Proxy Voting: What a Voting Record Does and Does Not Show

Proxy voting is the manager casting the votes attached to the shares a portfolio holds, on the resolutions companies put to their shareholders. Because the mandate is discretionary, those votes are cast without asking the holder each time. A voting record shows how many votes were cast and which way; on its own it shows nothing about why, or about what was achieved.

There is a transfer hiding inside that sentence and it is worth slowing down for. Somebody's money bought the share. Somebody else decides how the share votes. The arrangement was written that way at the beginning and neither side has changed it since, so nobody in between stops to ask the first person. Two vote counts are the whole of the evidence. Dividing them in view, rather than reading a finished figure off somebody's summary, is what makes the limits of the result visible.

The running example is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore run by Faiz Ahmad Ansari for an invented charitable endowment whose investment committee is chaired by Rukmini Deshpande. Its equity sleeve carries Rs 300 crore across 28 names, settled in the selection sequence and used here without being rebuilt. Across one stated twelve month period the mandate voted on 214 resolutions arising from those holdings, and on 19 of them it voted against the board's recommendation. The two counts are the entire body of evidence. Every conclusion below rests on them, and they are all there is to rest anything on.

A manager's wider duty, and why discretion is the condition that creates it at all, is covered separately. Voting is one instrument of that duty, and the voting decision itself raises four questions: what casting a vote does to the holding, what a committee has to write down, how a decision is evidenced afterwards, and what happens when nobody casts anything at all.

What is a proxy vote, and who actually casts it here?

Companies put items to their shareholders. Each item is a resolutionAn item put to a company's shareholders for a vote at a meeting. A resolution is the unit that gets counted in any voting record., and every share carries a right to vote on it. The catch is geography and time: the meeting happens somewhere on a particular morning and a holder of shares in twenty eight companies cannot be in twenty eight rooms. So the vote is cast by proxyA vote cast on behalf of a shareholder who is not present at the meeting. The word describes the standing-in and no particular way of voting., which simply means somebody stands in and casts it on the shareholder's behalf. Standing in is the whole of what the word carries, and it says nothing about how the vote goes.

In a discretionary mandateAn arrangement in which the manager decides and acts without seeking the holder's approval for each decision. The votes attached to the holdings are among the decisions that move across. the standing-in is done by the manager, and it is done as a matter of course rather than by a fresh instruction each time. The person whose money bought the share is therefore not the person deciding how the share votes, and that separation is created by the arrangement rather than by any single vote. Faiz Ahmad Ansari does not telephone Rukmini Deshpande before each of the 214 items. He would not be running a discretionary mandate if he did.

Ten shops share one mall. Once a year the shopkeepers vote on the shared security contract. The meeting is at eleven on a Wednesday, the exact hour a shop cannot be shut. So a shopkeeper has choices. Attend and vote. Send a nephew holding a signed line saying how to vote. Send the nephew with no instruction at all, in which case somebody else has effectively chosen. Or send nobody, in which case the contract is settled by whoever turned up. Every one of those four is a real decision and only the last one feels like an absence of one.

One further thing about the act itself, and it is the part that surprises people. Casting a vote moves no money. The Rs 300 crore equity sleeve is exactly the same size the morning after a meeting as it was the morning before, the 28 names are the same names, and not one position has grown or shrunk because of anything the mandate said on a ballot. A vote changes the record of what the mandate did and changes nothing whatever about what the mandate holds. So a vote never appears in a return, a weight or a turnover figure. That is also why voting is so easily left undone. Nothing in the portfolio complains.

Why is casting the vote a duty rather than an option?

Because the right exists whether or not anyone reaches for it. A share that is never voted still carried a vote that day; it was simply left on the table. So declining is not a way of standing outside the question, it is a way of answering it without saying anything. Not voting is a decision, and in a discretionary arrangement it is a decision the holder took no part in and will usually never hear about. That last clause is the part that turns a small procedural point into a duty: the endowment cannot notice an omission it is never shown.

Two acts get mixed up here and they are not the same. An abstentionA recorded decision to cast no vote either way on a resolution. The decision itself leaves a trace. Simply not participating leaves none. is a positive act with a record attached: the holder was present in the process and chose to take no side, and the choosing is written down. Failing to vote is nothing at all, and it leaves nothing at all. Both acts hand the outcome to everybody who did vote, so neither is neutral in its effect. Only one of the two leaves anything a reviewer can look at afterwards. The difference matters more to the review than it does to the result.

There is a practical consequence for anyone building the process rather than reading about it. A voting arrangement that produces silence on an item cannot be distinguished, six months later, from a voting arrangement that considered the item carefully and concluded that no side was worth taking. Both look identical in the file, and the file is empty. The abstention is the cheap fix: it costs one line and it converts a silence into a statement.

What do 19 votes out of 214 actually come to?

A figure watched coming out of two numbers behaves differently afterwards, so the division is worth doing rather than reading the result off somebody's summary. Nineteen divided by two hundred and fourteen is 0.088785, and that decimal expansion is written out here so the rounding is visible rather than folded away. Multiplied by a hundred, the voting shareThe count of votes of one kind divided by the total number of resolutions voted on, expressed as a percentage. The base is the total, and the total is therefore always named beside it. is 8.8785 per cent, which is 8.9 per cent to one decimal place.

The other side of the division is the larger number and it is almost never the one printed. Turn the same year round. Two hundred and fourteen less nineteen is one hundred and ninety five. One hundred and ninety five divided by two hundred and fourteen is 0.911215, or 91.1 per cent, and those are the resolutions on which the mandate voted with the board recommendationHow a company's board asks its shareholders to vote on a resolution. A recommendation is a request rather than an instruction, and shareholders may go the other way.. Both sentences describe the identical twelve months and the identical 214 items, and which of the two a report prints is a choice somebody made rather than a fact the record forced on them.

The precision printed decides what the figure can still be traced back to, so the decimal is worth holding on to. At one decimal place, 8.9 per cent points at nineteen votes and at nothing else: eighteen would have come out at 8.4 per cent and twenty at 9.3 per cent, so neither could hide behind it. Printed as a whole number, 9 per cent now covers nineteen votes and twenty votes alike. For 20 votes the division gives 0.093458. One decimal place is the whole difference between a figure that names its own count and a figure that gestures at a neighbourhood of counts.

How big is a voting year set against a 28 name sleeve?

A count on its own has no size until it is put next to something. Two hundred and fourteen sounds like a great deal of voting until the equity sleeve's 28 names are set beside it. Divided out, 214 over 28 is 7.642857, so about 7.6 resolutions a holding across the stated twelve month period. Seven and a half items a company is a very different mental picture. Once the routine business is included, a handful of items per company across a year is what an ordinary annual meeting produces, and a handful of items is not a wall of contested decisions.

The scaling gives the size of the exercise and nothing whatever about its quality. The difference between size and quality is the single most useful thing to hold on to about any count. A mandate that read every one of the 214 items carefully and a mandate that pushed all 214 through a standing template would report the identical 214. The number measures how much voting happened. The count has no vocabulary at all for how the voting was done.

Sliding between bases without announcing it is how a reader ends up misled about scale. A second base is sitting close by, and it is worth making explicit. The 8.9 per cent is measured against resolutions. Somebody will eventually ask a different question: on how many of the 28 holdings did the mandate oppose the board at least once? The second question is measured against holdings, and this record cannot answer it. The record can put walls around it. If the nineteen opposing votes fell on nineteen different companies, that is 19 of 28. The division gives 0.678571, or 67.9 per cent of the holdings. If all nineteen fell on one company, it is 1 of 28, and the division gives 0.035714, or 3.6 per cent. The truthful answer is a range running from 3.6 to 67.9 per cent with the actual figure not supplied, and quoting either wall as though it were the answer would be an invention.

What does the count leave open?

Three things. A general caution is precisely what a later reader skips, so each of the three is worth naming on its own. The first is the resolution mixWhat the resolutions in a period were actually about. The mix decides what any share of the resolutions means, and it has to be recorded separately from the counts.. The record carries no categories for the 214 items at all, and it does not say how many were routine annual business and how many were genuinely contested. Without that split, the denominator of the 8.9 per cent is a bag whose contents are unknown.

The second is engagementDiscussion between a holder of shares and a company. Engagement may precede a vote, follow one, or happen without any vote at all, and it leaves no mark in a vote count.. The record holds none. So a vote cast after a long conversation with a company and a vote cast straight off a standing template are indistinguishable in the count, and in practice the conversation is often the part that changed something while the vote is only where the conversation surfaced. The channel most likely to have produced an effect is the one that leaves no trace in the arithmetic at all.

The third is outcomes. The record does not say what happened to any of the 214 items, so not one of the 19 can be followed to a result. The record carries no shareholding percentage anywhere either. The weight the mandate's votes carried in any of those decisions is unavailable too. Four unknowns, then, sitting between a computable figure and any statement about conduct.

Sorting statements is easier than arguing about them, so take four sentences somebody might write about this year and ask of each whether the two counts carry it. The mandate voted on 214 resolutions: carried, and the count says so. The mandate voted against the board nineteen times: carried by the same count. The mandate pays close attention to the companies it holds: not carried. Attention is not a quantity anywhere in the file. The mandate changed something at a company: not carried. No outcome is recorded, and no stake size exists to weigh a vote with. Two of the four are arithmetic and two are opinions wearing the arithmetic's clothes, and separating them takes about a minute.

Can the same 8.9 per cent be read two opposite ways?

It can. Suppose most of the 214 items were routine annual business of the kind on which almost every shareholder votes with the board. Then nineteen departures from that pattern is a substantial number of positions taken, and the year looks considered. Now suppose instead that a large share of the 214 were genuinely contested items where shareholders split. Then the same nineteen is a small number, and the year looks like a mandate going along with things. The identical figure supports both descriptions, and the fact that would decide between them was never written down.

The symmetry is what stops one reading feeling safer than the other. Reading a low share as deference assumes the resolutions were contested enough for opposition to have been available. Reading a high share as diligence assumes exactly the same thing from the other end. Both readings smuggle in the same missing fact. The cautious choice is to decline both readings and say why in writing. Neither reading is itself the cautious one.

Is a vote against the board a vote against the resolution passing?

No, and these two get run together constantly. A vote is a thing the mandate did. Whether the resolution passed is a thing everybody else did, added to it. On any item the mandate cast one holder's votes into a pool with every other shareholder's, and the arithmetic that settled the result belongs to the pool rather than to any single participant. A count of votes cast is a record of the mandate's own conduct and never a record of what happened next.

Nor is the count a measure of influence, for a reason that is easy to state and easy to forget. Influence needs a size, and this record carries no shareholding percentage in any company whose resolutions were voted. Without it, a vote cast by a Rs 23 crore position and a vote cast by a small one look identical in the tally. Weight cannot be computed without that percentage, and a tally read as influence is reading something the record does not contain.

What has to sit under a percentage before it is a record?

A line for every resolution, and four fields on each line: what the resolution was, what the board recommended, how the mandate voted, and why it voted that way. A line carrying all four fields is a voting recordThe resolution by resolution log that sits underneath any summary figure. Each line names the item, the recommendation, the vote and the reason for it.. Everything above it, including the 8.9 per cent, is a summary computed from that log. A percentage is the top line of a document that has to exist beneath it, and the mandate's record carries the top line and not the document.

The reason to insist on the four fields rather than three is the last one. Without a stated reason each line records that a position was taken and not that a position was reached, and reasons are the only field a reviewer can argue with a year later. A template cannot fake the reason field convincingly. A template produces the same sentence 214 times, and the repetition is visible in the log. So the right response to being shown a voting share is not to admire it or to doubt it, but to ask for the layer underneath and see what comes back.

There is also a timing point buried in the fourth field, and it decides whether a record can be built at all. A reason exists at exactly one moment, the moment the decision is taken. Everything downstream of that moment is arithmetic on a decision already made: the vote gets cast, the votes get counted, the counts get divided, and one line gets printed into a pack. No later step can manufacture a reason, so a reason not captured at the decision is lost to everybody afterwards, however carefully they reconstruct the year. A reason is therefore captured when the decision is taken, not when the year is reported.

Where are the rules on voting actually set?

Not in the mandate. The document Rukmini Deshpande's committee and Faiz Ahmad Ansari signed can say what the arrangement wants, and it can be argued with line by line, but whether any voting obligation, policy requirement or disclosure duty attaches to a discretionary arrangement is settled somewhere neither of them writes. A reader who needs such a requirement goes to the authority that publishes it and reads the current text there.

How is a voting line handled in an actual review meeting?

Three moves, in order, and they take about four minutes between them. First, whoever presents the stewardship figures reads the count and immediately says what base it sits on: 19 of 214 items voted on across the stated twelve months. The share is 8.9 per cent, and the same year is 91.1 per cent with the board. Naming both halves in one breath is the cheapest correction available, and it stops the meeting hearing a single flattering number.

Second, somebody asks what the 19 were about. If the answer arrives, the discussion becomes a real one about specific items. If the answer does not arrive, that is not a dead end: it is the finding, and it goes into the minutes as a finding rather than evaporating. An absence written down survives to the next review. An absence merely noticed in the room does not. An analyst reading the mandate from outside does the same thing with a shorter list, because the analyst has only the summary and knows it.

The household version is smaller and identical in shape. A residents' association sends round a notice about a maintenance contract before its annual meeting. The count of how many flats returned a form gives the turnout and nothing about whether anybody read the contract, and a resident who wants to know which way things went asks for the minutes rather than the turnout figure. Turnout is the summary; the minutes are the log. The distinction between a summary and a log is the same one at every scale.

What is proxy voting not?

Proxy voting is not a score. Nothing about a share of votes ranks one arrangement against another, and two mandates reporting different shares may simply have held different companies putting different items to their shareholders in the same twelve months. The discussion that often does the work leaves no line in the tally, so a vote count is not a measure of engagement either. And a vote count is not a claim about what any of it did to the portfolio's return. Return is a separate question, and one this record could not answer.

Proxy voting is, in the end, one of the few parts of a manager's conduct that leaves a countable trace by default. A countable trace is exactly what gets asked to carry more weight than it can. The finding is a refusal, and it is a result rather than an apology: the voting share for this invented mandate is computable to six decimal places and it is not interpretable at all, and a reader who computed it and then declined to read anything into it has done the entire job correctly.