Stewardship: The Duties of Running Money for Someone Else

Stewardship is the set of duties a manager takes on when the money belongs to somebody else and the decisions are made without asking first. The duties run to the holder's stated purpose, not to the manager's preference, and they are judged against what the mandate asked for and against what the holder actually received after costs.

Two halves sit inside that sentence and most reporting shows only one of them. The first half is conduct: did the person running the money stay inside what was agreed, and can anybody check it. The second half is arithmetic: what actually arrived at the holder once every cost of the arrangement had been taken out. An answer to only one of them describes either a well behaved process or a number with no accountability attached to it. Both answers belong to the same twelve months, and they can point in opposite directions.

The running example is the Anantara Multi-Asset Portfolio, an invented discretionary mandateAn arrangement in which the manager decides without asking the holder first, against a purpose and limits set in advance. of Rs 500 crore run for an invented charitable endowment. Rukmini Deshpande chairs the endowment's investment committeeThe holder's own decision making body: it appoints the manager, writes the instructions and reviews what happened., and Faiz Ahmad Ansari runs the mandate. Its stated shape is equity 60.0 per cent at Rs 300 crore, fixed income 30.0 per cent at Rs 150 crore and cash 10.0 per cent at Rs 50 crore, and those three sum to Rs 500 crore exactly.

What is stewardship, and what single condition creates it?

Money that belongs to the person deciding, decided by that same person, creates no duty to anybody. Money that belongs to somebody else, decided by that person, creates one immediately. Adding the second ingredient changes the shape of the duty again: where the manager has to ask before every decision, the holder is reviewing each one in advance and can simply refuse. Where the manager does not have to ask, nobody is reviewing anything in advance, and the review has to be built somewhere else.

The everyday version. A householder is travelling for three weeks and a wedding in the household falls inside that window. The neighbour gets the keys, the cash and a written list of the intended spending, and the householder will not be reachable for most of it. The conversation is not going to happen at the moment each decision is taken, so the list has to do the work the conversation cannot. On returning, the householder will not ask whether the neighbour felt the spending was sensible, but will read the list, read what was spent, and see whether the two match.

The structure scales without changing. Replace the neighbour with Faiz Ahmad Ansari, the keys with Rs 500 crore, the list with a mandate document, and the wedding with twelve months of markets. The arrangement exists precisely to spare the endowment from reviewing each purchase and sale in advance. So the standard the manager is judged against has to exist beforehand, in writing, and specific enough to be read the same way by two people who disagree.

Where do a manager's duties actually come from?

There are exactly two places, and keeping them apart is the whole of this section. The first is the mandate document itself. The endowment and the manager wrote it together and agreed it, and every clause of it appears below, to be argued with and tested line by line. The second is whatever a regulator requires of an arrangement like this one, and a regulator's text can only be pointed at.

The two feel identical when read and behave differently when relied on. A duty from the mandate is true because two parties agreed it, and it stops being true the day they agree something else. A duty set in regulation is true whether or not anybody agreed it, and changes when the rule changes rather than when the parties change. An account that blurs the two will sooner or later state a requirement that nobody checked, and the reader will act on it as though somebody had.

What does discretion do to accountability?

Discretion moves the whole of accountability onto the moment the decision was taken. A decision made without asking has to be defensible against the mandate as the mandate stood at that moment, on the information available at that moment. The standard is demanding and fair. It is the only standard the manager could actually have met at the time.

The tempting alternative is to judge by how the year came out, and it does not survive being applied consistently. A manager whose only defence of a decision is the result has no defence at all in the year the result goes the other way, and every manager eventually gets that year.

A driving test works the same way. The examiner does not fail a candidate because another car did something unpredictable at a junction, and does not pass one because a manoeuvre that had no business being attempted happened to come off. The examiner assesses what the candidate did with what could be seen, measured against a standard written before the candidate got in the car. Separating the decision from the result is the reason a mandate is written at all.

How does a constraint turn a duty into something testable?

By putting a number on one side and a limit on the other. A subtraction then settles the question and no judgement is needed. A duty phrased as an intention cannot be failed. A duty phrased as a comparison can be failed on a Tuesday afternoon by an assistant with a calculator, and that is what makes it a duty rather than an aspiration.

The Anantara mandate carries four of them, all invented and all agreed in advance. Equity must sit between 50 and 70 per cent of the portfolio. No single holding may exceed 5 per cent. No unlisted holdings. The fixed income sleeve must meet a minimum credit standing. The mandate states that standing as a policy rather than as a symbol. Each one is a mandate constraintA limit agreed in advance so that compliance is settled with a number rather than argued about., converting something the endowment cares about into something an outsider can check.

Turn each one into rupees and the check becomes a subtraction. The equity band of 50 to 70 per cent of Rs 500 crore is Rs 250 crore to Rs 350 crore, so the policy weight of 60.0 per cent, Rs 300 crore, sits Rs 50 crore above the floor and Rs 50 crore below the ceiling. The 5 per cent cap is Rs 25 crore. The largest holding in the stated year was 4.6 per cent of the portfolio, or Rs 23 crore, and Rs 2 crore of headroom is left. Nobody has to agree about intentions for those answers to come out the same way twice.

The fourth constraint behaves differently: its check is a statement rather than a subtraction, and that is not a defect. Some duties reduce to a number and some do not, and the ones that do are the ones a review actually runs. A duty nobody can test with a number tends to be reported as satisfied by the person it constrains. Being reported as satisfied is not the same as being satisfied.

One more thing decides whether a test means anything: the base. The 5 per cent cap is written against the portfolio, so the check divides by Rs 500 crore and the Rs 23 crore holding is 4.6 per cent. Measured against the Rs 300 crore equity sleeve instead, that identical holding is 7.7 per cent. Neither figure is wrong and they answer different questions, so a check that slides between the two bases without saying which one it used has quietly changed what it was testing. Which base applies is settled by the mandate, and this mandate says the portfolio.

Whose return counts, the portfolio's or the endowment's?

The endowment's, and they are not the same number. The portfolio produces one figure and the person whose money it is receives a different one. Both are correct, both describe the same twelve months, and they answer two different questions.

The gross returnWhat the portfolio itself produced, measured before anything the arrangement costs has come out of it. is a fact about the holdings. The gross return states what the securities did while they were held, and it is the right number for asking whether the selection and the allocation worked. The net returnWhat is left for the holder once every cost of the arrangement has been taken out. is a fact about the arrangement. The net return states what actually reached the endowment, and it is the only number that describes the person paying for all of it. Reporting the first and calling it the result is not a lie and is not an error of arithmetic; it is an answer to a question the holder did not ask.

The household version is a rented flat. The rent is Rs 30,000/- a month and that is what the tenant pays, but the owner of the flat does not receive Rs 30,000/-. Society charges, maintenance, the agent's cut, the months with no tenant and the tax all come out first. The flat yields one number. The landlord receives a smaller one.

What did the stated year cost, and what reached the holder?

The mandate carries two charges, both invented and both this arrangement's own commercial terms rather than any market level or anything a regulator sets.

The first is a management feeA charge struck on the size of the portfolio rather than on its result. of 1.25 per cent of assets. On Rs 500 crore that is Rs 6,25,00,000/-, and it is payable whatever the year did. The second is a performance feeA charge struck only on the part of the return that clears an agreed level. of 15 per cent of the return above a 10 per cent hurdleThe return level a performance fee begins above. Below it the charge is nothing.. The stated year returned 14.2 per cent, so the part above the hurdle is 4.2 percentage points. On Rs 500 crore, 4.2 points is Rs 21,00,00,000/-, and 15 per cent of that slice is Rs 3,15,00,000/-.

Add them. Rs 6,25,00,000/- plus Rs 3,15,00,000/- is Rs 9,40,00,000/-. On Rs 500 crore that is 1.88 per cent of assets, made up of 1.25 per cent from the management charge and 0.63 per cent from the performance charge. So the endowment's own return for the stated year is 14.2 less 1.88, or 12.32 per cent. The composite benchmark returned 12.6 per cent over the same twelve months. The portfolio finished 1.6 percentage points ahead of its benchmark gross and the endowment finished 0.28 percentage points behind it net, and the same twelve months produced both of those sentences.

The portfolio beat the benchmark and the holder did not. Nothing was misreported to produce that: no cost was hidden, no return was overstated, and both figures are exactly what they claim to be. The two answers differ because they answer two different questions, and only one of those questions is about the endowment.

The same twelve months in rupees rather than in points. The portfolio returned 14.2 per cent on Rs 500 crore, or Rs 71,00,00,000/-. The cost of the arrangement took Rs 9,40,00,000/- out of that. The charge is 13.2 per cent of the whole year's return, and Rs 61,60,00,000/- is left for the endowment. The benchmark's 12.6 per cent on the same Rs 500 crore would have been Rs 63,00,00,000/-. The distance between what the endowment received and what its own benchmark did is Rs 1,40,00,000/-, and that is what a net shortfall of 0.28 percentage points looks like once it is written as money.

The monitoring work on the same twelve months splits the gross excess by how much of it was simply carrying more market exposure. At a beta of 1.08 against a benchmark that returned 6.1 points above the 6.5 per cent risk-free rate, the part explained by exposure is 0.488 points and the residual is 1.112 points. On Rs 500 crore the residual is Rs 5,56,00,000/-. A record citing the leftover as 1.11 points is quoting Rs 5,55,00,000/-, and the Rs 1,00,000/- between them is where the rounding was taken. The full working of that split is settled where the excess is decomposed.

The gross excess of 1.6 points is Rs 8,00,00,000/-. The residual of 1.112 points is Rs 5,56,00,000/-. The cost of the arrangement is Rs 9,40,00,000/-. The charge exceeded not only the gross excess over the benchmark but also the part of it that market exposure does not explain, and that is the comparison the endowment is actually in.

Whether the arrangement was worth it depends on what the alternative would have delivered over the same twelve months, and the record holds no alternative: not a cheaper mandate, not an unmanaged holding of the same benchmark, not a different manager. Without one, every verdict is a guess wearing a decimal point. The arithmetic reaches minus 0.28 points net and stops there, and stopping there is the finding rather than a failure to reach one.

Does it matter which return the hurdle is measured against?

Measuring against a different return changes the charge, so yes. The mandate measures its 10 per cent hurdle against the 14.2 per cent the portfolio produced, giving the 4.2 point slice and the Rs 3,15,00,000/- computed above.

Write the same clause the other way, measuring the hurdle against a return that has already had the management fee taken out. The reduced return is 14.2 less 1.25, or 12.95 per cent. The slice above the hurdle is then 2.95 points. On Rs 500 crore the slice is Rs 14,75,00,000/-, and 15 per cent of it is Rs 2,21,25,000/-. Total charges become Rs 8,46,25,000/- or 1.6925 per cent of assets, and the endowment's net figure becomes 12.5075 per cent against the same benchmark of 12.6. The shortfall narrows from 0.28 points to 0.0925 points and does not disappear.

The charge rate never moved, the return never moved, and the amount changed by Rs 93,75,000/- purely because of which number the hurdle was measured against. That is why the base a hurdle sits on belongs in the body of a mandate where the committee reads it, and not in a footnote where a reader will pass over it as a technicality.

What has to sit in a record before anyone can check it?

Four items, and the fourth is the one that gets left out. What was decided. On what basis. Inside which constraint. And what would have changed the decision. Omitting the last one leaves a note explaining why the decision was right. A note explaining what the decision depended on is a different document.

The fourth item is what makes a record testable rather than persuasive. A note saying the holding was bought because a particular condition was expected to hold lets a reviewer ask what happened to that condition. A note saying only that the holding looked attractive leaves nothing to check, and the review becomes a conversation about whether the manager seems sensible. A record that cannot be falsified by anything is not a record of a decision; it is a description of a mood.

Now the timing rule. A reason supplied after the result is known is not a reason, so all four items are written when the decision is taken or they are not written at all. The trouble is not honesty. Once the result is known it has already changed what looks obvious, so a reason written up later is sincere and untestable at the same time. Think of a student writing down a prediction before a match and a student explaining the score afterwards. Both may be sincere. Only one of them can be marked.

What do the two vote counts actually carry?

The Anantara mandate voted on 214 resolutions across its holdings during the stated year, and the record carries a second count, 19 occasions on which it voted against the board's recommendation. Voting is the one part of stewardship that leaves a visible trace by default. A visible trace is exactly what gets over read.

A count is a summary of a record that would have to exist underneath it, and what is available here is the summary and not the record. How those counts are read, what they can support and what they cannot, is worked through where the voting record itself is examined, later in this sequence.

The boundary of the evidence belongs here. The record holds two counts and nothing beneath them: no categories for the resolutions, no engagement record at all, no cost for the derivatives overlay, and no stated policy on any environmental or social criterion for this mandate. So nothing in that record supports any statement about what those resolutions were about, or about what was discussed with anybody. Naming an absence is what stops it being quietly filled with something plausible, and a plausible filling is indistinguishable from a fact once it has been printed twice.

What is stewardship not, and what does it never promise?

Stewardship is not a promise about a result. It is not a claim of skill. It is not a substitute for the arithmetic. Every duty in the Anantara mandate could be met exactly, every decision recorded at the time with all four items, every constraint tested and passed, and the endowment could still finish the year at 12.32 per cent against a benchmark of 12.6 and be unhappy about it. The stated year finished exactly there.

Duty discharged and outcome liked are two independent readings, so all four combinations occur, and two of them look like contradictions to anybody who has quietly merged the two ideas into one. A manager can breach a constraint in a year the holder enjoys, and the enjoyment does not repair the breach. Treating a good year as evidence of good conduct is the same error as treating a bad year as evidence of bad conduct, run in the more comfortable direction.

The everyday version is a journey rather than a test. Someone can drive within every limit and arrive late because a bridge was closed; someone else can speed the whole way and arrive on time. Assessing drivers by arrival time eventually promotes the second one. The reason a mandate is written down at all is to make the assessment survive the years when conduct and outcome disagree.

How does anybody use this in a room, on a Tuesday?

By changing the order of two sheets in the committee pack. The change costs nothing and is the whole intervention. An investment committee like the one Rukmini Deshpande chairs asks for the holder's own figure at the top of the pack, the cost line that produced it underneath, and only then the gross figure with the constraint checks beside it. The number that requires the subtraction now sits at the top of the pack, so the missing subtraction becomes impossible.

An analyst reading somebody else's report does the same thing in reverse. The analyst finds the cost, finds the period, finds the base every percentage was struck on, and rebuilds the holder's figure before reading a word of the commentary. If the report gives a result without giving the cost, the result is about the portfolio and the report has not yet said anything about the person paying for it. The observation is not an accusation. It is a description of which question the report answered.

A household runs the identical check with a pen. Whatever the savings arrangement is, write down what went in and every charge that came out, including the ones deducted before the statement was printed, and compute what is left. Then the conduct question: was there anything written down beforehand saying what this money was for, and can anybody check the decisions against it. The two questions are the same two questions at Rs 500 crore and at Rs 5,000/-, and the second one is the one almost nobody writes down at either size.