Side Pocket: Ring-Fencing an Illiquid Holding

A side pocket is a separate class of units holding one position the manager has designated as not reliably valuable. On designation every investor's units split in the same proportion, so nobody can redeem out of the pocket and nobody subscribing afterwards buys into it. In Nilgiri Absolute Return Fund, invented, a Rs 40,00,00,000 designation split a Rs 5,00,00,00,000 fund into Rs 4,60,00,00,000 of dealing units and the pocket.

The whole idea was invented on a street long before anybody wrote it into a trust deed, so start on a street rather than in a fund. Four neighbours put money into two things together: a small shop, and a plot of land at the edge of town. The shop is easy. The shop takes money at the counter every evening, it has rent and wages, and two buyers have made offers in the last year, so a quarter of the shop can be valued today and the valuation defended. The plot is not easy. There is a dispute over the title, nobody will make an offer while the dispute runs, and the last honest sentence anybody can say about it is that it is worth something and nobody knows what.

Now one of the four wants out. He is moving cities and he wants his share settled this month. And here is the problem, and it is not a valuation problem at all. Whatever figure the other three put on the plot when they settle him is a figure they cannot check, and every rupee of error in it comes out of, or goes into, their own pockets. Put the plot at a high number and the man leaving walks away with money the remaining three will never see. Put it at a low number and he has been short-changed by neighbours who then quietly gain. No number is knowable, so no number is safe.

The four neighbours, if they are sensible, split the arrangement in two. They settle the departing neighbour in full on the shop, today, at a figure everybody can check. And they leave his one-quarter interest in the plot exactly where it is, alive, unsettled, and untouched, until the dispute clears and the plot is genuinely sold to somebody. On that day, whenever it comes, he gets a quarter of whatever the plot actually fetched. Not a quarter of a guess. A quarter of a price.

The arrangement those four neighbours have just built is a side pocketA separate class of units holding one position that cannot be reliably valued, in which no dealing happens.. Everything below is the same idea run inside a fund, with the arithmetic done to the rupee on one invented vehicle, and with the four questions a reader should be able to answer afterwards: what pulls the trigger, how the units split, what the pocket costs while it sits there, and what happens on the day it ends.

What is a side pocket, and what is it actually separating?

A side pocket is a second class of units inside a fund that already has one. The whole structural fact is that plain, and it is worth saying before any arithmetic arrives. Most readers first meet the phrase in a story about something going wrong, so they arrive expecting a punishment rather than a partition.

Nothing leaves the fund when a pocket is created. Nothing is sold, nothing is written off, nothing is transferred to the manager, and no investor is asked to agree to anything on the day. One line of the fund's holdings stops being counted in the class where dealing happens and starts being counted in a class where dealing does not happen. The assets sit exactly where they sat the day before. The change is in which class of units has a claim on which asset.

The fund in every worked figure here is Nilgiri Absolute Return Fund, an invented open-ended vehicle managed by Nilgiri Alternatives Advisors Private Limited and held by Nilgiri Trusteeship Services Private Limited as trustee. Its net asset value is struck by Kolar Fund Services Private Limited as administrator. In a fund of this Indian shape there is no partnership and no general partner: the role a reader will see called the general partner elsewhere is discharged between the manager and the trustee, and the contract is a trust deed rather than a partnership agreement. Its net assets at the record date are Rs 5,00,00,00,000.

One long position in that fund, carried at Rs 40,00,00,000, holds shares that have been suspended from tradingNo longer quoted, so no price exists to mark the holding against.. Rs 40,00,00,000 against Rs 5,00,00,00,000 of net assets is exactly 8.0 per cent of the fund, and that 8.0 per cent is the number every part of the mechanism below turns on.

So there are two classes once that move is made, and each one has a different set of rules attached to it. The class that can still deal holds everything that has a price, and the class that cannot deal holds the one thing that does not. The picture below is the structure. The amounts are the easy part, so read the rules inside the two right-hand boxes instead.

Why does a fund need a second class at all?

Because of the departing neighbour, and the problem he creates is a fairness problem rather than a valuation problem. Fairness and valuation get confused constantly, and the difference between the two is the whole of this section.

An open-ended fund does something a closed-end one never has to do: it puts a single number on itself, repeatedly, and then pays and charges real money against that number. Every dealing date it strikes a price, somebody joins at that price, somebody leaves at that price, and cash moves. If one holding inside the fund cannot be reliably valued, then that single number contains a figure nobody can check, and the money moving against it is real money belonging to people who are still there.

Run it forward. Suppose the suspended position is carried too high. Whoever redeems on that dealing date is paid out of a pot that is smaller than the pot the fund thought it had, so the leaver is overpaid, and the shortfall is carried by the investors who stayed. Suppose instead it is carried too low. Now the leaver is underpaid, walks away with less than a fair share, and the investors who stayed are quietly enriched at his expense. The transfer is caused by the uncertainty and not by the direction of the error, so no carrying figure avoids it.

Notice what the transfer rules out. A better estimate does not solve it. A more careful estimate makes the error smaller, and a smaller error is still an error, and money still moves in one direction or the other on every dealing date until the position is finally sold. The only way to stop money moving on a figure nobody can check is to stop dealing on that figure at all. The second class stops the dealing on that figure, and stopping it is the only thing the second class does.

One word is worth setting down before the mechanism goes further, and it is the word most people reach for and the word most easily misused. To ring-fenceSeparating one holding so that its uncertainty cannot reach any other price. a holding is to separate it so that its uncertainty cannot reach any other price. A fence is not a protection, not a promise about outcomes, and not a statement that the holding is bad. A fence keeps two things apart. The fence says nothing at all about which side of it is the better place to stand.

What puts a holding into a pocket, and who pulls the trigger?

One condition, and it is narrower than most readers expect. The manager designates the holding as one that cannot be reliably valued. Reliable valuation is the test in this fund's own documents, and the test is worth reading twice for what it does not say.

The test does not say the holding has fallen. A price on a live screen exists and anybody can check it, so a position can halve every day of the week and still be perfectly valuable. It does not say the holding is large. Size decides how much of everybody moves, and it decides nothing about whether the trigger is pulled. It does not say the holding is bad, a mistake, or in trouble. It says a price cannot be got.

In this fund the trigger is factual and dull, as it should be. The shares behind one long position have been suspended from trading. There is no quotation. The last traded figure belongs to a day that has gone and to a market that has stopped. Palani Valuation Advisors, an invented limited liability partnership (LLP), is the independent valuation agent this manager uses. Putting a defensible figure on a holding with no price is a real craft with real machinery behind it. The valuation machinery is covered separately. Only its outcome matters for the split: the manager has said this position cannot be reliably valued, and the fund's own documents give the manager that power.

The act itself has a name. The designationThe manager's act of moving a holding into the pocket, on the day it is done. is the manager's decision to move the holding across, and it takes effect on the day it is made. Everything worked below is measured from that day.

The question worth asking about a pocket is never whether one exists, but who is allowed to create one, against what written test, and what the investors are told on the day it happens. The trigger, the decision and the notice sit in the fund's constitutional papers rather than in the arithmetic, and they are where a careful reader spends their attention. Whether an arrangement of this kind is permitted at all, and on what conditions, is a matter for the Securities and Exchange Board of India, named in the sources block below.

How do the units split on the day it is designated?

Proportionally, across the whole register, on the same day, in the same ratio, whoever the holder is. There is no selection, no application, no queue and no exception. The manager's own units divide on the identical rule, and so do the units of somebody who had already asked to leave that quarter.

The ratio is the one number already on the table. The designated position is Rs 40,00,00,000 and the fund's net assets are Rs 5,00,00,00,000, so 8.0 per cent of the fund crosses into the pocket and 92.0 per cent stays in the class where dealing happens. The dealing classThe units that can still be issued and cancelled at a struck price. is therefore Rs 4,60,00,00,000, and Rs 4,60,00,00,000 plus Rs 40,00,00,000 is Rs 5,00,00,00,000 to the rupee.

Now put one investor under it. Take a holder with Rs 25,00,00,000 in this fund on the day before the designation. Eight per cent of Rs 25,00,00,000 is Rs 2,00,00,000, and ninety-two per cent is Rs 23,00,00,000. So that investor now holds Rs 23,00,00,000 in the dealing class and Rs 2,00,00,000 in the pocket. Rs 23,00,00,000 plus Rs 2,00,00,000 is Rs 25,00,00,000, exactly what the investor held the day before. Nothing was created and nothing was destroyed. One holding became two.

Both columns add back exactly, and that is the check a reader can run on any pocket anywhere: the two parts must reconstruct the whole, at the fund level and at every holder's level, to the rupee. If they do not, something other than a designation has happened.

Now the part that trips people, and it is worth slowing down for because three different percentages are live in this one example and they have three different denominators. Rs 2,00,00,000 is 5.0 per cent of the Rs 40,00,00,000 pocket. The same Rs 2,00,00,000 is 8.0 per cent of the investor's own Rs 25,00,00,000 holding. And the investor's whole Rs 25,00,00,000 is 5.0 per cent of the fund's Rs 5,00,00,00,000 of net assets. The investor's share of the pocket equals its share of the fund, at 5.0 per cent, and the pocket's share of the investor equals the pocket's share of the fund, at 8.0 per cent, and those two sentences are the entire definition of a proportional split. A percentage quoted here without its denominator attached has said almost nothing, and the same figure of Rs 2,00,00,000 reads as 5.0 per cent, 8.0 per cent or 0.4 per cent depending on the denominator.

What does the dealing class look like the next morning?

Smaller by exactly what left it, and this is the single most useful fact in this guide for anybody who suspects a pocket of being stagecraft.

Before the designation, every price this fund struck rested on Rs 5,00,00,00,000 of net assets, of which Rs 40,00,00,000 was a figure nobody could check. After it, every price rests on Rs 4,60,00,00,000, and every rupee of that has a price behind it. The base got smaller by the whole of the designated amount, immediately, on the day, with no phasing and no averaging.

The shrinkage answers the suspicion, so say it plainly. Creating a pocket makes the number the fund deals on go down, not up, and it goes down by the full carrying amount of the position that moved. A mechanism that made a fund look better would have to make its dealing base bigger or leave it alone. A pocket shrinks it on day one.

One caution on how to read the smaller number. The dealing base is smaller because the class holds less, not because anything has been marked down. The position is still carried at Rs 40,00,00,000; it is simply carried somewhere else. And because this fund is open-ended and units are issued and cancelled continuously, the only honest way to state any of this is in rupees at the fund and holder level. The unit count therefore moves from one dealing date to the next, and a fund total divided by a moving unit count would not stay true for long.

What does somebody subscribing the next day actually buy?

Dealing units, and nothing else. No share of the pocket, not a rupee of it, not at any price. The rule is easy to state and hard to feel, so it is worth going back to the plot of land.

A fifth neighbour turns up next month and wants in. What is he buying? The shop, obviously: a share of a business with a countable till and a defensible figure. Is he buying a share of the disputed plot? He cannot be, and he should not want to be. Nobody can put a defensible figure on the plot, so nobody can tell him what to pay for a share of it. Any number he pays would be a transfer between him and the four already there, in one direction or the other, decided by nothing. So he buys into the shop, at a shop price, and the plot stays with the four who were there when the dispute began.

The fund does exactly that. A subscription received after the designation date is issued dealing units at the dealing price, struck on Rs 4,60,00,00,000, and it carries no interest in the pocket whatsoever. The pocket belongs, in fixed proportions, to whoever held units on the designation date and to nobody else. The pocket is a closed set from the day it is made, and it stays closed until the day it ends.

That single rule is the reason a pocket exists at all: it makes the class of people exposed to the unknowable holding a fixed list, decided on one day, that cannot grow and cannot shrink. Everything else follows from it. Everything in the dealing class has a price, so the class can take in new money and let old money out all it likes. The pocket does neither, and nothing in it has a price.

How is the pocket carried while it sits there, and what does it cost?

At its designation valueThe value the position is carried at from the moment it enters the pocket. of Rs 40,00,00,000, the figure the position was already carried at on the day it crossed. The designation value does not move afterwards. Nothing exists to strike a price against, so there is no quarterly remark, no fresh estimate each dealing date, and no price struck on the pocket at all.

Now the fact the rest of this depends on. Rs 40,00,00,000 is an estimate rather than a price, and what the position eventually turns into cash for may be more than that or less than that. Whether that figure is conservative, prudent, generous or right is not something anybody can say. The carrying figure stands in for a price nobody can observe, and it is the best that can be said until somebody actually pays for the holding. Both directions of that are worked further down.

Andrew Ang's Asset Management, 2014, is where the idea that illiquidity is a property of an asset to be described rather than an accident to be apologised for is set out at length. The framing of illiquidity as a describable property is what is borrowed; what illiquidity does to any return is a separate subject.

The cost of a pocket is easier and rather more surprising. Under this fund's own contracted terms, the pocket bears no management fee and no performance fee for as long as it sits there. The manager is not paid on the pocket while it is a pocket. Charging no fee on the pocket has an arithmetic consequence that runs the opposite way to what most readers expect. This fund's management fee is 2.00 per cent a year charged on net assets. On Rs 5,00,00,00,000 that is Rs 10,00,00,000 for a year. With Rs 40,00,00,000 sitting in a pocket that bears none, the charge runs on the Rs 4,60,00,00,000 dealing class instead and comes to Rs 9,20,00,000, which is Rs 80,00,000 less. Creating a pocket reduces the manager's fee while the pocket is open, and it keeps reducing that fee for as long as the position cannot be sold.

Two different fee mechanisms are now running in the same fund, and they are different objects. The dealing class carries a performance fee of 20.0 per cent a year subject to a high-water mark. A high-water mark is carried forward on a unit price so that the same gain is never charged twice. The full record of that mark for this fund runs to five years and is covered separately. At the end of Nilgiri Absolute Return Fund's own Year 4 the unit stood at Rs 118.00 after fee, and at the end of its Year 5 at Rs 122.00 after fee.

The pocket has no unit price, so it can have no high-water mark. Its fee is charged once, at the end, on the excess of the cash actually received over the Rs 40,00,00,000 designation value, and on nothing else. A mark carried forward and a single charge on a single excess are not the same machinery, and a reader who assumes the pocket simply inherits the fund's high-water mark will get the next section wrong.

When is the pocket released, and who is paid?

When the position is realisedTurned into cash, which is the only event that dissolves a pocket., and on no other occasion. Not when the manager decides the worst is over, not when a fresh estimate looks firmer, not at the end of a year and not at the request of anybody holding pocket units. Cash arrives, the class is dissolved, and the proceeds go to whoever held pocket units on the designation date, in the proportions fixed on that day.

So a pocket has exactly three events in its whole life, and it is worth counting them because the middle one is the one people find hardest.

Now the arithmetic of the payment itself, worked upwards first. Suppose the suspended position is eventually sold for Rs 52,00,00,000. The excess over the Rs 40,00,00,000 designation value is Rs 12,00,00,000. The performance fee is 20.0 per cent of that excess, being Rs 2,40,00,000, and it is charged on the excess alone rather than on the whole amount received. So Rs 52,00,00,000 less Rs 2,40,00,000 is Rs 49,60,00,000, and that is what goes to the holders of pocket units.

The same arithmetic taken down to the one investor: it held Rs 2,00,00,000 of the Rs 40,00,00,000 pocket, or 5.0 per cent of the pocket. Five per cent of Rs 49,60,00,000 is Rs 2,48,00,000. Against the Rs 2,00,00,000 it was carrying, that is 1.24 times. The shape of that payout has a kink in it at exactly Rs 40,00,00,000, and the kink is the whole of the fee arrangement drawn as a picture.

What if it is sold for less than it was carried at?

Then the pocket holders receive less than they were carrying, no fee is charged on anything, and the dealing class is exactly where it was. Only the pleasant direction would leave a false shape of the mechanism behind, so the other half runs on the same numbers.

Suppose the position is sold for Rs 10,00,00,000 rather than Rs 52,00,00,000. There is no excess over the Rs 40,00,00,000 designation value, so there is no performance fee: not a reduced one, not a deferred one, none. The whole Rs 10,00,00,000 goes to the pocket holders. The same investor's 5.0 per cent share of the pocket is Rs 50,00,000, against the Rs 2,00,00,000 it was carrying, or 0.25 times.

Put the two outcomes beside each other and read the bottom line rather than the top ones. The same pocket returns 1.24 times in one outcome and 0.25 times in the other, and in both of them the dealing class stands at Rs 4,60,00,00,000 and does not move by a rupee. The fence exists for exactly that. The fence works in the direction nobody wanted just as reliably as in the direction everybody hoped for, and a fence that only held in one direction would not be a fence at all.

Who is on which side of the fence, and does it favour either?

Neither, and the arithmetic is what says so rather than anybody's good intentions. Go back to the reason the second class exists at all. Without it, one figure nobody can check sits inside every price the fund strikes, and every rupee of error in it moves money between whoever deals that day and whoever stays. The fence does not make that money appear or disappear. The fence stops the transfer.

Read the two sides carefully. Somebody leaving is settled fully on the dealing class, at a price where every rupee has a price behind it, and keeps a pocket interest that pays when and only when the position is turned into cash. Somebody staying is no longer funding, or being funded by, an estimate that leaves the building. And somebody arriving next month buys into a class where nothing is unknowable. None of those three is better off than before the designation: what has happened is that a transfer between them, in whichever direction the error happened to run, has been stopped.

The claim is a narrow one and it should stay narrow. It says nothing about whether the position was a good one, whether the manager should have held it, whether the pocket is a comfortable thing to hold, or how long anybody will wait. A pocket is a partition, and a partition has no opinion.

What is actually worth looking for in a document that mentions one?

This is the practical end of the subject, and it is the end that matters most, because almost nobody reading this will ever create a pocket and a good many will one day read a document that mentions one. Analysts do this for a living. So do the people at an institution deciding whether to keep money with a manager, and so does anybody at home reading a factsheet properly rather than glancing at the headline figure.

Four things, in this order, and every one of them is a question about the document rather than about the outcome.

First, the trigger, read as written. A document that says a holding may be designated where the manager considers it cannot be reliably valued has given the test. A document that says a holding may be designated where the manager thinks it appropriate has given almost nothing. The wording between those two extremes is where the real information is, and it takes thirty seconds to find.

Second, who pulls it and whether anybody has to agree. Is the designation the manager's alone? Does it need the trustee to be told, or to agree? Is there an investor committee, and is it consulted before or informed after? None of these arrangements is better than another in the abstract. The reader's job is to know which one the document in hand sets out.

Third, the fee treatment, in both places. Does the pocket bear the management fee while it sits there, or not? Is the performance fee charged on the excess over the designation value, as in the invented fund worked here, or on the whole amount received, or not at all? The difference between charging 20.0 per cent on Rs 12,00,00,000 of excess and 20.0 per cent on the whole Rs 52,00,00,000 received is Rs 2,40,00,000 against Rs 10,40,00,000, on the identical outcome. The difference is not a detail.

Fourth, what investors are told and when. A designation happens to an investor's holding without that investor's agreement, so the notice arrangement is the whole of the protection against finding out late. The four things worth reading are the trigger, the decision, the fee and the notice, and every one of them is written down in the document rather than discoverable from the outcome.

One last practical note, and it is the one people most often want. How long a pocket lasts has no honest answer. A pocket ends when the position is turned into cash, and there is no schedule for that. The absence of a schedule is exactly why the position was pocketed. Anybody quoting a usual duration for a side pocket is quoting something nobody knows.