Compliance Monitoring: Checking a Portfolio Mandate

Compliance monitoring tests the portfolio as it actually stands on a stated date against every constraint the mandate wrote down, and against the stated base each constraint was written on. In the Anantara Multi-Asset Portfolio the largest holding is 4.6 per cent of the portfolio against a 5 per cent cap, and 4.6 sits inside 5. The identical holding is 7.7 per cent of the equity sleeve.

Two figures, one holding, no disagreement between them. The difference between the two figures is not disagreement but the base each was measured against. The Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore, is run by Faiz Ahmad Ansari for an invented charitable endowment whose investment committee is chaired by Rukmini Deshpande. 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. The three sleeves sum to Rs 500 crore exactly. The policy weights are that stated shape, and the actual weights drift away from them between one rebalancing and the next. The drift is precisely how a portfolio walks into a breach nobody decided on.

What is a compliance check actually testing?

A compliance check tests obedience, and only obedience. Compliance monitoringThe routine of comparing a portfolio as it actually stands on a date against the constraints written into its mandate, one constraint at a time, and recording the result. takes the portfolio as it stands on a date, takes the constraints as they were written, and asks of each one in turn: does the number on this date sit inside the number in the document? The question is narrow, and the narrowness is the feature. Anybody can re-run the same question later and reach the same answer, and that repeatability is what makes it worth doing at all.

Think of the goods vehicle at a weighbridge. The bridge does not ask whether the load was worth carrying, whether the route was sensible, or whether the driver is any good. The bridge asks one thing: is the weight under the number painted on the sign? The answer is yes or no, anybody standing there gets the same answer, and the answer says nothing whatsoever about the wisdom of the journey. A compliance check is a weighbridge for a portfolio.

A compliance check is not a quality check, not a performance check and not a risk opinion, and mistaking it for any of the three is the reason clean compliance records sit beside bad outcomes so often. Every one of the Anantara Multi-Asset Portfolio's four mandate constraintsA limit written into the document that sets up the mandate, which every later decision has to be taken inside. It is a promise about shape, not a forecast about return. can pass on the same date that the portfolio is badly built, poorly diversified and exposed to something nobody in the room has thought about. Passing means the document was obeyed. Passing does not mean the document was any good.

The four constraints of the Anantara mandate are: equity between 50 and 70 per cent of the portfolio; no single holding above 5 per cent of the portfolio; no unlisted holdings; and a minimum credit standing on the fixed income sleeve, stated as a policy rather than as a rating symbol. All the arithmetic below works on those four constraints.

Why does the base a limit is written against decide the answer?

Because every limit is a ratio, and a ratio is two numbers, not one. A cap of 5 per cent is a numerator over a denominator, and somebody chose that denominator when the mandate was drafted. The numerator is easy: the value of the holding, Rs 23 crore in the Anantara Multi-Asset Portfolio on the stated date. The denominator is the choice, and it is the choice that quietly decides whether anybody is told they breached a limit today.

Measured against the whole portfolio, the baseThe denominator a limit is measured against. Two per cent of what? The answer to that question is the base, and a percentage without one is not yet a measurement. is Rs 500 crore and the holding is 23 over 500, or 4.6 per cent. Measured against the Rs 300 crore equity sleeveOne asset class portion of a portfolio treated as a unit, such as the equity part or the fixed income part. A sleeve is smaller than the portfolio, so a weight measured against a sleeve is always the larger number. that holding sits inside, the same Rs 23 crore is 23 over 300, or 7.7 per cent to one decimal place. The same holding produces two different and both entirely correct answers depending on the base, and a monitoring routine that does not carry the base alongside the limit is checking an unknown quantity.

The Anantara mandate writes its concentration capA limit on how much of a stated base a single holding, or a group of holdings, is allowed to be. It restrains one position; it says nothing about what the positions have in common. against the whole portfolio, so the test is 4.6 against 5 and the position is comfortably inside. Had the identical sentence been written against the equity sleeve, the identical position would sit 2.7 points over. Nothing about the portfolio would have changed. Nothing would have been bought or sold. Only the sentence in the document would be different.

Set side by side, the two bases show the same pattern repeating all the way down the portfolio. The top ten holdings of the Anantara Multi-Asset Portfolio are Rs 155 crore, or 31.0 per cent measured against the portfolio and 51.7 per cent measured against the sleeve. Both numbers are honest. Neither is usable until the base it stands on is stated.

How much room does a 5 per cent cap actually leave?

Now to the arithmetic that catches almost everybody, and it catches them because the wrong answer feels finished. HeadroomHow far a position can move before it touches its limit. It is the distance between where a ratio sits today and where the ratio becomes a breach. is the distance between where a position sits today and where it breaches. The cap allows 5 per cent of Rs 500 crore, or Rs 25 crore. The holding is Rs 23 crore. So there appears to be Rs 2 crore of room, and Rs 2 crore on a Rs 23 crore holding is a rise of 8.70 per cent.

The 8.70 per cent answer is wrong, and it is wrong for the same reason the base problem exists. If that one holding rises and nothing else does, the portfolio does not stay at Rs 500 crore. The portfolio grows by exactly the gain on the holding. The denominator moves with the numerator, so a headroom computed on a frozen denominator is always too small.

Do it properly. Let the holding rise by a fraction x. The holding becomes 23 times one plus x. Only that one holding changed, so the portfolio becomes 500 plus 23x. Set the ratio equal to the cap:

23 times one plus x, all over 500 plus 23x, equals 0.05. Multiply out and 23 plus 23x equals 25 plus 1.15x. Collect the terms and 21.85x equals 2, so x equals 0.0915, or 9.15 per cent. Solving with the denominator allowed to move gives 9.15 per cent of price movement rather than 8.70 per cent, and the naive answer errs in the direction of raising a false alarm.

A solved equation that has not been checked is a guess with algebra on it, so the check comes before the belief. At a rise of 9.1533 per cent the holding becomes Rs 25.11 crore and the portfolio becomes Rs 502.11 crore. Divide: 25.11 over 502.11 is 5.00 per cent, exactly the cap. The check holds. The naive figure understates the true room by 0.46 points of price movement. Half a point of price movement is easily a normal week.

Which constraint gives way first?

Monitoring has a harder job than testing four numbers. Monitoring has to know which of the four is close. Constraints written in different units bind at different points, and the one that gives way first is very often not the one that looks tightest when the list is read. In the Anantara Multi-Asset Portfolio the equity band has 10 points of slack in either direction while the concentration cap has 0.4 points, so the cap looks like the tight one and the band looks like the loose one. The intuition is right here, but intuition is not the method, and on a differently shaped portfolio it would be wrong.

The method is to push each constraint arithmetically and compare. Take a broad equity rise: the whole Rs 300 crore sleeve gains a fraction x, fixed income and cash stay still. The equity band gives way when 300 times one plus x, over 500 plus 300x, reaches 0.70. Solving gives 300 plus 300x equals 350 plus 210x, so 90x equals 50 and x is 55.6 per cent. The concentration cap gives way when 23 times one plus x, over 500 plus 300x, reaches 0.05. Solving gives 23 plus 23x equals 25 plus 15x, so 8x equals 2 and x is 25.0 per cent.

Check the crossing point. At a 25.0 per cent sleeve gain the equity is Rs 375 crore and the portfolio is Rs 575 crore, so equity is 65.2 per cent of the portfolio and still comfortably inside the 50 to 70 band. The largest holding is Rs 28.75 crore, exactly 5.00 per cent. The concentration cap gives way at a 25.0 per cent sleeve gain while the asset class band survives to 55.6 per cent, so the binding constraintThe limit that a portfolio reaches first as conditions move. It is the one actually shaping decisions, whatever order the constraints appear in the document. here is the cap, and the document lists the band first.

How Portfolio Risk Is Monitored: what gets refreshed between reporting dates?

Between one reporting date and the next, monitoring reads exposures rather than outcomes. Monitoring refreshes the weight of each asset class against its band, the size of the largest single holding, the concentration in the top ten, the credit standing of the fixed income sleeve, and the presence of anything the mandate excludes. The return, the ratios and the split of the excess return are not refreshed, and the omission is deliberate rather than lazy.

A return records what already happened. An exposure shows the risk the portfolio carries right now, and only a current risk can still be acted on. If the largest holding has climbed to 4.9 per cent, somebody can sell part of it this afternoon. If last month's return was disappointing, nobody can do anything to last month. The interval between reporting dates is the window in which action is still possible, so the monitoring in that window looks at the things action can still reach.

The household version is a familiar one. Between salary dates the household does not review last month's spending; it checks what is left in the account and what is committed before the next credit arrives. The balance and the commitments are exposures. Last month's spending is an outcome. One of them can still change what happens next week.

Did a trade cause the breach, or did the market?

Two portfolios can report the identical number on the identical date and be carrying two entirely different problems. A holding sitting at 5.1 per cent against a 5 per cent cap got there one of two ways. Somebody bought it, or its price rose.

An active breachA breach that a decision created, such as a purchase that took a position over its limit. Somebody acted, so something in the control that should have stopped the action did not work. is one a decision created. Somebody placed an order that took the position over. A control that should have refused the order did not refuse it. A passive breachA breach that arrived because prices moved, with nobody trading. Nothing was decided, so the repair is a trigger that catches drift rather than a control on orders. is one that arrived because prices moved while nobody acted at all. The portfolio drifted into it.

The arithmetic separates them cleanly, and this is the payoff of the headroom work above. The money was already inside the portfolio, so switching Rs 2 crore of cash into the holding does not change the portfolio total. The holding becomes Rs 25 crore, the portfolio is still Rs 500 crore, and 25 over 500 is exactly 5.00 per cent. One rupee more and it breaches. So the naive 8.70 per cent figure is not a wrong answer at all; it is the right answer to the trade question and the wrong answer to the price question. A price rise needs 9.15 per cent because the denominator travels with the numerator. A purchase needs 8.70 per cent because the denominator does not move.

A passive breach is not a smaller event than an active one, it is a different event, and recording both simply as a breach destroys the only information that would say what to repair. The active case points at the order control. The passive case points at the drift trigger that should have flagged the position while it was still at 4.9 per cent. Fixing the first will never prevent the second.

How to Document Portfolio Compliance: what does every line have to carry?

A compliance record is not a note that somebody looked. A compliance record is a specification, and every line of it has to carry seven things: the constraint exactly as the mandate words it, the base it is measured against, the value on the date, the date itself, where the value came from, who checked it, and the headroom remaining. Seven fields, one line, one constraint.

The field that gets left out is the base. A line without its base cannot be re-run later by anybody, so it cannot be relied on, and a line nobody can rely on was never a record in the first place. Six months later, nobody in the room was in the room. The only thing standing between the reader and a guess is whether the line says what the denominator was. A line reading that the largest holding was 4.6 per cent and inside the cap is a sentence. A line reading that the largest holding was Rs 23 crore over the Rs 500 crore portfolio, being 4.6 per cent against a 5 per cent cap written on the portfolio, taken from the holdings record as at the stated date, is a record.

Written out for the Anantara Multi-Asset Portfolio on the stated date, the four constraints produce four lines, and two of them cannot be filled in at all from what the holdings record carries.

The equity line hides the same trap in a friendlier form. Ten points of room in each direction is Rs 50 crore either way. Moving Rs 50 crore between equity and cash leaves the portfolio at Rs 500 crore, and that makes the figure exact for a trade. The Rs 50 crore figure is not exact for a price move, for the reason already established. Even the comfortable line needs its base and its mechanism written beside it.

What did the check fail to test at all?

Two of the four constraints. The ban on unlisted holdings needs a listing status for every holding, and no listing status is carried in this record. The minimum credit standing needs a credit standing for every fixed income position, and none is carried either. The two constraints were not tested, not passed, and not failed.

A blank in a compliance record is read as a pass by whoever reads it next, and that is exactly how an untested constraint quietly becomes an assumed one, so the check ends by naming what it could not reach. Marking them untested costs one word per line and preserves the difference between a thing that was checked and a thing that was skipped.

How does an investment committee use this on a Tuesday?

Rukmini Deshpande's committee opens the pack at the constraint that is closest, not the constraint that is listed first. Reading the closest constraint first is most of the practical value of everything above. The pack she reads carries each constraint with its base beside it, the value on the date, and the headroom expressed in the units the thing actually moves in: points for a band that changes by trading, per cent of price movement for a cap that changes by drift.

Faiz Ahmad Ansari, running the mandate, uses the same arithmetic in the other direction. Before placing an order he wants to know the trade headroom, the Rs 2 crore figure. A purchase leaves the denominator alone. A price move carries the denominator with it, so the drift trigger sitting in the monitoring system uses the other figure, 9.15 per cent. Two numbers, two questions, and a system that carries only one of them will either block trades it should allow or miss drift it should catch.

A lender reading a borrower's covenant pack does the identical work. A covenant written as a ratio has a base, the base is in the document, and the first question on any reported breach is whether the reported ratio was struck on the base the covenant names. A household does it too, in miniature. The rule "no more than a tenth of savings in any one place" needs somebody to say whether savings means the deposits alone or the deposits plus the retirement account. The same money is inside the rule on one reading and outside it on the other.

What has a clean check established, and what has it not?

A clean check has established that the portfolio obeyed what was written, on one date, against the bases the mandate named. The finding is real and useful, and it is worth having. Obedience on one date is also the entire extent of the claim.

A clean record says nothing about whether what was written was sensible, whether the risks that actually matter were the ones constrained, or whether the portfolio is exposed to something nobody ever thought to write a limit against. The Anantara mandate caps any single holding at 5 per cent of the portfolio. Ten holdings each sitting exactly at that cap would be Rs 250 crore, or 50 per cent of the portfolio, and every line of the check would pass. The top ten actually sit at Rs 155 crore, or 31.0 per cent, so there are 19 points of permitted concentration the mandate would not object to. A per name cap restrains one name. A per name cap says nothing about what the names have in common.