Tail Risk in Alternatives: The Loss That Ends the Fund

A tail risk in a fund like this is not one large loss. The kind that ends a fund is a loss that sets three of the fund's own mechanisms working against each other: gross exposure rises without a single trade, the gate turns redemption requests into a queue, and a position that cannot be sold is ring-fenced, leaving a smaller fund holding more of what nobody wanted. Every figure here belongs to Nilgiri Absolute Return Fund, invented.

One fact surprises people, and everything else in this guide follows from it. Two funds can lose exactly the same rupees on exactly the same day. One of them files its next report, carries on, and nobody outside its investor list ever hears about it. The other one stops existing in anything like the shape it had that morning. The difference between those two outcomes is not the size of the loss. The deciding factor is whether the loss switches on machinery that the fund built into its own contract, and once that machinery is running it does not need a second loss to keep going.

What makes a loss the kind that ends a fund rather than one it recovers from?

Think about a household first. The shape is identical and the rupees are small enough to feel. A household brings in Rs 40,000 a month and pays Rs 12,000 of rent. One month the hours get cut and the income falls to Rs 30,000. Notice what did not happen: the landlord did not raise the rent. The rent is exactly what it was. But it was 30 per cent of the income and it is now 40 per cent, and it got there entirely on its own, with nobody signing anything. Then the second thing happens. The household starts selling what it can sell quickly, and what it can sell quickly is not the same as what it would like to sell. The old cycle in the corner goes; the half-finished plot of land nobody wants does not. So the plot was one part of what the household had in the morning and is a bigger part of what is left by evening, and again nobody bought or sold it.

The household holds the whole mechanism at a scale small enough to keep in the head. A single event knocked the income down, and then two ratios moved on their own afterwards, in a direction nobody chose. Tail riskThe kind of loss that changes what a fund is rather than what it is worth. in a fund is that same sequence, run with a contract underneath it. The loss arrives once. The rest is arithmetic, and the arithmetic keeps going after the loss has stopped.

The fund worked in this guide is Nilgiri Absolute Return Fund, invented, managed by Nilgiri Alternatives Advisors Private Limited. The fund is open-ended and registered as a Category III Alternative Investment Fund. Long positions of Rs 6,50,00,00,000 and short positions of Rs 2,50,00,00,000 sit against net assets of Rs 5,00,00,00,000. Its prime broker is Marudhar Securities Private Limited. The broker holds the positions, lends the stock that is sold short, lends the cash that funds the long book and sets the margin.

What exactly does an adverse move of ten per cent mean here?

One sentence carries the whole argument, and it is worth being slow about. Read past it and everything downstream comes out wrong. An adverse move of 10.0 per cent against the fund on every position means two things happening at the same moment. The long positions fall by 10.0 per cent and the short positions rise by 10.0 per cent, simultaneously, and both of those are losses.

An adverse move of that kind is not a market going down, and not a market going up either. The two halves of the book move apart from each other, and the fund pays on both sides at once. The long book of Rs 6,50,00,00,000 falls to Rs 5,85,00,00,000, so Rs 65,00,00,000 is gone. The short book of Rs 2,50,00,00,000 rises to Rs 2,75,00,00,000, and because a short position loses money when the price goes up, that rise of Rs 25,00,00,000 is also a loss. Rs 65,00,00,000 plus Rs 25,00,00,000 is Rs 90,00,00,000.

The other reading is a different question, not a smaller version of the same one. Suppose instead the market simply rose 10.0 per cent and every position went up with it. The long book would rise to Rs 7,15,00,00,000, a gain of Rs 65,00,00,000. The short book would still go to Rs 2,75,00,00,000, a loss of Rs 25,00,00,000. Net assets would end at Rs 5,40,00,00,000, the fund would be Rs 40,00,00,000 better off, and gross exposure would read Rs 9,90,00,00,000 over Rs 5,40,00,00,000, being 183.3 per cent. The calculation is real and the answer is perfectly correct. The question it answers is a different one, and mixing the two produces a number that reconciles with nothing else.

Is the loss struck against the gross book or against the net figure?

Against the gross book, always, and this is where most readers lose the thread. A fund like this one publishes two exposure figures and they are both true at the same time about the same positions. Gross exposureEverything at work, long and short added together, against net assets. is the long book and the short book added together: Rs 6,50,00,00,000 plus Rs 2,50,00,00,000 is Rs 9,00,00,00,000, and against net assets of Rs 5,00,00,00,000 that is 180.0 per cent. Net exposureLong less short, against net assets. is the long book less the short book: Rs 6,50,00,00,000 less Rs 2,50,00,00,000 is Rs 4,00,00,00,000, being 80.0 per cent.

Think of a vegetable seller on a street corner. She has Rs 50,000 of her own money in the business and she has taken another Rs 90,000 of stock from a wholesaler on credit, so there is Rs 1,40,000 of vegetables on the cart. Now ask what a bad week is applied to. A bad week is applied to every crate on that cart. Every crate can rot. A bad week is not applied to her Rs 50,000, the amount left over once the wholesaler is paid. Her own money is the answer to a different question: how much of the cart is hers. The money at work and the money that is hers are two separate numbers, and the weather only knows about the first one.

A long-short book has exactly that shape. Every rupee in either book is exposed to something moving. A short position is not a cancellation of a long position; it is a second position, with its own way of losing money, sitting alongside the first. So when every position moves 10.0 per cent against the fund, the arithmetic is 10.0 per cent of Rs 9,00,00,00,000, giving Rs 90,00,00,000. Against net assets of Rs 5,00,00,00,000 that single move costs 18.0 per cent, and 18.0 divided by 10.0 is 1.80, the gross exposure ratio itself. The match is not a coincidence, and it is worth saying plainly: the gross exposure ratio is the multiplier that turns a move into a loss.

Why does gross exposure rise after a loss when nobody has traded?

Go back to the household and the rent. Nothing was done to the rent. The income moved, and the ratio moved with it, and the household could have sat completely still all month. A fund's exposure ratios behave the same way, and the reason is that the loss does not fall evenly on the top and the bottom of the fraction.

Follow it in rupees. The long book falls from Rs 6,50,00,00,000 to Rs 5,85,00,00,000. The short book rises from Rs 2,50,00,00,000 to Rs 2,75,00,00,000. Add them and the gross book is now Rs 8,60,00,00,000, or Rs 40,00,00,000 smaller than it was. The fund bore both halves of the loss, so net assets are down by the whole Rs 90,00,00,000, to Rs 4,10,00,00,000. So the top of the fraction fell by Rs 40,00,00,000 and the denominatorThe figure a ratio is divided by, which here falls at the same moment the numerator does. fell by Rs 90,00,00,000. Rs 8,60,00,00,000 over Rs 4,10,00,00,000 is 209.8 per cent. Gross exposure rose from 180.0 to 209.8 per cent and the manager did not place a single order.

Now the part that makes this worth working through rather than leaving as a footnote. Net exposure did something else over the same event. Net exposure was Rs 4,00,00,00,000 over Rs 5,00,00,00,000, being 80.0 per cent. Now it is Rs 3,10,00,00,000 over Rs 4,10,00,00,000, being 75.6 per cent. Net exposure went down. One loss, one moment, two exposure numbers, and they moved in opposite directions. A person reading only the net figure would close the report thinking the book had got a little quieter. From that number alone the conclusion is completely reasonable, and it is the opposite of what happened to the other one. Both halves have to sit in the same view, and a document offering one exposure number has left its reader to do a division it does not have the inputs for.

What do the two exposure readings do as the move gets bigger?

The two readings sit there at one setting. The question a reader always asks next is whether the gap between them widens or closes as the move grows, and that is a question better moved than asserted. The control below applies one adverse move to every position at once, from nothing up to 30.0 per cent, and redraws the two books, the fund and both exposure readings at every step.

Two things are worth watching as the control moves, and they are the entire point of it. The first is that the two markers travel in opposite directions along their own scales, starting at opposite ends and moving towards each other. The second is that the gross marker does not travel at an even pace. The book and net assets are both falling, and not at the same rate, so each further step of the control moves the marker further than the step before did. The gap between what a fund has at work and what a fund is worth widens on its own, and it widens faster the further the move goes.

What does a gate do to a fund once investors start asking to leave?

A gate turns a request into a queue, and the queue is longer at the next dealing date than it was at this one. The queue is the whole mechanism, and being precise matters here: a gate is not a failure of anything. The gateA cap on how much may be redeemed at one dealing date, with the excess carried forward. is a term this fund's own documents contain, agreed in advance by everybody who subscribed, and it is doing exactly what it was written to do.

Nilgiri Absolute Return Fund's contracted gate has two limbs. No more than 20.0 per cent of any one investor's holding, and no more than 25.0 per cent of the fund's net assets, may be redeemed at a single dealing date. Both are the fund's own terms, not a standard and not a requirement. Only the second limb matters to the arithmetic here. The loss moves that one. Dealing dates, the notice it takes to reach one, and the order in which this fund's four liquidity terms actually bite on a single request are all covered separately. The gate serves here as machinery already understood.

Here is the part that catches people. The cap is 25.0 per cent of net assets, and net assets are no longer Rs 5,00,00,00,000 but Rs 4,10,00,00,000. So the most that can be paid at this dealing date is Rs 1,02,50,00,000, where the same contractual term would have allowed Rs 1,25,00,00,000 the day before the loss. The cap is struck on what the fund is worth now, so a fund that has lost money can also pay out less. Anything requested above that figure is not refused and is not cancelled. The excess is scaled backCut by the same proportion for everybody so the total fits the cap. for everybody by the same proportion and carried to the next dealing date, where it joins whatever is asked for there.

Think of a caterer taking wedding bookings. Only so many plates can be served on one date. The bookings above that number are not thrown away; they go to the next available date, and that date now has its own new bookings plus everything carried over. Nobody has been treated unfairly and every booking is still alive. The next date is simply fuller than it would have been, and the date after that inherits whatever the next one could not take.

What happens when the only things left are the things that cannot be sold?

To pay the Rs 1,02,50,00,000 the gate allows, the manager has to raise Rs 1,02,50,00,000 of cash, and cash comes from selling. Here is where a fund's holdings stop being one undifferentiated pool and split into two groups that behave completely differently: the things somebody will buy today at a price that can be struck, and the things nobody will.

Nilgiri Absolute Return Fund holds one long position, carried at Rs 40,00,00,000, whose shares have been suspended from trading. Nobody is quoting a price for it because there is no market open in it. The fund's documents allow the manager to designate a holding that cannot be reliably valued as a side pocket. A side pocket is a separate class in which no subscription and no redemption is accepted until the holding is realised. The trigger for a side pocket, the way the units split, the valuation of the pocket and its release are all covered separately. Only the arithmetic consequence matters here.

Follow the Rs 40,00,00,000 through the three moments. Before any of this it was Rs 40,00,00,000 against net assets of Rs 5,00,00,00,000, being 8.0 per cent of the fund. After the loss it is the same Rs 40,00,00,000 against Rs 4,10,00,00,000, being 9.8 per cent. After the redemptions the gate allowed have been paid, net assets are Rs 3,07,50,00,000, and the same Rs 40,00,00,000 is 13.0 per cent. The one position nobody could sell went from 8.0 to 13.0 per cent of the fund without anybody buying a single share of it.

ConcentrationHow much of what is left sits in one position. has arrived by subtraction rather than by purchase, and the step is the one readers miss most often. Every instinct says a position gets bigger only when more of it is bought. The household selling things to cover a bad month shows the same movement. The plot of land nobody wants was one item among many in the morning. By evening, with the cycle and the two gold bangles gone, it is a much larger fraction of what the household still has. Nobody made a decision about the plot. Every decision was about the other things.

Why does the financing not shrink when the fund does?

Because it is a debt, and a debt is a number written in a contract rather than a share of anything. Nilgiri Absolute Return Fund's long book of Rs 6,50,00,00,000 is larger than its net assets of Rs 5,00,00,00,000, and the gap of Rs 1,50,00,00,000 between them is cash lent by Marudhar Securities Private Limited, the fund's prime broker. The lending is what makes a Rs 5,00,00,00,000 fund able to hold Rs 9,00,00,00,000 of positions in the first place.

Now put the loss through it. Net assets fall by Rs 90,00,00,000 to Rs 4,10,00,00,000. The Rs 1,50,00,00,000 does not fall by anything. Nobody repaid any of it, and a lender does not write down a loan because the borrower had a bad week. The loan was 30.0 per cent of net assets and is now 36.6 per cent. The one item in the whole chain that does not shrink with the fund grows as a share of everything else purely by standing still.

Compare it with the management fee, the opposite kind of number. The fee is 2.00 per cent a year on net assets, so it was Rs 10,00,00,000 and is now Rs 8,20,00,000, and after the redemptions the gate allowed it would be Rs 6,15,00,000. A fee expressed as a proportion shrinks when the fund shrinks. A loan expressed in rupees does not. The contrast is the whole of what leverage does to a smaller fund, and it needs no view about anybody's judgement to hold.

There is a second item alongside it, and the important thing about it is who sets it. The fund's contracted margin requirementWhat the broker demands the fund posts, which the fund does not set. with its prime broker is 15.0 per cent of the gross value of a position. The rate belongs to this one invented arrangement and to nothing wider. On the gross book of Rs 9,00,00,00,000 that is Rs 1,35,00,00,000, and on the smaller book of Rs 8,60,00,00,000 it is Rs 1,29,00,00,000. So the requirement in rupees actually fell. As a share of net assets it went from 27.0 to 31.5 per cent. Net assets fell by more. And the rate itself belongs to the broker: it is a term of the fund's agreement with Marudhar Securities Private Limited, and the fund cannot change it by deciding to.

How do the four steps feed each other?

Set the four steps out in order and what stands out is how ordinary each one is. Not a single step in this chain requires an unusual event, a manager who did something foolish, a counterparty who behaved badly or a market that did anything it has not done before. Each step is a division or a subtraction, and each one is caused by the step in front of it.

Step one is the loss. Every position moves against the fund and the loss is struck against the whole gross book, so Rs 90,00,00,000 leaves a fund whose net assets were Rs 5,00,00,00,000. Step two happens with no decision at all: gross exposure rises to 209.8 per cent because the denominator fell faster than the numerator. Net exposure falls to 75.6 per cent and tells the opposite story about the same event. Step three is people, and it is the only human step in the sequence: investors ask to leave, the gate caps one dealing date at Rs 1,02,50,00,000, and what is asked for above that is scaled back and carried. Step four is what raising that Rs 1,02,50,00,000 leaves behind: the sellable things are sold, the unsellable thing is ring-fenced, and it is now 13.0 per cent of a fund that is Rs 3,07,50,00,000.

Nothing in that sequence needed a second loss to keep it moving. The sequence is what people mean when they say a loss ended a fund rather than hurt it. The event was one moment. The mechanism ran on afterwards, out of the fund's own contract and its own arithmetic, in a fixed order, and every step of it was written down in advance by the people who set the fund up.

Why is no probability put on any of it?

Because there is no basis for one, and a number without a basis would be worse than no number at all. A fund coming apart naturally raises a further question: how often does that happen? The honest answer is that the frequency is not known.

Three limits hold this account in place and each of them is doing real work. The first is frequency. The whole meaning of the word tail is that the event is rare, and rare events are exactly the ones a short record cannot count, so a rate cannot be read off a handful of cases. The second is provenance. The arithmetic was constructed to make a mechanism visible, and a constructed case can show how a mechanism works without showing how often it fires. The third is judgement. Whether the fund was well run, whether the gate was a good term, whether the leverage was too much: not one of those questions is settled by arithmetic, and arithmetic is what a chain of divisions supplies.

One statement survives all three limits, and it is the certain one: if a fund with a book larger than itself takes a loss on that whole book, the arithmetic that follows is fixed, and it does not need anybody's permission or anybody's mistake to run. The statement describes a mechanism rather than warning anybody off a kind of investment.

How does somebody reading a fund document actually use this arithmetic?

Far more people read a fund's documents than ever run a book, and here is where the arithmetic earns its keep. Somebody sitting with an offering document or a quarterly factsheet is doing arithmetic, whether or not they name it, and its three parts are now in place.

The first part is the relationship between the two exposure numbers. If a document gives net exposure and not gross exposure, the loss on a move against every position cannot be worked out from it. The move applies to the sum of the two books, and net exposure is what is left when one is taken from the other. If it gives gross exposure, the multiplier is sitting there in plain sight: a gross exposure of 180.0 per cent means a 1.0 per cent adverse move across the book is 1.80 per cent of net assets, and the arithmetic scales from there. A reader who has both numbers can reconstruct both books; a reader who has one has neither.

The second part is that every one of these ratios has a date attached to it and the date belongs to the denominator. Net assets are struck by the fund's administrator on a particular day. In this arrangement that is Kolar Fund Services Private Limited, where Ashwin Baliga strikes the net asset value, and every exposure percentage published anywhere is that day's book divided by that day's net assets. The 180.0 per cent and the 209.8 per cent worked here are the same fund on two different days, and the difference between them is entirely in the second number of the division.

The third part is the terms. A gate, a notice period, a dealing calendar and a side pocket are contractual terms sitting in the fund's own documents, and what they say determines what a loss turns into afterwards. Two of them are used here; how each one works, and the order in which all four bite on a single request, are covered separately. The arithmetic underneath them is what has been added here: a cap expressed as a percentage of net assets is a cap that moves when net assets move, and a position moved into a pocket is a position that goes on being the same rupees while everything around it changes.

One last thing about what a document will show afterwards, stated so it does not surprise anybody. The fund's net asset value a unit falls with its net assets. The fall takes it below the high-water mark that the performance fee is charged against, so no performance fee arises until that mark is regained. How the mark works, and the record behind it, are covered separately. The management fee is a percentage of net assets rather than a percentage of a gain, so it goes the other way: it falls from Rs 10,00,00,000 a year to Rs 8,20,00,000 without anybody renegotiating a single term.