This log starts on 10 October 2026. Corrections made
before that date are not listed.
This page lists errors in the figures and statements published on
proxyvotesnow.com that have been corrected. Each entry says what was wrong,
how much of the site it affected, what changed, and which section of the
methodology describes the rule that now applies.
Newest first. The date is the day (UTC) the correction reached the live
site; company pages are redrawn within a day of it.
If a figure on this site does not match the SEC filing it cites, please
tell us.
Vote results re-read from companies’ 8-K filings; outcomes shown only where the filing states them
What was wrong. The software that reads shareholder vote results from Form 8-K Item 5.07 sometimes shifted figures by one column (for example Illinois Tool Works 2024 and 2025), gave an item the title of the item next to it (for example Fastenal, Applied Materials, NetApp, Disney and Domino’s), missed rows (Paramount Group 2024), or read text that is not a name as a director. Pages also showed a pass or fail result worked out from a simple majority of votes cast even where the filing said otherwise. For example, Air Products’ 2025 Proposal 4 received 63% of votes cast and was shown as “Majority: yes” with no result from the filing; the company reports that it was not approved because it needed more than a majority of votes cast.
Scope. 12,299 meetings were re-read. About 2,600 of them gained vote tallies they had lacked and about 4,950 had at least one item’s result changed. 211 meetings that showed “no results” now show the filing’s results. Company pages whose outcome text contradicted their label fell from 32 to 5.
What changed. Results are read with a new reader. An outcome is shown as Approved or Not approved only where the filing states it, and the sentence it was read from is kept. Where the filing gives only counts, the page says whether FOR was more than half of the votes cast and does not call that the result. Where a filing’s stated outcome and its own figures disagree, both are shown and the disagreement is marked. Separately, official results were added for 4,124 meetings from mid-2023 to early 2025 that had none.
Company pages: one entry per ballot item, and only the season’s meetings
What was wrong. Company pages listed each wording that fund filers used for a ballot item as a separate item, so a meeting could show more items than were on its ballot: Apple’s 2025–2026 page showed 22 items against 12 in the company’s Form 8-K. A season’s page could also include items from a meeting in another season: Air Products’ January 2025 meeting appeared on three season pages.
Scope. Company pages showing items from a meeting outside their season fell from 2,493 to 0. Pages showing one row of an official tally on two separate items fell from 2,081 to 203.
What changed. Each item in the company’s official tally is shown once. Other wordings matched to the same item are listed under it with their own figures. Wordings that could not be matched, and that few managers reported, are kept in a collapsed list and not counted as items. Meetings outside the season are no longer shown on that season’s page; the page says which season they belong to. This change did not alter any vote count, percentage or outcome.
Manager pages: shareholder proposals identified from the filings, and the board’s position stated, not assumed
What was wrong. The “backed E&S proposals” figure on each manager page counts environmental, social, workforce and diversity proposals put on the ballot by shareholders. A proposal counted only if its wording said “shareholder proposal”, which missed shareholder proposals worded otherwise, for example Microsoft’s 2025 “European Security Program Censorship Risk Audit”. The pages also implied that the board opposed every shareholder proposal. Some boards support them, for example U-Haul’s 2025 stockholder ratification proposal, and some make no recommendation.
Scope. The figure on every manager page for the 2024, 2025 and 2026 seasons (537 manager-seasons). For the 20 largest managers it moved by between −3.6 and +2.6 percentage points.
What changed. Who put a proposal on the ballot is now taken from the filer’s own answer on Form N-PX (Item 1(h)) where it gives one; otherwise it is inferred and marked as inferred. The definition of the figure is unchanged: environmental, social, workforce and diversity proposals only. The board’s position is taken from the “management recommendation” field, read only for funds whose use of the field passes an automatic check and never at a meeting with a proxy contest.
Fund votes of foreign companies shown on U.S. company pages
What was wrong. Nine company-season pages showed the fund votes of a different, foreign company: MasTec 2025 and 2026, MercadoLibre 2025, Nutrien 2024 to 2026, Ormat Technologies 2025 and 2026, and Steel Dynamics 2025. The votes belonged to companies such as Proya Cosmetics, CATL, Saudi National Bank, Smartfit and Yifeng Pharmacy. The site could not read some foreign security identifiers, so a single fund record filed with the wrong identifier attached those companies to the U.S. company.
Scope. Nine company-season pages.
What changed. Foreign security identifiers are now read and validated with their check digit, and each foreign company is kept separate. The nine pages now show only the company’s own votes.
Official results and share counts taken from another company’s filings
What was wrong. Some company pages cited another company’s Form 8-K as the official meeting result, for example Allianz showing The Hartford’s, Alimentation Couche-Tard showing Casey’s, L’Oréal showing International Flavors & Fragrances’ and NN Group showing NN Inc’s. Some “% of shares outstanding” figures used another company’s share count: Accenture 2025 used BioCryst Pharmaceuticals’ (208,543,000 shares), and Apple Hospitality REIT 2025 used Apple Inc.’s (about 14.9 billion). The cause was matching filings to companies partly by name.
Scope. 783 stored official-result rows and 177 stored share counts, read by 120 company-season pages.
What changed. A filing is now used for a company only when the SEC registrant that filed it is the issuer of that company’s securities. The mismatched rows were set aside, not deleted. Share counts were fetched again from each company’s own filings for the same year; where none could be found, no percentage is shown. Most of the affected companies are based outside the United States and file no Form 8-K, so their pages now say that no official result is on file. Accenture 2025 now uses 625,676,922 shares and Apple Hospitality REIT 2025 uses 237,914,339.
Fund support now counts withheld votes as opposition
What was wrong. On company pages the share of fund votes cast FOR an item was computed as FOR ÷ (FOR + AGAINST). Withheld votes were grouped with abstentions and left out, although the methodology page says withheld votes are the opposition in a For/Withhold election. Some director elections therefore read as near-unanimous fund support when most reporting funds had withheld. For example, at Air Products’ January 2025 meeting the nominee Charles Cogut showed 100% FOR beside a breakdown showing 35.5% FOR. Separately, where a fund reported the same shares under two directions, the vote counted was chosen with the N-PX “management recommendation” field. In proxy contests this replaced some funds’ votes with the “do not vote” line for the other proxy card.
Scope. The fund figure moved by 5 percentage points or more on 6,717 director-election items across 2,587 company pages. On manager pages the “against the board” figures of the 20 largest managers changed by at most 0.04 points, because they already counted withheld votes as opposition. Official 8-K results were not affected.
What changed. Fund support is FOR ÷ (FOR + AGAINST + WITHHELD). Every direction code a filer uses is mapped by a fixed table, and a code that could mean two things is shown as unknown rather than guessed. Where the same shares are reported under two directions, the card actually voted is counted; the “management recommendation” field is no longer used for this. Air Products’ Cogut item now shows 36% FOR.
Thirty company pages withdrawn while a company-matching fault was fixed
What was wrong. Thirty company-season pages showed fund votes or an official meeting result that belonged to a different company. The cause is described in the two entries of 11 October.
Scope. 30 company-season pages, among them MasTec 2025, Nutrien 2025 and Allianz 2024.
What changed. The 30 pages were taken off the site and out of the sitemap on 10 October. After the repair of 11 October each page was checked against the company's own security identifiers. All 30 passed and were restored the same day.
Four statements in the site’s explanatory text corrected
What was wrong. (1) Ownership: company pages headed their holdings tables “beneficial ownership” and said the percentages were “the ones that were true then”, that is, at the meeting. Form 13F reports what an institutional manager held at the end of a calendar quarter in accounts over which it has investment discretion; that is not beneficial ownership, and some rows come from filings dated after the meeting. (2) Form N-PX scope: manager and company pages said that funds and institutional managers disclose how they voted on every proxy. Registered funds do; institutional managers report only their votes on executive pay (say-on-pay, say-on-frequency and golden parachutes), so a manager with no director vote on file has not left anything out. (3) Directors and shareholder proposals: manager pages said that withholding is the only way to oppose a director in an uncontested election, although a vote AGAINST is possible where a company elects directors by majority vote. They also said that shareholder proposals reach the ballot over the board’s objection; boards usually recommend against these proposals, but not always. (4) The N-PX “management recommendation” field: manager pages showed it as “65.8% wrong” on shareholder proposals and the methodology page called it unreliable. Form N-PX Item 1(l) defines the field as whether each vote was cast for or against management’s recommendation: it records agreement with the board, not the board’s own position. The 65.8% figure came from reading it as the board’s position; that reading was the error, not the field.
Scope. (1) The holdings and ownership section of every company page that has one. (2) The introductory text of manager pages, the managers index and company pages. (3) The explanatory text on manager pages. (4) One section of every manager page. Also the methodology page. Apart from the removed 65.8% table, no figure on the site changed.
What changed. (1) The tables are now headed “holdings and ownership”, each row names the form it came from, and the pages say that each holding is as of its own filing, which may be dated before or after the meeting. (2) The text states what each kind of filer reports. (3) The pages say that a director can be opposed by voting AGAINST under majority voting and by withholding under plurality voting, and that a vote FOR a shareholder proposal is usually, not always, a vote against the board’s recommendation (see also the entry of 12 October on shareholder proposals). (4) The table and the claim were removed and replaced by a description of what the field records; this is a retraction.