Shareholder Democracy
Why It MattersHow It Works

This is a beta report. We are still tuning how these are researched and written, and a correction from you changes what is published.

What You Care About · Dependency ReportBeta

How far does reducing political polarization depend on shareholder democracy?

Written September 25, 2026 by claude-opus-5 · method 1.6

Pivotal

Owners voting their values can deliver most of this outcome.

For 1 of 8 obstacles shareholder democracy is pivotal; for 1 it is necessary; for 2 it is helpful; for the other 4 it is independent.

Reducing partisan animosity in the United States runs through eight obstacles. A shareholder vote reaches four of them and is the whole of the lever on one: what the companies that own America's television news choose to put on air.

Fox News viewership shifted American ideology and party identity rightward, and a 0.05 rise in its ratings raised Republican vote shares by 0.5% in elections, on a 2024 study of two decades of ratings and results. Local television news reaches more people still, 25 million nightly viewers as of 2019, and five companies controlled 37% of those stations as of 2016, with Sinclair alone holding 191 of them and reaching almost 40% of the population. Stations Sinclair acquired moved towards national politics at the expense of local coverage and raised their coverage of polarizing national topics, and the study that tested audience taste against owner preference as rival explanations found the owner's preferences in the content.

That is the clearest case on this map of an obstacle that is nothing but corporate conduct. Congress has had the Federal Communications Commission review broadcast ownership since 1996, and that remit reaches who may hold a license rather than what the license holder broadcasts, so the owners are the only people who can direct it. All four companies named here are ones a vote reaches: Sinclair's 2026 proxy statement records 48,254,031 Class A shares each carrying one vote, Nexstar's records 30,538,965 common shares each carrying one, and Gray Media's gives one vote to a common share and ten to each Class A share.

The feeds come second. An audit of Twitter's engagement-based ranking found it amplifies out-group hostile content that users say makes them feel worse about the other party, and that users do not prefer the political posts it selects for them. A 2023 field experiment on X found the algorithmic feed shifting political opinion in a conservative direction while promoting conservative content and demoting traditional media. The same experiment found no significant movement in affective polarization when the algorithm was switched either on or off, and an earlier switch-off on Meta's platforms found no political effects. Ranking is a decision taken inside the company, and changing it is necessary without being sufficient on the evidence available.

The record of owners asking is longer than the record of owners winning. Trillium Asset Management urged Alphabet's shareholders to support algorithm disclosure in 2022 and filed the 2023 proposal through its ESG Global Equity Fund, and it told shareholders the 2022 proposal had taken 56% of non-insider votes and had not been implemented. The tally at that meeting was 117,998,457 for and 483,701,300 against, which is the arithmetic of shares that carry ten votes each. As You Sow asked Meta's shareholders in 2024 for a report on prohibiting political advertising and on restoring the enhanced actions of the 2020 election cycle that reduced the amplification of false and divisive information, and it drew 152,703,615 votes against 4,855,808,278. Meta's shareholders asked in 2025 for the results to be published by share class and were refused, 1,028,098,566 to 3,959,996,157.

Corporate political money is a real lever and a smaller one. Outside groups spent $4.5 billion of non-party money in the decade after Citizens United, including $963 million from groups that do not disclose their funders, and the undisclosed part reaches no shareholder. What a vote does reach is what a listed company gives and what it pays in dues, and measuring corporate and trade group positions on lobbied bills against their giving finds their revealed preferences more conservative than their contributions imply. Disclosure and congruency votes have run at Meta, Alphabet, UnitedHealth Group and Coca-Cola.

Four obstacles sit with actors that have no shareholders. Partisanship has become one of Americans' most salient social identities, and no company decides whom a voter counts as their own. Segregation between the parties has risen year over year at every level from congressional districts to census blocks, and the sorting is done by households choosing where to live. Nomination rules are written by state legislatures and by the parties, and most evidence finds primaries make little difference to how members behave. District lines are drawn by legislatures and commissions, the national bias from gerrymandering in the 2020 cycle came to about two seats, and the Senate's disregard of population is in the Constitution.

A member who wants to move this subject with the shares they hold has a short list of addresses rather than a long one, and the first name on it is a station group.

No person has reviewed this report. Every claim was checked against the page it cites. If we got something wrong, tell us. How we score dependency.

What stands in the way, scored

Eight obstacles, ordered by how far each depends on how shares are voted and then by the weight of the evidence behind it. The families are the ones a systematic review of 135 causal studies uses, and the media family is split in two here because a broadcaster and a ranking algorithm are reached in different ways.

Ordered by how far each depends on shareholder democracy, and within that by severity as the analyst ranked it, drawing on What drives affective polarization? A systematic literature review.

1. Nationalized partisan broadcast and cable newsPivotal

Companies involved: Fox Corporation (shareholders vote here); Sinclair (shareholders vote here); Nexstar Media Group (shareholders vote here); Gray Media (shareholders vote here)

EDGAR full-text search for a shareholder proposal at Sinclair carrying the word polarization, and for one at Fox Corporation carrying the word misinformation, returned neither. The proposals on this subject have gone to Meta and Alphabet instead.

No lobbying by these broadcasters on the ownership rules or on any other measure bearing on this obstacle was found in the documents read, so the rating rests on the first question alone.

The causal evidence here is the strongest in the literature the review surveys: ratings and returns across two decades for cable, and station acquisitions as a natural experiment for local television, which together tie editorial decisions to ideology, party identity and vote share.

Fox News viewership shifted American ideology and party identity rightward, and a 0.05 rise in its ratings raised Republican vote shares by 0.5% in elections, on a 2024 study of two decades of ratings and results. Local television news carries a larger audience still, 25 million nightly viewers as of 2019, and its ownership is concentrated: five companies controlled 37% of local stations as of 2016, and Sinclair alone held 191 stations reaching almost 40% of the United States population. What that ownership does to the news has been measured rather than assumed. Stations Sinclair acquired turned towards national politics at the expense of local coverage and raised their coverage of polarizing national topics, and the study that set audience taste against owner preference as rival explanations found the owner's preferences in the content. This obstacle is what these companies broadcast, which is their own conduct and nobody else's: Congress has had the Federal Communications Commission review broadcast ownership since 1996, and that remit reaches who may hold a license rather than what the license holder airs. A vote reaches all four of the station groups and networks named here, one vote for each common share at Nexstar and Sinclair's Class A, and ten for each of Gray Media's Class A shares.

Confirmed for this obstacle (11)
  • Fox News viewership shifted American ideology and party identity rightward. source
  • A 0.05 rise in Fox News ratings increased Republican vote shares by 0.5% in elections. source
  • Local television news drew 25 million nightly viewers, more than national cable programmes such as Fox News and MSNBC, when the study was published in 2019. source
  • As of 2016 five companies controlled 37% of local television stations in the United States. source
  • Sinclair Broadcast Group owned 191 stations reaching almost 40% of the United States population when the study was published in 2019. source
  • The researchers tested whether the rightward slant of acquired stations followed audience tastes or the owner's preferences, and concluded that ownership preferences influence content. source
  • Stations acquired by Sinclair reported more often on national news at the expense of local topics and increased their coverage of polarizing national topics. source
  • Congress has required the Federal Communications Commission to review broadcast ownership regulations since 1996, which is a remit over who may own stations rather than over what they broadcast. source
  • Sinclair's 2026 proxy statement records 48,254,031 Class A shares, each carrying one vote on each proposal at the annual meeting. filing
  • Gray Media's 2026 proxy statement records one vote for each common share and ten votes for each Class A share. filing
  • Nexstar Media Group's 2026 proxy statement records 30,538,965 common shares outstanding, each carrying one vote. filing
2. Engagement-ranked feeds at the dominant platformsNecessary

Companies involved: Meta Platforms (shareholders vote here); Alphabet (shareholders vote here); X (privately held, no shareholder vote)

Lobbying by these platforms against a named measure on ranking or amplification was not found in the documents read. What the record does carry is their shareholders asking for disclosure and being refused, which is the first question rather than the second.

Ranked second on the strength of the designs rather than the size of the effect: audits and field experiments show the ranking amplifies hostile content, and the same experiments find no measurable movement in affective polarization when the ranking is switched.

A preregistered audit of Twitter's engagement-based ranking found it amplifies emotionally charged, out-group hostile content that users themselves say makes them feel worse about the other party, and that users do not prefer the political posts it selects for them. A 2023 field experiment on X found that moving users onto the algorithmic feed raised engagement and shifted their political opinions towards more conservative positions, and that the algorithm promotes conservative content while demoting posts by traditional media. The same experiment found no significant effect on affective polarization or self-reported partisanship from switching the algorithm on or off, and an earlier switch-off on Meta's platforms found no political effects either. A 2026 model of ranking systems finds that weighting a feed towards likes and reshares raises engagement and raises misinformation and ideological polarization with it. A ranking rule is written, weighted and changed inside the company that runs the feed, so changing it is something only the owners can direct, and the measured evidence stops short of showing that changing it alone would be enough.

Confirmed for this obstacle (14)
  • In a preregistered audit, Twitter's engagement-based ranking amplified emotionally charged, out-group hostile content that users said made them feel worse about their political out-group, compared with a reverse-chronological baseline. source
  • The same audit found that users do not prefer the political posts the engagement-based algorithm selects for them. source
  • In a 2023 field experiment on X, moving users from a chronological feed to the algorithmic feed increased engagement and shifted their political opinions towards more conservative positions. source
  • In that experiment neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship. source
  • The researchers found that X's algorithm promotes conservative content and demotes posts by traditional media. source
  • A 2026 model of ranking systems finds that weighting a feed towards likes and reshares raises engagement while also raising misinformation and ideological polarization. source
  • At Alphabet's 2022 annual meeting a stockholder proposal on algorithm disclosures drew 117,998,457 votes for and 483,701,300 against. filing
  • Trillium Asset Management told Alphabet shareholders that the algorithm disclosure proposal received 56% of non-insider votes at the 2022 annual meeting and had not been implemented. filing
  • The 2023 algorithm disclosures proposal at Alphabet was filed by the Trillium Asset Management ESG Global Equity Fund. filing
  • At Alphabet's 2023 annual meeting the algorithm disclosures proposal drew 2,177,846,989 votes for and 10,618,493,044 against. filing
  • As You Sow asked Meta shareholders in 2024 to vote for a report on prohibiting political advertising and on restoring the enhanced actions of the 2020 election cycle that reduced the amplification of false and divisive information. filing
  • That proposal drew 152,703,615 votes for and 4,855,808,278 against at Meta's 2024 annual meeting. filing
  • At Alphabet's 2026 annual meeting a proposal for a report on AI-generated misinformation drew 1,145,766,202 votes for and 11,179,823,623 against. filing
  • Trillium Asset Management urged Alphabet shareholders to vote for the 2022 algorithm disclosure proposal, filing its case with the Securities and Exchange Commission as an exempt solicitation. filing

Shareholder advocacy on record

  • 2022: Filer not established at Alphabet Inc. Supported by Trillium Asset Management, LLC. Asked the company to Algorithm disclosures, Proposal 15: quantitative and qualitative disclosure of how the company's algorithmic systems target and deliver content. 117,998,457 for, 483,701,300 against, of votes cast; Trillium reported the proposal as taking 56% of non-insider votes. Not implemented, as Trillium told shareholders in May 2023. Source
  • 2023: Trillium Asset Management ESG Global Equity Fund at Alphabet Inc. Asked the company to Algorithm disclosures, Proposal 12. 2,177,846,989 for, 10,618,493,044 against, of votes cast. We found no documented change. Source
  • 2024: Filer not established at Meta Platforms, Inc. Supported by As You Sow. Asked the company to Item 13, a report on prohibiting political advertising and on restoring the enhanced actions of the 2020 election cycle that reduced amplification of false and divisive information. 152,703,615 for, 4,855,808,278 against, of votes cast. We found no documented change. Source
  • 2022: Filer not established at Meta Platforms, Inc. Asked the company to Proposal 7, a report on the external costs of misinformation. 143,673,070 for, 5,118,039,014 against, of votes cast. Source
  • 2025: Filer not established at Meta Platforms, Inc. Asked the company to Proposal 8, a report on hate targeting marginalized communities. 727,562,405 for, 4,253,031,703 against, of votes cast. Source
  • 2025: Filer not established at Meta Platforms, Inc. Asked the company to Proposal 7, publication of voting results by class of shares. 1,028,098,566 for, 3,959,996,157 against, of votes cast. Not approved, so the split between founder and outside shares on every other proposal stays unpublished. Source
  • 2026: Filer not established at Alphabet Inc. Asked the company to Proposal 13, a report on AI-generated misinformation. 1,145,766,202 for, 11,179,823,623 against, of votes cast. Source

Already working on this

Trillium Asset Management (filer and public supporter)☆ FollowAs You Sow (public supporter)☆ Follow
3. Corporate and undisclosed political moneyHelpful

Companies involved: Meta Platforms (shareholders vote here); Alphabet (shareholders vote here); The Coca-Cola Company (shareholders vote here); UnitedHealth Group (shareholders vote here); Visa (shareholders vote here)

The review finds institutional and socioeconomic mechanisms comparatively thin in the causal literature, and the money that is documented is mostly money no shareholder reaches.

Outside groups spent $4.5 billion of non-party money in the decade after Citizens United, including $963 million from groups that do not disclose their funders, and the undisclosed part is the part no vote reaches. What a vote does reach is what a listed company gives and what it pays in trade association dues. Measuring corporate and trade group positions on lobbied bills against their giving finds their revealed preferences more conservative than their contributions imply, so a bipartisan donation record is not evidence of moderation, and the study concludes such interests may not reduce polarization overall. Shareholders have put this to votes: lobbying disclosure at Meta in 2023 and at Alphabet in 2022, a political contributions congruency report at UnitedHealth Group in 2022, and political expenditures values alignment at Coca-Cola in 2023, that last one filed by Clean Yield Asset Management for a client who holds the shares. Visa told its shareholders in 2022 that the Center for Political Accountability had scored its disclosures 100 and called it a Trendsetter, which is what the ask looks like once a company has met it.

Confirmed for this obstacle (9)
  • In the decade after Citizens United, outside groups spent $4.5 billion that was not party money, including $963 million by groups that do not disclose their funders. source
  • Measuring corporate and trade group positions on lobbied bills against their contributions finds their revealed preferences are more conservative than their giving implies. source
  • The same study concludes that a moderate contribution record need not imply moderate policy preferences and that such interests may not reduce polarization overall. source
  • A proposal for a report on lobbying disclosures at Meta's 2023 annual meeting drew 731,006,998 votes for and 4,289,286,570 against. filing
  • A lobbying report proposal at Alphabet's 2022 annual meeting drew 114,337,748 votes for and 488,159,890 against. filing
  • Clean Yield Asset Management filed the 2023 shareowner proposal on political expenditures values alignment at The Coca-Cola Company on behalf of its client Julie Kalish. filing
  • That proposal stood as Item 7 on the ballot at Coca-Cola's 2023 annual meeting. filing
  • A shareholder proposal for a political contributions congruency report stood as Proposal 5 on the ballot at UnitedHealth Group's 2022 annual meeting. filing
  • Visa told shareholders that the Center for Political Accountability assessed its disclosures in 2022 for the CPA-Zicklin Index and designated the company a Trendsetter with a score of 100. filing

Shareholder advocacy on record

  • 2023: Filer not established at Meta Platforms, Inc. Asked the company to Proposal 6, a report on lobbying disclosures. 731,006,998 for, 4,289,286,570 against, of votes cast. Source
  • 2022: Filer not established at Alphabet Inc. Asked the company to Proposal 5, a lobbying report. 114,337,748 for, 488,159,890 against, of votes cast. Source
  • 2023: Clean Yield Asset Management at The Coca-Cola Company Asked the company to Item 7, political expenditures values alignment, filed for a client who holds the shares. Not established in the documents read. Source
  • 2022: Filer not established at UnitedHealth Group Asked the company to Proposal 5, a political contributions congruency report. Not established in the documents read. Source
  • 2022: Filer not established at Visa Supported by Center for Political Accountability. Asked the company to Disclosure of political contributions and trade association payments, scored on the CPA-Zicklin Index. Visa reported a score of 100 and a Trendsetter designation. Source

Already working on this

Clean Yield Asset Management (filer)☆ FollowCenter for Political Accountability (index publisher)☆ Follow
4. Economic inequalityHelpful

No corporate lobbying against measures that would narrow the distribution was found in the documents read for this obstacle.

The review places socioeconomic mechanisms among the thinnest in the causal literature, and the strongest series available here runs both ways between inequality and polarization.

Nine decades of United States data relate a precise measure of inequality to polarization, while a cruder measure shows no such relation, and the feedback runs in both directions. What listed companies pay their workers and their executives is part of the income distribution, so this is not an outcome no company touches. The share of that distribution sitting with companies a shareholder vote reaches is not established in the documents read, and on the evidence here the connection matters at the margin rather than sitting on the critical path.

Confirmed for this obstacle (1)
  • Nine decades of time series show a more precise measure of inequality related to polarization, with feedback running in both directions. source
5. Partisan identity and out-group animosityIndependent

The systematic review of 135 causal studies from 2012 to 2025 finds psychological and identity explanations dominating the literature, which is why this sits above the institutional obstacles on severity while the vote reaches none of it.

Partisanship has become one of Americans' most salient social identities, with partisans more negative towards the opposing party on both explicit and implicit measures, and those biases carrying into everyday decisions. Nothing a company makes, sells or employs decides which party a voter takes as their own. The media obstacles above feed this one, and they are rated where they are rather than here; the identity itself is not a thing a resolution at an annual meeting can reach.

Confirmed for this obstacle (2)
  • Partisanship has become one of Americans' most salient social identities, with partisans more negative towards the opposing party on both explicit and implicit measures. source
  • A systematic review of 135 causal studies published between 2012 and 2025 reports that psychological, media and identity explanations dominate the literature while institutional and socioeconomic mechanisms remain comparatively thin. source
6. Partisan geographic and social sortingIndependent

Measured now for every registered voter rather than by survey, which is why it ranks high on evidence, and the deciding actors are households rather than institutions.

A measurement covering more than 180 million registered voters finds a large share of them living with virtually no exposure to voters of the other party where they live, and tracking every registered voter in states recording party registration between 2008 and 2020 shows that segregation rising year over year at every level from congressional districts down to census blocks. The actors here are named rather than assumed: tens of millions of households choosing where to live, each for their own reasons, and no shareholder vote reaches any of them.

Confirmed for this obstacle (2)
  • A measurement covering more than 180 million registered voters finds that a large proportion live with virtually no exposure to voters of the other party where they live. source
  • Tracking every registered voter in states recording party registration between 2008 and 2020 shows geographic segregation between Democrats and Republicans rising year over year at every level from congressional districts to census blocks. source
7. Nomination systemsIndependent

Placed low on severity by the evidence rather than by its prominence in debate: most studies of primary elections find small to no effect on how members behave.

Concerns about polarization and about how hard it is for moderate candidates to win primaries have driven several reform efforts in recent decades, and most empirical evidence on the influence of primary elections finds small to no effect on member behavior. Nomination rules are written by state legislatures and by the parties themselves, so the instrument of change is a statute or a ballot measure rather than a resolution at an annual meeting.

Confirmed for this obstacle (2)
  • Most empirical evidence on the influence of primary elections finds small to no effect on how members of Congress behave. source
  • Concerns about polarization and about the difficulty moderate candidates have in winning primaries have driven several electoral reform efforts in recent decades. source
8. District design and Senate malapportionmentIndependent

Simulation against non-partisan baselines puts the national partisan effect of gerrymandering at about two seats, which is why this ranks last on severity among the eight.

Partisan gerrymandering was widespread in the 2020 redistricting cycle, and most of the electoral bias it created cancels at the national level, giving Republicans two additional seats on average. The Senate is rare internationally for disregarding population differences among the units it represents, and reweighting 804 key Senate votes taken between 1961 and 2019 shows a systematic effect from state equality. Lines are drawn by legislatures and commissions and the Senate's composition is written into the Constitution. None of those actors has shareholders.

Confirmed for this obstacle (2)
  • Simulating alternative maps against the enacted ones finds partisan gerrymandering widespread in the 2020 redistricting cycle, with most of the electoral bias it creates cancelling at the national level. source
  • The United States Senate is rare internationally for disregarding population differences among the units it represents, and reweighting 804 key Senate votes taken between 1961 and 2019 shows a systematic effect from state equality. source

What owners can do

  • Read the proxy statements of the station groups in your portfolio. Sinclair, Nexstar and Gray Media each put a vote on every common share, and no resolution on editorial standards or on the nationalization of local news has yet been put to any of them.
  • Vote alongside Trillium Asset Management on algorithm disclosure at Alphabet, where the 2022 proposal took 56% of non-insider votes and has still not been implemented.
  • Vote alongside As You Sow at Meta on political advertising and on the election-cycle measures the company applied in 2020 and later dropped.
  • Support publishing Meta's and Alphabet's voting results by share class, which is how a member finds out whether the shares they voted were outvoted by the founders or by other outside holders.
  • Vote for political spending disclosure and congruency reports at every company you hold, which is the only part of election money a shareholder reaches at all.

Where this sits on our maps

Organizations to know

  • As You Sow (shareholder advocacy on record) It filed the notice urging Meta's shareholders to vote for a report on political advertising and on restoring the 2020 election-cycle brakes on amplification.☆ Follow
  • Trillium Asset Management (shareholder advocacy on record) It filed the algorithm disclosure proposals at Alphabet and reported the 2022 result as 56% of non-insider votes, unimplemented.☆ Follow
  • Clean Yield Asset Management (shareholder advocacy on record) It filed the 2023 political expenditures values alignment proposal at Coca-Cola for a client who holds the shares.☆ Follow
  • Center for Political Accountability (shareholder advocacy on record) Its CPA-Zicklin Index is the scale companies report themselves against on political disclosure, and Visa cited its 2022 score to shareholders.☆ Follow
  • Sinclair (publicly listed) 191 local stations reaching almost 40% of the population as of 2019, and 48,254,031 Class A shares each carrying one vote in 2026.
  • Nexstar Media Group (publicly listed) 30,538,965 common shares outstanding in 2026, each carrying one vote.
  • Gray Media (publicly listed) One vote for each common share and ten for each Class A share in 2026.
  • Fox Corporation (publicly listed) Owner of the cable channel whose ratings the 2024 study ties to ideology, party identity and vote share.☆ Follow
  • Meta Platforms (publicly listed) Runs the feeds the amplification obstacle is about, and the company whose shareholders have voted on political advertising, misinformation and hate content.☆ Follow
  • Alphabet (publicly listed) Runs the other dominant ranking systems, and the company where algorithm disclosure took a majority of non-insider votes in 2022.☆ Follow

Sources

  1. Engagement, user satisfaction, and the amplification of divisive content on social media | PNAS Nexus | Oxford Academic
  2. The political effects of X’s feed algorithm | Nature | Springer Nature Link
  3. Ranking for engagement: How social media algorithms fuel ...
  4. Alphabet Inc. 8-K 2022-06-03 (U.S. Securities and Exchange Commission)
  5. https://www.sec.gov/Archives/edgar/data/1652044/000121465923006504/r54230px14a6g.htm
  6. Alphabet Inc. 8-K 2023-06-08 (U.S. Securities and Exchange Commission)
  7. Annual Meeting: May 29, 2024
  8. Meta Platforms, Inc. 8-K 2024-05-31 (U.S. Securities and Exchange Commission)
  9. Alphabet Inc. (GOOG, GGLAP, GGLBP, GOOGL, GOOGM, GOOGN) (CIK 0001652044) 8-K 2026-06-11 (U.S. Securities and Exchange Commission)
  10. From viewers to voters: Tracing Fox News' impact on ...
  11. Media Consolidation Means Less Local News, More Right ... (Edmund L. Andrews)
  12. Does Local News Stay Local?: Online Content Shifts in Sinclair-Acquired Stations
  13. Media Consolidation & Political Polarization (Mary R. Hornak)
  14. Sinclair, Inc. DEF 14A 2026-04-23 (U.S. Securities and Exchange Commission)
  15. GRAY MEDIA, INC DEF 14A 2026-03-26 (U.S. Securities and Exchange Commission)
  16. NEXSTAR MEDIA GROUP, INC. DEF 14A 2026-04-30 (U.S. Securities and Exchange Commission)
  17. Partisanship as Social Identity; Implications for the Study of Party Polarization (Shanto Iyengar, Masha Krupenkin)
  18. What drives affective polarization? Asystematic literature review
  19. The measurement of partisan sorting for 180 million voters | Nature Human Behaviour
  20. Sources and Extent of Rising Partisan Segregation in the U.S. – Evidence from 143 Million Voters (Jacob Brown, Enrico Cantoni, Ryan Enos, Vincent Pons, Emilie Sartre)
  21. More money, less transparency: A decade under Citizens United • OpenSecrets
  22. Moderation or Strategy? Political Giving by Corporations and Trade Groups
  23. Meta Platforms, Inc. (META) (CIK 0001326801) 8-K 2023-06-02 (U.S. Securities and Exchange Commission)
  24. F49231px14a6g.htm
  25. COCA COLA CO (KO) (CIK 0000021344) DEF 14A 2023-03-10 (U.S. Securities and Exchange Commission)
  26. UNITEDHEALTH GROUP INC (UNH) (CIK 0000731766) DEF 14A 2022-04-22 (U.S. Securities and Exchange Commission)
  27. VISA INC. (V) (CIK 0001403161) DEF 14A 2022-12-01 (U.S. Securities and Exchange Commission)
  28. Institution of Nomination and the Policy Ideology of Primary Electorates (Seth J. Hill)
  29. Top-Four Primaries Help Moderate Candidates via Crossover Voting: The Case of the 2022 Alaska Election Reforms (Sarah E. Anderson, Daniel M. Butler, Laurel Harbridge‐Yong, Renae Marshall)
  30. Widespread Partisan Gerrymandering Mostly Cancels Nationally, but Reduces Electoral Competition (Christopher T Kenny, Cory McCartan, Tyler Simko, Shiro Kuriwaki, Kosuke Imai)
  31. The Conservative Policy Bias of US Senate Malapportionment | PS: Political Science & Politics | Cambridge Core (Richard Johnson Affiliation: Queen Mary, University of London, UK)
  32. Income Inequality and Political Polarization: Time Series Evidence Over Nine Decades (Federal Reserve Bank of Dallas, John V. Duca, Jason L. Saving)
  33. Meta Platforms, Inc. 8-K 2025-05-30 (U.S. Securities and Exchange Commission)
  34. Meta Platforms, Inc. (META) (CIK 0001326801) 8-K 2022-05-27 (U.S. Securities and Exchange Commission)
  35. B55220px14a6g.htm

About this report

The rating answers two questions: how far the outcome depends on companies changing what they make, how they make it and what they sell, and how far it depends on companies dropping their political influence against it. The higher answer sets the rating: Independent, Helpful, Necessary or Pivotal.

This is the deep report. A frontier model researched it against documents captured as it read them, and every fact it recorded carries an excerpt from one of those documents. Each excerpt was matched to its document twice: on the machine that did the work, and again here. The ratings are analytical judgments under our method, not findings of any organization named here.

56 claims were checked against the document each cites: 56 held and 0 did not.

What we read. 136 documents were held for this report and every quotation was checked against them.

Every claim, and what became of it (56)
  • Confirmed · In a preregistered audit, Twitter's engagement-based ranking amplified emotionally charged, out-group hostile content that users said made them feel worse about their political out-group, compared wit
  • Confirmed · The same audit found that users do not prefer the political posts the engagement-based algorithm selects for them.
  • Confirmed · In a 2023 field experiment on X, moving users from a chronological feed to the algorithmic feed increased engagement and shifted their political opinions towards more conservative positions.
  • Confirmed · In that experiment neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship.
  • Confirmed · The researchers found that X's algorithm promotes conservative content and demotes posts by traditional media.
  • Confirmed · A 2026 model of ranking systems finds that weighting a feed towards likes and reshares raises engagement while also raising misinformation and ideological polarization.
  • Confirmed · At Alphabet's 2022 annual meeting a stockholder proposal on algorithm disclosures drew 117,998,457 votes for and 483,701,300 against.
  • Confirmed · Trillium Asset Management told Alphabet shareholders that the algorithm disclosure proposal received 56% of non-insider votes at the 2022 annual meeting and had not been implemented.
  • Confirmed · The 2023 algorithm disclosures proposal at Alphabet was filed by the Trillium Asset Management ESG Global Equity Fund.
  • Confirmed · At Alphabet's 2023 annual meeting the algorithm disclosures proposal drew 2,177,846,989 votes for and 10,618,493,044 against.
  • Confirmed · As You Sow asked Meta shareholders in 2024 to vote for a report on prohibiting political advertising and on restoring the enhanced actions of the 2020 election cycle that reduced the amplification of
  • Confirmed · That proposal drew 152,703,615 votes for and 4,855,808,278 against at Meta's 2024 annual meeting.
  • Confirmed · At Alphabet's 2026 annual meeting a proposal for a report on AI-generated misinformation drew 1,145,766,202 votes for and 11,179,823,623 against.
  • Confirmed · A ranking rule is not a market condition. It is written, weighted and changed inside the company that runs the feed, which is why the design sits with the people who own it.
  • Confirmed · Fox News viewership shifted American ideology and party identity rightward.
  • Confirmed · A 0.05 rise in Fox News ratings increased Republican vote shares by 0.5% in elections.
  • Confirmed · Local television news drew 25 million nightly viewers, more than national cable programmes such as Fox News and MSNBC, when the study was published in 2019.
  • Confirmed · As of 2016 five companies controlled 37% of local television stations in the United States.
  • Confirmed · Sinclair Broadcast Group owned 191 stations reaching almost 40% of the United States population when the study was published in 2019.
  • Confirmed · The researchers tested whether the rightward slant of acquired stations followed audience tastes or the owner's preferences, and concluded that ownership preferences influence content.
  • Confirmed · Stations acquired by Sinclair reported more often on national news at the expense of local topics and increased their coverage of polarizing national topics.
  • Confirmed · Congress has required the Federal Communications Commission to review broadcast ownership regulations since 1996, which is a remit over who may own stations rather than over what they broadcast.
  • Confirmed · Sinclair's 2026 proxy statement records 48,254,031 Class A shares, each carrying one vote on each proposal at the annual meeting.
  • Confirmed · Gray Media's 2026 proxy statement records one vote for each common share and ten votes for each Class A share.
  • Confirmed · Nexstar Media Group's 2026 proxy statement records 30,538,965 common shares outstanding, each carrying one vote.
  • Confirmed · What a broadcaster puts on air is its own conduct rather than a condition acting on it, and no regulator sets it: the ownership rules reach who may hold the license, not what the license holder says.
  • Confirmed · Audience demand is a preference rather than an actor, so it cannot carry the claim that this obstacle sits beyond a shareholder vote, and the study that tested the two explanations against each other
  • Confirmed · Partisanship has become one of Americans' most salient social identities, with partisans more negative towards the opposing party on both explicit and implicit measures.
  • Confirmed · A systematic review of 135 causal studies published between 2012 and 2025 reports that psychological, media and identity explanations dominate the literature while institutional and socioeconomic mech
  • Confirmed · Nothing a company makes, sells or employs decides which party a voter takes as their own, and no shareholder vote reaches the identity itself.
  • Confirmed · A measurement covering more than 180 million registered voters finds that a large proportion live with virtually no exposure to voters of the other party where they live.
  • Confirmed · Tracking every registered voter in states recording party registration between 2008 and 2020 shows geographic segregation between Democrats and Republicans rising year over year at every level from co
  • Confirmed · Where a household lives is decided by the household, and the actor beyond a shareholder vote here is named rather than assumed: tens of millions of movers, each choosing for their own reasons.
  • Confirmed · In the decade after Citizens United, outside groups spent $4.5 billion that was not party money, including $963 million by groups that do not disclose their funders.
  • Confirmed · Measuring corporate and trade group positions on lobbied bills against their contributions finds their revealed preferences are more conservative than their giving implies.
  • Confirmed · The same study concludes that a moderate contribution record need not imply moderate policy preferences and that such interests may not reduce polarization overall.
  • Confirmed · A proposal for a report on lobbying disclosures at Meta's 2023 annual meeting drew 731,006,998 votes for and 4,289,286,570 against.
  • Confirmed · A lobbying report proposal at Alphabet's 2022 annual meeting drew 114,337,748 votes for and 488,159,890 against.
  • Confirmed · Clean Yield Asset Management filed the 2023 shareowner proposal on political expenditures values alignment at The Coca-Cola Company on behalf of its client Julie Kalish.
  • Confirmed · That proposal stood as Item 7 on the ballot at Coca-Cola's 2023 annual meeting.
  • Confirmed · A shareholder proposal for a political contributions congruency report stood as Proposal 5 on the ballot at UnitedHealth Group's 2022 annual meeting.
  • Confirmed · Visa told shareholders that the Center for Political Accountability assessed its disclosures in 2022 for the CPA-Zicklin Index and designated the company a Trendsetter with a score of 100.
  • Confirmed · Disclosure votes reach the money a listed company gives and the trade association dues it pays. They do not reach the undisclosed spending, which is the larger part of the problem and sits with actors
  • Confirmed · Most empirical evidence on the influence of primary elections finds small to no effect on how members of Congress behave.
  • Confirmed · Concerns about polarization and about the difficulty moderate candidates have in winning primaries have driven several electoral reform efforts in recent decades.
  • Confirmed · Nomination rules are written by state legislatures and by the parties themselves, and the reform on offer is a ballot measure or a statute rather than a resolution at an annual meeting.
  • Confirmed · Simulating alternative maps against the enacted ones finds partisan gerrymandering widespread in the 2020 redistricting cycle, with most of the electoral bias it creates cancelling at the national lev
  • Confirmed · The United States Senate is rare internationally for disregarding population differences among the units it represents, and reweighting 804 key Senate votes taken between 1961 and 2019 shows a systema
  • Confirmed · District lines are drawn by legislatures and commissions and the Senate's composition is written into the Constitution, so the actors are named and none of them has shareholders.
  • Confirmed · Nine decades of time series show a more precise measure of inequality related to polarization, with feedback running in both directions.
  • Confirmed · What listed companies pay their workers and their executives is part of the income distribution, and the share of that distribution sitting with companies a vote reaches is not established in the docu
  • Confirmed · At Meta's 2025 annual meeting a proposal to disclose voting results by class of shares drew 1,028,098,566 votes for and 3,959,996,157 against.
  • Confirmed · At the same meeting a proposal for a report on hate targeting marginalized communities drew 727,562,405 votes for and 4,253,031,703 against.
  • Confirmed · Meta's 2022 annual meeting held a vote on a report on the external costs of misinformation, which drew 143,673,070 votes for and 5,118,039,014 against.
  • Confirmed · A proposal that a majority of outside shares supports can still lose where founders hold shares carrying ten votes each, which is why the reach of a vote at these two companies is a question about the
  • Confirmed · Trillium Asset Management urged Alphabet shareholders to vote for the 2022 algorithm disclosure proposal, filing its case with the Securities and Exchange Commission as an exempt solicitation.

First drafted September 25, 2026, reviewed and republished September 25, 2026. Scoring method revision 1.6.

Which passes have run, and what ran them

The passes can run on different models, so each one records its own. Where a model was asked for by a name that moves, the exact build that answered is shown.

  • The deep run · September 25, 2026 · Read 17 documents and wrote 56 claims, each with an excerpt from one of them. The run's own verifier repaired 0 and dropped 0 over 4 passes.local runner · claude-opus-5

Build the MovementExplore what you care about