Home › About
Proxy Votes Now shows how funds and asset managers voted at U.S. company shareholder meetings, beside the result the company itself reported.
It joins two SEC filings. On Form N-PX, funds and institutional managers disclose how they voted. On Form 8-K, under Item 5.07, a company reports the result of its meeting. For each ballot item, a company page shows the reporting funds’ votes and, where an 8-K has been matched, the official tally, with a link to each filing on SEC EDGAR. Ownership figures from Forms 13F, 13D and 13G and the proxy statement are shown beside them.
Manager pages show each asset manager’s record for a season: how often it voted against the board on the board’s own proposals, and how often it backed environmental and social proposals that shareholders put on the ballot.
The voting policies section collects the proxy-voting guidelines that asset managers publish, with each annual edition kept as a PDF, and a cross-manager analysis of how the 2026 policies changed.
Votes, official results and ownership come from public filings on SEC EDGAR. Voting policies are the documents the managers publish on their own websites. The methodology below explains how each figure is derived, and the coverage page shows, season by season, how many meetings have a matched official result.
New N-PX filings are collected every week, on Saturday night (UTC). New 8-K results and share counts are collected every Sunday, and company pages are rebuilt from them over the following days. Schedule 13D filings are collected daily. N-PX filings for a season that ends on June 30 are due by August 31, so most of a season’s fund votes arrive in September, and late filings keep arriving after that.
Short explainers on how shareholder votes are counted and what the SEC filings behind this site show.
The upcoming votes pages list shareholder meetings from the proxy statements that companies and funds file with the SEC (Forms DEF 14A, DEFR14A, DEFC14A and DEFM14A, with the additional materials filed on Forms DEFA14A and DFAN14A), collected every day.
The meeting date, record date and ballot are read from the proxy statement by software. Before a ballot is shown it is checked against the proxy’s own statements: the items must be numbered without gaps and agree with the number of items the notice and the proxy card give, and every item must have a title and a board recommendation. When the check does not pass, the page gives the meeting facts and says “Ballot: see the proxy statement”, with a link, instead of a list that may be wrong. In a test on 129 proxy statements from the 2026 season, checked by hand, the ballots that passed had no wrong board recommendation and no missing or extra item, and one wrong meeting date. Item descriptions and the names of shareholder proponents were sometimes wrong, so an item’s kind is labelled only for the kinds that were read reliably, and a proponent is named only when one was found in the proxy. Which funds vote on each item of a fund meeting is not shown; the proxy statement says.
Last year’s results are the company’s own Form 8-K tallies, counted as described below; how the largest managers voted last year comes from their Form N-PX filings. None of it is a prediction: the pages say how the board recommends voting and what happened last year, not how any fund or shareholder will vote, and they carry no proxy-adviser recommendations.
To count its readers, this site uses its own small script. It sets no cookies, stores nothing in your browser and sends nothing to anyone else, and it counts nothing at all if your browser sends Do Not Track or Global Privacy Control. Once a page has been open for a few seconds, or you scroll or click, it reports the page, the name of the site that linked you here, your screen width, your language, how long the page stayed open and how far you scrolled. From your IP address the server works out your country and, outside the European Union, the EEA, the UK and Switzerland, your region and the organization that owns your network, such as a company or a university (never a home or mobile internet provider), from public internet registry records. For a business internet line outside the EU, the EEA, the UK and Switzerland, the server may also ask the public internet registry, through rdap.org, who the block of addresses the line belongs to is registered to: it sends the address of the block, never your own address. Visitors from the EU, the EEA, the UK and Switzerland are recorded by country only. Your IP address is not stored. A code that changes every day counts unique visitors per day without following anyone from one day to the next; the browser is recorded only as its name and device type. Individual page views are deleted after 30 days and only daily totals are kept. IP Geolocation by DB-IP.
The tables this site derives from SEC filings — the per-item and per-manager vote aggregates, the official results as tabulated here, and the CSV downloads — are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You may copy, adapt and republish them, including commercially, provided you credit Proxy Votes Now and link to the page you used. The underlying filings are public records of the U.S. Securities and Exchange Commission (SEC website policy); company names and trademarks belong to their owners. The licence covers the data, not the site’s design or code.
How to cite. Name the site, the page title, its URL and the date shown on the page, for example: Proxy Votes Now, “Pfizer Inc. 2025-2026 Proxy Voting Records”, https://proxyvotesnow.com/company/pfizer-inc/2026 (version of November 8, 2026). Where a figure is an official result, cite the company’s Form 8-K that the page names as well. Every company page carries its own “Cite this page” line.
Questions, and figures that do not match the filing they cite, go to data@proxyvotesnow.com.
How each figure on the site is derived, so that you can check it.
Some summaries and comparisons are produced with the help of automated tools, then checked against the source filings and documents; policy-change comparisons are also reviewed before publication.
Form N-PX is the annual SEC filing in which registered funds disclose how they voted on every ballot item at every company they held. Institutional investment managers also file it, to report their executive-pay votes. The per-fund and per-manager votes on this site are those filings, as filed.
N-PX does not follow the calendar year. Each filing covers meetings held from July 1 to June 30 and is due by August 31. A year on this site is therefore a proxy season: “2026” means the 2025–2026 season, that is, meetings from July 1, 2025, to June 30, 2026. A meeting held in the autumn of 2025 is in the 2026 season, not 2025.
N-PX is not a census of shareholders. It covers the institutions required to file it. Individuals, most foreign holders, insiders and many other investors do not file. A total of fund votes is the position of the reporting funds only. It is never the meeting result, and the site does not present it as one.
A U.S. public company reports the outcome of a shareholder meeting on Form 8-K under Item 5.07. Where this site shows an official tally for a ballot item, the figures were read from that filing, and the result links to the filing on SEC EDGAR so you can compare the two directly.
A company is identified by the issuer number in its CUSIP, the first six characters of the security identifier that fund filers report, not by its name. An 8-K is not used for a company when the SEC registrant that filed it is, by its current and former registered names, a different company from the issuer of that security. The same check applies to the share counts in section 5. Companies that report to the SEC as foreign private issuers file no Form 8-K, so their pages usually show no official result. Some companies incorporated abroad, such as TE Connectivity, file 8-Ks like U.S. companies and are covered.
Mistyped CUSIPs. Some filers report a CUSIP that cannot be right: it fails the check digit every CUSIP carries, or a spreadsheet has dropped its leading zeros (Amazon’s 023135106 filed as 23135106). We read such a value as another security only when one small slip explains it — restoring one or two leading zeros, changing one character, or swapping two neighbouring characters — and when exactly one security that already appears often in the data, reached that way, held a meeting on the same date. Anything else is left as filed. The CUSIP the filer reported is kept unchanged; the correction is stored beside it.
Where no 8-K has been matched to a meeting, the page says so and shows only the reporting funds’ own total, labelled as fund support with no official result. No outcome is stated in that case.
The 8-Ks are found in two ways: by searching SEC EDGAR’s full-text index, week by week, for every report filed under Item 5.07 since June 2023, and from the company’s own list of filings. Each week of filings is searched again until the search has completed for it. An amended report (8-K/A) for the same meeting is kept beside the original.
Some reports set out their figures in a way that is not read as a table at first: the whole report laid out as one page-wide table, the results written as lines such as “For: 44,264,494” under each nominee, or the results placed in an exhibit attached to the 8-K. Such a report is read a second time in that layout. The result is used only when every item read has a vote for and a vote against or withheld, and the items read account for at least four-fifths of the vote figures in the report; otherwise the meeting is shown with no official result rather than with part of one. Results read from an exhibit link to the exhibit. An 8-K that another meeting of the same company already cites is not used again.
One entry per ballot item. Fund filers describe the same ballot item in different words. Where a meeting has a matched 8-K, each item in the company’s tally is shown once, with its official result. A fund-reported description is joined to an item only on evidence: the same text, the same director nominee, or a match to that item’s row of the tally. A smaller description matched to a row that already has one is listed under that item with its own fund figures, and is not added into them. Nothing is joined because two descriptions merely read alike.
A description that matches no row of the tally, and that far fewer managers reported than the items on the ballot (under a quarter of a typical item at that meeting), is most likely one of those items in other words. Such descriptions are not shown on the company page; they are kept, with their own figures, in the page’s data download (CSV), marked unreconciled, together with lines reported under a date on which no meeting is on file. They are not counted in the number of items, in the page’s summary of its items or in any comparison between items. Nor are they counted in the page’s totals of asset managers and fund votes, which count only the items shown and the descriptions listed under them. A description that many managers reported but that matches no row is shown as an item of its own, with no official result: usually the tally as read here is missing that item.
A season’s filings sometimes also report a meeting held outside its July 1 to June 30 window. Those lines are not shown on that season’s page and are not counted in its number of items; the page says how many there were and which season the meeting belongs to. They are not in its totals of asset managers and fund votes either.
Status labels. Each item on a company page carries one label saying how far its result has been checked against the company’s own filing:
A second label, Fund totals may overlap, is added where the item’s fund votes come from reports that can cover the same shares: fund reports together with institutional-manager reports, or fund shares that add up to more than the shares outstanding or the votes cast at the meeting. No combined fund share total is shown for such an item. The label concerns the funds’ figures only; an official result is not affected. Labels describe what is on the page; they change no figure.
A few companies report Item 5.07 in percentages instead of vote counts. Those figures are shown as the company published them, because the filing does not always say what the percentage is a percentage of.
How many meetings have a matched 8-K, season by season, is on the coverage page.
Support = FOR ÷ (FOR + AGAINST) Captioned “of votes cast” wherever it appears.
This is the basis most companies and most press coverage use, but it is not the only one. The same vote can be quoted as a share of votes cast, of shares present, or of shares outstanding, and the three can differ widely. If a figure here disagrees with one you have seen elsewhere, check the denominator first. A figure on this site that is a share of shares outstanding is labelled as such.
On the asset manager pages. Every figure is a count of ballot items, one fund’s vote on one proposal at one meeting, never of shares.
An outcome is shown as Approved or Not approved only where the 8-K itself says so: in the item’s heading (“The appointment of KPMG LLP ... was ratified”), in the sentence that follows its tally (“Based on the votes set forth above, the proposal was not approved”), or in a list at the top of the report (“Shareholders approved Proposals 1 through 4”). The sentence it was read from is kept with the result. A sentence about another item, or one that names several items without saying which outcome is which, is not used.
Calculated outcomes. Where the filing gives the vote counts but does not state the outcome in words, the page shows Passed (calculated) or Failed (calculated): the company’s filing reports the vote counts but does not state the outcome, and the site calculates it from the counts using the voting standard for that item where the company’s 8-K, or failing that its proxy statement, states one (a majority of the votes cast: votes for exceeded votes against; a majority of the shares present or represented: votes for exceeded votes against plus abstentions), and otherwise by whether votes for exceeded votes against. The standard used is named beside each calculated outcome. Where an item needed a majority of all shares outstanding or a supermajority, the counts alone cannot show that it passed, so no outcome is calculated unless votes for did not exceed votes against; an election decided by plurality gets no calculated outcome. Wherever the filing states the outcome, its own words are shown instead. The stored data are not changed: the label is worked out, on display, from the counts shown beside it.
A simple majority of votes cast is not always the legal standard. Some items need a supermajority, or a majority of all shares outstanding, or a separate vote of each share class, and director elections may be decided by plurality. Where a filing says an item was not approved although FOR was more than half of the votes cast, the page shows the filing’s word beside the tally. Where the filing states the standard that applied, the page quotes it. Where it does not, and the item is of a kind that often needs more than a majority of votes cast (a charter or bylaw amendment, a merger, an increase in authorised shares), the page notes that a stricter requirement may have applied; that note is the site’s inference, not the filing’s. Where the filing’s word and its own numbers cannot both be right and the filing gives no reason, both are shown and the disagreement is marked; the site does not choose between them.
In a director election the site records “elected” or “not elected” only where the filing says so, including the list of winners in a contested election. A nominee who received less than a majority of votes cast under a resignation policy is neither, unless the filing says which. Where the filing names as many winners as it says there were seats (“Election of four directors ... A, B, C and D were elected”), the other nominees are recorded as not elected.
Preliminary and final results. Some companies first file preliminary results, for example “the estimated preliminary voting results reported by the Company’s proxy solicitor”, and later file the final count in an amended 8-K (8-K/A). Where the final 8-K/A is on file, its figures are used and the page names the 8-K/A as the source. Where only the preliminary filing exists, its figures are kept and recorded as preliminary, as the filing itself describes them.
Results at a glance and dissent. The top of each company page summarises the meeting from the company’s own Form 8-K: how many items it reports, and how many it states, or the site calculates, as passed or failed; the company’s items it reports as not approved or not elected; and the items with notable, significant or high dissent. Dissent is judged on the 8-K tally, by type of item. Support is FOR ÷ (FOR + AGAINST), or FOR ÷ (FOR + WITHHELD) on a withhold ballot; against is the rest of those votes; abstentions and broker non-votes are in neither. A shareholder proposal with 15% support or more has notable dissent, 25% significant and 40% high. A director nominee with 10% or more against or withheld is notable, 20% significant. Say-on-pay under 80% support is notable, under 70% significant. An equity or compensation plan with 15% or more against is notable, 25% significant. Any other item the board put forward with 15% or more against is notable. The summary lists the items in each tier. On the search page these thresholds can be changed under “Dissent thresholds”; the company pages always use the ones above. In a contested election the nominees are listed together, each marked as the company’s or a shareholder’s nominee. Votes on how often to hold say-on-pay are not judged. The funds’ votes reported on Form N-PX cover only some of the shareholders and are never used for this, so a meeting with no 8-K tally on file gets no dissent judgement.
Holder and ownership figures come from three kinds of filing, and they do not measure the same thing. Each row on a company page names the form it came from.
None of these is shares voted. Company pages do not yet show each row’s as-of date, and some rows come from filings dated after the meeting, so read them as the latest disclosures held for that season, not as positions on the meeting date. Dating each row and choosing the disclosure nearest before the meeting is planned.
Percentages of a company are computed against a share count for the same meeting year as the holding, taken from the company’s own SEC filings. Where no reliable share count is available, no percentage is shown. A season can hold a second meeting in another calendar year (a special meeting, or a late report of an earlier one); its items are measured against that year’s own share count, and where none is on file no comparison with shares outstanding is shown for them.
To check an official result, open the 8-K linked beside it and find Item 5.07. The vote counts on this site should match the filing exactly, and the support percentage is the arithmetic in section 3 applied to those counts.
To cite a figure, cite the page URL together with the SEC filing it names, with the filing’s form type and date. Cite the season as a range (for example “the 2025–2026 proxy season”) so that a reader does not take it for a calendar year. For a fund vote, the primary source is that filer’s Form N-PX on SEC EDGAR.
Company pages are rebuilt periodically and the current season keeps changing as late filings arrive, so note the date you retrieved a figure.
Downloads. Each company page offers its data as two CSV files. One has a row per ballot item: the company’s Form 8-K tally, the outcome in the filing’s own words, the support percentage with the two numbers it is computed from, the stated voting standard, and the summed fund votes. The other has a row per asset manager per item: shares by direction, the number of funds, and the vote values exactly as filed. Every row links its SEC filing. The column guide is at company-csv-README.txt.
Form N-PX is filed by each fund registrant and each institutional manager, not by the firm a reader thinks of as the manager. One firm can file under many names, and one registrant can file for funds of many unrelated firms. The manager pages therefore group the filers before counting.
Percentages are rounded half up: 99.15 is shown as 99.2 and 99.14 as 99.1. A figure that is not exactly 100% is never shown as 100, and one that is not exactly 0% is never shown as 0: they stop at 99.9 and 0.1, because “100% FOR” would say that nobody voted otherwise.
Questions, corrections and feedback are welcome at data@proxyvotesnow.com.
If a figure on this site does not match the SEC filing it cites, please send the page address and the filing, and say which number differs. The filing is the authority, and a reported mismatch is the most useful thing a reader can send.