The Participation Gap

It takes a significant amount of effort to get votes for a mutual fund proxy campaign. Calls, mailers, and continued outreach all add up. Often, the problem is getting enough votes to allow fund business to continue, rather than winning the issue at hand. According to the Investment Company Institute’s 2019 survey, an average of 94 percent of votes cast on proposals requiring a 1940 Act majority favored the proposal, excluding abstentions and broker non-votes. ^[1]

Those who do vote usually show strong support for the issue.

In the ICI’s sample of campaigns from 2012 to 2019, 38 percent of proposals requiring a 1940 Act majority needed at least one meeting adjournment to reach quorum. The 145 campaigns reported by sixty-four member firms cost $373 million. Those firms represented roughly 76 percent of U.S. registered fund assets. Follow-up solicitation costs were estimated at $158 million to $229 million, or 42 to 61 percent of total costs. These were follow-up costs overall, not just spending after an adjournment. ^[2]

One campaign cost $107 million. In a November 2024 article, the ICI said that one of two campaigns exceeding $100 million would have cost nearly $200 million in then-current dollars. That means 2024 dollars, and the article does not establish that this was the same $107 million campaign. ^[3]

More recent research shows that the expense remains substantial. The ICI’s March 2026 report estimated costs of $675 million to $1.14 billion for surveyed campaigns conducted from 2020 through 2025. These numbers include special meetings across fund types and annual meetings for closed-end funds. They are not a mutual-fund-only total or an industry-wide census of spending. ^[4]

Filings also show how difficult it can be to get enough votes. In a supplement dated December 23, 2024, JPMorgan reported that its Large Cap Growth Fund’s meeting had been adjourned to January 30, 2025. ^[5] In November 2025, Vanguard told Health Care Index Fund shareholders that insufficient participation had led to an adjournment until January 6, 2026. ^[6]

It is an old problem. Analytics may help an outreach team determine whether another contact would be useful. There is a commercial case for testing that idea, but we need evidence that it can produce additional votes at a lower cost.

Quorum Problems in the Numbers

Shareholders must vote on many aspects of a mutual fund, including changes to fundamental policies and new advisory agreements. For matters requiring a “majority of the outstanding voting securities,” the Investment Company Act of 1940 defines approval as the lesser of two thresholds:

• 67 percent or more of the voting securities present at a meeting where holders of more than 50 percent of outstanding voting securities are present or represented by proxy; or
• More than 50 percent of all outstanding voting securities.

This approval formula is not a universal quorum rule. Director elections and other ballot items can have different requirements. ^[7] The practical point is that a proposal can have strong support among those voting and still lack the participation needed to move fund business forward.

According to Broadridge’s 2025 ProxyPulse report, retail investors voted 28 percent of the shares they owned during the corporate proxy season, the lowest level in nine years. Institutional investors voted 76.6 percent, the lowest in at least ten years. ^[8] These figures measure shares, not individuals. They describe corporate voting, not measured participation in mutual fund governance ballots.

The ownership picture helps explain why participation matters so much to mutual funds. In its March 2026 report, the ICI stated that more than 125 million retail investors held fund shares in 2025. At year-end 2024, retail investors accounted for 88 percent of mutual fund assets and 94 percent of long-term mutual fund assets. ^[9] It also cited historical voting averages of about 30 percent for retail investors and 80 percent for institutional investors. ^[10] Those figures provide context, but they do not tell us whether a particular fund has reached quorum. That requires the fund’s own ownership, voting, and approval data.

An adjournment brings additional costs: notices, follow-up mailers, and continued calling. The ICI has described proxy campaigns as burdensome and costly. ^[11] In its 2019 sample, 95 percent of proposals requiring a 1940 Act majority were ultimately approved. ^[12] That is a different measure from the 94 percent average support among votes cast. One measures how many proposals passed; the other measures support among returned votes. Together, they leave us with a reasonable question: how much are campaigns spending just to secure enough participation?

Why Outreach Falls Short

A proxy campaign needs to get its materials to registered shareholders and beneficial owners who hold shares through intermediaries. Broadridge distributes proxy materials on behalf of more than 1,000 broker-dealers and custodian banks. ^[13] Transfer agents and other providers support the registered-holder side. Follow-up depends on the fund, the intermediaries, the timetable, and the permissions available for contacting shareholders.

Phone outreach is one option. A team can call shareholders, leave permitted messages, and follow up through other approved channels. Position size is a reasonable place to start because one large holding can contribute more shares than many small holdings. But is position size always the best guide to the next call?

We need to know whose participation is likely to change because of that contact. Someone who votes regularly may not need an extra nudge. A large holder who has been difficult to reach may still be worth another attempt, but only if that attempt has a reasonable chance of producing a vote.

The ICI raised this question in its 2019 research. It described target-date funds for investors retiring in 2050 or later that required multiple solicitation rounds. Most funds for investors already past their target retirement dates reached quorum after one solicitation round. The report identified age, asset balances, and employment status as possible contributing factors. ^[14] That is a useful hypothesis. It does not prove that targeting those attributes will improve results.

Experienced solicitors already exercise judgment about whom to contact and when to follow up. I want tools that help them learn from those efforts and apply what they know across a campaign with 50,000 shareholders. We also need to find out whether a model’s priorities are better than their current process.

Communication preferences are another place to look. If everyone goes through the same sequence, recorded preferences may help a team avoid wasted attempts. That includes how people want to be contacted, when contact is permitted, and how often to try. The explanation should make the proposal easier to understand, without changing its substance or pressuring someone to vote a particular way.

Can the same people, with better information about the next useful contact, secure more participation?

A Case for Targeting

Start with data a campaign is authorized to use: contact attempts, delivery failures, response timing, recorded preferences, position size, and whether a vote was returned. Transfer-agent records may add useful information where access and permitted use have been established. Having data somewhere in the fund ecosystem does not mean a solicitation firm can obtain it, combine it, or reuse it across clients. Contracts, privacy obligations, and intermediary arrangements all matter.

With enough useful data, we may be able to predict who is likely to vote. But that is only part of the job. We want to know whose participation a contact could increase. If a model only identifies people who would have voted anyway, it may spend money without adding a single vote.

Past campaigns can help, with some limits. Shareholders buy and sell. Proposals differ. A response to one campaign may tell us little about the next. Earlier targeting decisions can also skew the data. If large holders received most of the calls, their response rates do not show what would have happened if other holders had received equal attention.

There is a lot to learn from campaign history, but a model needs to be tested on later campaigns, including new shareholders and different proposals. Start with a simple comparison using position size and recent contact history. Keep added complexity only if it produces a repeatable improvement. More data creates an opportunity to learn; it does not guarantee a better model after every campaign.

The system could suggest the next permitted contact based on voting impact, estimated response, channel, and cost. The goal is participation whether a shareholder votes for or against the proposal. A low score must never remove required notices or someone’s ability to vote. Contact limits, applicable opt-outs, and a human route for questions belong in the process. An opt-out from optional follow-up is not a reason to withhold required communications.

Run the model against what the team is doing now. Compare similar groups, and where practical, randomly allocate eligible additional follow-up while preserving required communications. Measure additional shares voted per dollar alongside attempts, complaints, and opt-outs. Give experienced operators a way to override a recommendation, record why they did it, and review the result.

Time to the required threshold and the need for adjournments also matter. One successful campaign is not enough to prove the case. We need to see whether the improvement holds across campaigns and whether it justifies the added cost.

How AI May Be Able to Assist

Machine learning can help identify patterns in contact and voting history. It does not have to be overly complicated. Start with something the team can understand, then see whether a more complex approach earns its place. AI may also help with the documents and work surrounding a campaign.

Proxy materials can be lengthy and hard to follow. Solicitors need to understand the proposals and the limits of what they can say. A document assistant could retrieve relevant passages from approved materials, cite them, and help prepare an answer for human review. That could make it easier to learn a campaign and find the right explanation.

The answer still needs checking. AI can produce a confident response that the source does not support. Keep answers traceable to approved documentation and give the team a clear escalation route when a question is uncertain or requires individual advice.

AI can also flag possible exceptions, such as missing consent evidence where required or language that departs from an approved script. Clear restrictions should be enforced through rules wherever possible. If a system relies on a model to catch everything, it is asking too much of the model. The flags need review, and missed flags remain possible.

Where AI Should Not Take Over

Consider a shareholder who is upset about a proposed change in advisory fees. That person needs someone who knows when to stop explaining and listen. AI should help prepare the solicitor, not replace that conversation.

Or a shareholder may ask how a proposed transaction would affect their taxes. The team can point to relevant disclosures and direct individual tax questions to a qualified adviser. A generated answer is not a substitute for that advice.

AI should not replace an understanding of the law. Proxy solicitation has rules that need to be interpreted and translated into controls. Software can flag possible concerns, but qualified counsel and experienced compliance professionals need to decide what applies and approve the process.

Nor should AI be presented as a complete answer to structural problems. If materials are confusing or voting is cumbersome, better targeting will only go so far. Clearer approved explanations and easier voting routes may help. Ownership structures and statutory requirements remain constraints that a model cannot remove.

The useful question is which part of the workflow can be improved with data the operator is allowed to use and controls the operator can support.

Compliance

The rules depend on the activity, the communication channel, and the participant’s legal role. A fund, broker-dealer, transfer agent, and solicitation provider do not necessarily have the same obligations. Before configuring the system, identify the requirements for contacts, data use, and records.

The Telephone Consumer Protection Act regulates certain automated or prerecorded calls and texts. Under its private-action provision in section 227(b)(3), a claimant may recover actual monetary loss or $500 per violation, whichever is greater. A court may increase the award to as much as three times that amount for a knowing or willful violation. When the $500 measure applies, that can mean $1,500 per violation. These are civil damages, not a flat schedule of fines for every call or text. ^[15]

Application depends on the technology, message, recipient, consent, and any relevant exemption. The Eleventh Circuit vacated the FCC’s 2023 one-to-one consent restrictions on January 24, 2025, before their scheduled January 27 effective date. ^[16] That did not eliminate other applicable consent and calling requirements.

Recordkeeping also depends on the entity and the records involved. For covered broker-dealer records, the SEC’s 2022 amendments to Rule 17a-4 permit an electronic audit-trail alternative to non-rewriteable, non-erasable storage. Among its requirements, the alternative must allow the original record to be recreated if it is changed or deleted. ^[17] That is not a blanket retention mandate for every solicitation provider, and recreation alone does not satisfy every condition of the rule. Counsel and compliance teams should determine the duties and retention periods that apply.

For AI-assisted work, preserve approved source material, generated drafts where appropriate, the reviewer’s decision, and the version actually sent. Record the model and instructions when needed for review. Set access and deletion rules for that history rather than keeping everything indefinitely.

Build the approved controls into the product. Enforce contact restrictions, retain required evidence, log overrides, and route exceptions to accountable people. Separate client data and limit access by role. An audit trail helps someone review what happened. It cannot, by itself, prove that a campaign complied with the law.

How We May Evolve Over the Next Five Years

There is room to improve the way votes are solicited. We know that getting enough participation can take substantial money and time. We do not know that every firm uses the same methods, and a historical adjournment rate is not a measurement of today’s campaigns. Start with an operator’s actual costs and participation problems.

Here are four areas I would focus on.

First, stop collecting data without a clear permitted use. Know what can be stored and reused in future campaigns. Keep a record of its source and the decisions made with it. That gives a team something useful to learn from without assuming that every piece of data belongs in a model.

Second, do not enrich data without knowing why. Start with authorized campaign records and preferences shareholders have actually expressed. If outside data is added, establish its source, accuracy, permitted use, and benefit. Do not assume someone is wealthy because of a ZIP code or substitute that guess for knowing how they want to be contacted.

Third, capture the needed evidence while the work happens. A reviewer should be able to follow a communication back to its source and approval. Retain records according to the participant’s role and applicable requirements. The test is whether someone can conduct a real review, not whether the product calls itself “compliance-grade.”

Fourth, capture shareholder intent in a form that can support future voting. Advanced Voting Instructions deserve particular attention here. They belong in a broader discussion of tools that give shareholders more control over participation. Pass-through voting is a related but different development. It lets fund investors influence how their proportionate holdings are voted at portfolio companies’ shareholder meetings. In March 2026, Broadridge reported that pass-through voting was available across more than 600 funds representing more than $8 trillion in assets. ^[18] Those figures describe availability, not votes cast. These programs do not automatically provide votes on the fund’s own advisory agreement, board, or other governance proposals.

Advanced Voting Instructions and Shareholder Intent

Advanced Voting Instructions (AVI) let investors set voting directions before proxy materials arrive. Broadridge describes a process based on categories of proposals for equities and funds: ballots are prefilled from those instructions, delivered, and affirmatively submitted by the shareholder. That final submission matters. A stored preference is not yet a cast vote. ^[19]

I see AVI as a major driver of the next stage of shareholder participation. Some shareholders have clear preferences but little time or inclination to work through each campaign from scratch. Capturing those preferences gives them a starting point they can review and change when a ballot arrives. Where instructions remain applicable and their use is authorized, a platform could carry that expressed intent forward across campaigns. That could reduce the repeated effort required to participate.

Propensity modeling and preference capture address complementary parts of the problem. Modeling helps a solicitor decide where another contact could produce participation. AVI helps a shareholder act on choices they have already expressed. Together, they create an opportunity to combine better outreach with an easier path from intent to a submitted vote. That combination should be part of the product roadmap and tested alongside improvements to calling and document workflows.

The infrastructure should preserve the instruction, who supplied it, when it was given, and what it covers. Shareholders need a clear way to review, change, or withdraw it. A system should make any proposed application of an instruction visible and keep a record of the ballot ultimately submitted. AI could help organize instructions and flag uncertain matches for review; the shareholder’s choice must remain explicit. For a fund campaign, the team would need to establish which proposal types and voting channels support the workflow. Success should be measured in additional participation and reduced effort, not just preferences collected.

There are smart, experienced people who know how to solicit votes. Better tools could help them work more effectively. Faster completion and lower costs are results we need to demonstrate, not assumptions we should sell.

Could smaller solicitation firms get useful analytics without building a large proprietary system? A focused product could combine authorized contact history, next-contact recommendations, document assistance, support for AVI through compatible voting providers, and reviewable records. It could support the compliance team’s work, but it could not guarantee compliance. Its value would need to exceed the costs of integration, licensing, data, and oversight.

People do not vote for many reasons. They may lack time, find the materials confusing, or feel that the issue does not matter to them. Better outreach will not solve every cause. What we can test is whether better information helps a team choose its next useful contact, whether AVI makes it easier for shareholders to act on their stated preferences, and whether those improvements secure more participation at an acceptable cost. That is an opportunity worth building around.

About the Author

Henry V. Alfano is the Founder and CEO of Wild Eye AI, where he designs AI systems for regulated industries. His current work focuses on shareholder intelligence for proxy solicitation, including participation modeling, proxy document analysis, and auditable workflows. He is based in Manhattan Beach, California.