% of legislative seats.”

Multi-member proportional arrangements address the latter problem more directly, but increasing district magnitude also changes constituency accountability. Carey and Hix's work is valuable precisely because it rejects the assumption that the most proportional possible system is automatically optimal on every democratic dimension; their empirical argument is that low-magnitude PR can offer a favorable compromise between representation and accountability.

Nor would proportional representation abolish agenda-setting by money, media, experts, activists, or party leadership. Parties would still choose platforms; candidates would still require resources; journalists would still allocate attention; coalition partners would decide which campaign promises survive bargaining. Electoral-system reform therefore changes the architecture of alternative generation and conversion, not the existence of political filtering itself.

This is why the democracy-versus-agency hypothesis should not be reduced to “America has the wrong voting system.” Electoral rules are only one source of agenda constraint. Candidate recruitment, campaign resources, information systems, interest-group organization, legislative agenda control, and policy veto points can produce similar perceptions even under proportional elections.

Overall assessment of the democracy-versus-political-agency hypothesis

The hypothesis survives serious scrutiny, but in a qualified form.

The strongest version—

“Americans are alienated primarily because elections do not provide meaningful choices, and this drives them toward rejection of the political system”—

goes beyond the available evidence.

The better-supported version is:

A political system can possess meaningful electoral competition while leaving some citizens with a persistent perception that none of their preferred alternatives is capable of entering effective governing competition. That perceived lack of instrumentality is associated with lower democratic satisfaction and plausibly contributes to low external efficacy and broader political discontent.

Several findings support that conclusion.

First, agenda control is real and analytically distinct from final voting. Political science has long recognized that influence includes determining which matters receive action, and empirical studies of interest groups and legislative activity show that mobilizing attention and effort can be consequential apart from changing final roll-call votes. Media agenda-setting research likewise establishes that the salience structure within which voters make choices is itself politically produced.

Second, American candidate and party choice is neither completely predetermined nor unconstrained. Primary voters demonstrably alter candidate supply, and their nomination choices can substantially alter general-election outcomes. At the same time, voters encounter candidates only after entry, resource, visibility, and nomination filters have operated. Strategic electoral incentives further distinguish nominal alternatives from viable ones.

Third, dissatisfaction with political representation is widespread enough to be politically significant. Gallup's October 2024 finding that 58% of Americans said another major party was needed because the two existing major parties did a poor job representing the public cannot establish the causal hypothesis, but it rules out the assumption that dissatisfaction is confined to a negligible fringe.

Fourth, the distinction between losing and lacking an instrumental option has empirical support. Winner–loser effects are real and substantial: citizens often become less satisfied when their side loses. But recent evidence also indicates that representation matters particularly when one's congruent political alternative can realistically influence government. Ideological similarity without governing instrumentality appears much less satisfying. That is the strongest direct support for the user's proposed distinction.

Fifth, the hypothesis must coexist with strong counterevidence against a purely “constrained choice” account. Satisfaction can move sharply after ordinary electoral victory or defeat. Perceiving larger differences among competing parties actually magnifies the winner–loser satisfaction gap rather than eliminating it. Contemporary alienation therefore cannot be explained simply by saying that voters see the parties as interchangeable.

Sixth, unequal policy responsiveness is a legitimate concern but should not be exaggerated into a finding of zero mass influence. Gilens and Page provide important evidence of disproportionate elite and organized-business influence, while Enns demonstrates that correlated preferences make estimates of independent group influence much harder to interpret than popular summaries often suggest. The defensible conclusion is unequal and mediated responsiveness, not empirical proof that ordinary voting never matters.

Seventh, the proposed endpoint—anti-system or anti-democratic change—is the least well-established portion of the theory. Political dissatisfaction can produce quite different reactions. Demands for a third party are demands for another electoral alternative; primary insurgencies seek to capture existing parties; electoral-reform advocacy seeks different institutional rules; abstention represents withdrawal. None is equivalent to willingness to abandon free electoral competition. Satisfaction with democracy itself should not be conflated with commitment to democratic principles.

The resulting model is therefore not:

electoral democracy → fake choice → inevitable anti-system radicalization.

It is closer to:

**electoral procedures

→ for some citizens, persistent nonrepresentation → lower perceived external efficacy and electoral instrumentality → political discontent → multiple possible responses.**

• upstream filtering of issues, policies, candidates, and parties
• unequal or poorly perceived responsiveness

That distinction has substantial explanatory value because it identifies a kind of dissatisfaction that ordinary winner–loser models can miss.

A person saying “I voted, my preferred viable coalition lost, and I will try again next election” retains belief in electoral correction even while being intensely dissatisfied with the result.

A person saying “I have voted repeatedly, but no politically viable organization represents the combination of positions I want, and actors closer to my preferences never reach an effective decision stage” is expressing a different problem: not dissatisfaction with a decision, but dissatisfaction with the production of the decision set.

Empirically, political science has much stronger tools for measuring the first experience than the second. Surveys routinely record vote choice, winner/loser status, trust, efficacy, and satisfaction. They much less often reconstruct a respondent's ideal policy bundle and ask whether a sufficiently congruent candidate existed, whether that candidate was viewed as viable, whether the respondent strategically deserted that candidate, whether the preferred policy entered any party's agenda, and whether repeated absence of such options changed confidence in future electoral correction.

That is the principal research gap.

A decisive U.S. test would need longitudinal data distinguishing at least four quantities: pre-election ideal preferences; perceived ideological congruence with every available alternative; perceived probability that each alternative could win or influence policy; and post-election external efficacy, legitimacy, and support for different kinds of political change. It would then have to compare voters whose viable side genuinely lost with otherwise similar citizens whose preferred alternative never became viable. The recent instrumentality research points strongly in this direction, but that design has not yet become the standard way American anti-system sentiment is studied.

The bottom-line assessment is therefore:

The democracy-versus-political-agency hypothesis is empirically credible and probably captures a real component of contemporary political alienation, especially the distinction between losing a contest and lacking an instrumental representative option. It is not, however, established as the dominant cause of U.S. anti-system sentiment, and the evidence is weakest precisely at the final causal steps from constrained choice to rejection of the political system.

The clearest empirical finding is not that American citizens lack all political agency. They plainly possess consequential agency at elections, primaries, party insurgencies, public-opinion shifts, and organized collective action. The more defensible finding is that agency is distributed unevenly across stages: citizens have much more direct authority over selection among surviving alternatives than over the upstream processes by which issues, policies, candidates, resources, nominees, and viable parties become the alternatives from which they ultimately select.

That gap between electoral choice and agenda control is theoretically coherent, empirically observable in several adjacent literatures, and important enough that explanations of contemporary political discontent which measure only “who won the election?” are likely incomplete.