Market Neutral: Stripping Out Market Direction

Market neutral is the third of eight hedge fund approaches. The book is built with a long side and a short side of the same size, so net exposure is nil and a market move cancels between them to the rupee. The difference between how the two sides perform is what is left, and that difference is the only thing the fund gains or loses on.

Start on a street rather than in a fund. Two vegetable stalls sit at the same market, six feet apart, selling much the same produce to much the same people. When the rain comes, both of them have a bad week. When a festival comes, both of them have a good week. If somebody had put money behind the first stall's takings and, at the same moment and in the same amount, had arranged to be paid whatever the second stall's takings fell by, then the rain would cost them nothing and the festival would earn them nothing. The weather would have been removed from their result entirely, and what would be left is one question only: did stall one do better or worse than stall two? The whole of the market neutral construction is in that question. Everything below is the same sentence at a larger size, in rupees, with the arithmetic written out.

The stall arrangement quietly assumes two things, and both of them have to be taken apart before the construction can be trusted. The first is that the two stalls really do get the same weather. The second is that arranging to be paid when a stall's takings fall is a thing somebody can actually do, at a cost, with a counterparty on the other side of it. Neither assumption is free, and neither is the same as the construction being empty of risk.

What is a market neutral book neutral to?

To the market move. Not to loss, not to error, not to anything else, and the precision of that answer is the single most useful thing to hold on to. Market neutralA book built with equal long and short sides so that a market move cancels between them. is not a description of how a book behaves. Market neutral describes how a book was built. Somebody chose the size of one side, then chose the size of the other side to match it, and the matching is the mechanism. Nothing about the securities in the book, nothing about the manager's skill and nothing about the state of the market makes a book neutral. Sizing does.

The word neutral is best read as a construction rather than as a property. A house is not warm because of the weather; it is warm because somebody built walls. If the walls come down, the weather is back, immediately and without warning. In the same way, if the two sides of this book stop being the same size, the book stops being neutral that same day, and no amount of describing it as neutral in a document written last year changes the arithmetic.

The figure that carries this is net exposureThe long side less the short side, measured against the fund's net assets. In this construction it is nil., which is simply the long side less the short side. When those two are equal, the subtraction gives nil, and nil is what makes the market move cancel. Nil needs saying plainly. Nil does not mean the fund holds nothing. Nil does not mean the fund cannot have a bad year. Nil means one specific arithmetic quantity, long less short, has been set to zero, and the consequences of that are exactly as narrow as the definition.

Neutral is not another word for safe. Matching the two sides removes the market's direction from the result. Matching does not remove the risk that the two sides move apart from each other, it does not remove the cost of running the position, and it does not remove the possibility of losing a great deal of money on a book that is described in a document as having nil exposure. Each of those is worked out in rupees below rather than asserted.

Why does making the two sides the same size do the whole job?

Because a market move is a proportion, and a proportion applied to two equal amounts in opposite directions gives the same rupee figure twice, once with a plus in front of it and once with a minus. There is nothing cleverer than that in the mechanism, and a reader who has understood a subtraction has understood the approach.

The book worked here is the following. Nilgiri Absolute Return Fund, an invented open-ended fund managed by Nilgiri Alternatives Advisors Private Limited, with a trustee, Nilgiri Trusteeship Services Private Limited, and a sponsor, Nilgiri Financial Holdings Private Limited, all three invented as well, has net assets of Rs 5,00,00,00,000 at its record date. The fund does not run this approach. Nilgiri runs approach 6, long-short equity, and its own book is long Rs 6,50,00,00,000 against short Rs 2,50,00,00,000. The case worked below is a counterfactualA constructed case showing what would follow if something were different, clearly labelled as not having happened.: the same Rs 5,00,00,00,000 of net assets, built neutral instead. The neutral book did not happen. Every figure below belongs to that constructed case and to nothing else.

The book is long Rs 5,00,00,00,000 and short Rs 5,00,00,00,000, against net assets of Rs 5,00,00,00,000. Long less short is Rs 0. The equality of the two sides is not a feature of the construction, it is the entire construction, and it reduces to a single drawing: three bars, two of which are the same length on purpose.

Read the bracket carefully. The bracket marks the only thing that has to be true. The two sides are the same size. The two sides need not be the same securities, need not be in the same businesses, and need not be related at all. The construction fixes the amount, and everything the approach claims follows from the amount and from nothing else. A document can therefore describe such a book in a single line and be complete.

What happens to this book when the market rises, and when it falls?

A reader shown only the rise walks away with a wrong idea. A reader shown only the fall walks away with a worse one. Both cases have to be worked, in that order. Take the rise first. Suppose the whole market rises 10.0 per cent and both sides of the book rise with it. The long side is Rs 5,00,00,00,000, so it gains Rs 50,00,00,000. The short side is also Rs 5,00,00,00,000, and a short position that rises is a loss, so it loses Rs 50,00,00,000. Add them: Rs 0.

Now the fall, and this is the reading people skip. Suppose the market falls 10.0 per cent instead. The long side loses Rs 50,00,00,000. The short side gains Rs 50,00,00,000. Add them: Rs 0 again. The construction does not lean towards falling markets any more than towards rising ones. Reading it as a way of sitting out a fall is the commonest misreading there is. A fall is removed with exactly the same completeness as a rise. The good weeks are removed with it.

The word for what has just happened is cancellationThe offsetting of a move on one side of the book by the opposite move on the other side., and it is worth being exact about what did the cancelling. The cancelling was not done by a clever choice of securities. The cancelling was done by the fact that 10.0 per cent of Rs 5,00,00,00,000 and 10.0 per cent of Rs 5,00,00,00,000 are the same number. Change either amount and the cancellation becomes partial. Change the assumption that both sides moved 10.0 per cent and the cancellation stops being about the market at all. The case where the two sides part company is worked below.

One sentence about the short side. A price that has been sold can keep rising without limit, and a price that has been bought can only fall to nil. A short position's loss is therefore not bounded the way a long position's is. The asymmetry, how the stock is borrowed, and what happens when a lender asks for it back are the mechanics of selling short, and they are covered separately. Here the short side is used as a quantity of rupees.

If the market move cancels, what is left to gain or lose on?

The spreadThe difference between how the long side performs and how the short side performs., and nothing else. Work it. Suppose the market rises 10.0 per cent, the short side moves exactly with the market, and the long side does 2.0 percentage points better, so it moves 12.0 per cent. The long side gains Rs 60,00,00,000. The short side loses Rs 50,00,00,000. The result is Rs 10,00,00,000, or 2.0 per cent of the fund's Rs 5,00,00,00,000 of net assets.

Now run the same spread in a falling market. The falling case is the reading that proves the claim. The market falls 10.0 per cent. The short side falls 10.0 per cent with it, so it gains Rs 50,00,00,000. The long side falls only 8.0 per cent, still 2.0 percentage points better than the market, so it loses Rs 40,00,00,000. The result is Rs 10,00,00,000. The market moved 20.0 percentage points between those two cases and the answer did not move at all. Two arithmetic lines carry the central claim of the approach.

A single identity sits underneath both of those, and it is worth memorising. The identity replaces every worked example above. The result equals the size of one side multiplied by the spread between the two sides. Rs 5,00,00,00,000 multiplied by 2.0 per cent is Rs 10,00,00,000, and that holds whatever the market did. The market term appears once with a plus and once with a minus, and leaves the identity untouched. A reader holding that sentence can work any case below without being shown it.

Does neutral mean the book is small?

No, and the two figures that answer that question are so far apart that a reader given only one of them has been handed a picture of a fund that is not there. Net exposure on this constructed book is Rs 0. Gross exposureThe long side plus the short side, measured against the fund's net assets. In this construction it is 200.0 per cent., which is the long side plus the short side rather than long less short, is Rs 10,00,00,00,000, being 200.0 per cent of the fund's Rs 5,00,00,00,000 of net assets. Both sentences are true of the same positions on the same afternoon.

Think about what that means physically rather than as a ratio. Somebody has bought Rs 5,00,00,00,000 of shares and has sold short another Rs 5,00,00,00,000 of shares, so there are Rs 10,00,00,00,000 of live positions sitting with a broker, every rupee of which can move on any given day. The fund behind them is worth Rs 5,00,00,00,000. If the whole gross book moved 1.0 per cent in one direction with nothing offsetting it, that would be Rs 10,00,00,000, or 2.0 per cent of the fund. Nil net exposure says nothing whatever about how much is at work, and the only figure that says how much is at work is the gross one.

What does the total do while the two sides swing?

Nothing at all, and that is a sentence nobody believes from reading it. The control below moves one thing, the market, from a fall of 20.0 per cent to a rise of 20.0 per cent. Two very large bars swing in opposite directions as the control moves. A third bar, drawn underneath them, sits at Rs 10,00,00,000 and never moves by a pixel, at any setting, in either direction. The question below comes before the control for a reason: the prediction is worth more than the demonstration.

The control has a twin, and the pair is worth knowing about. The difference between the two is exactly one figure. On a book whose two sides are not the same size, the same control produces a total that swings. Nilgiri Absolute Return Fund's own book, invented, is long Rs 6,50,00,00,000 against short Rs 2,50,00,00,000. Long less short leaves a net exposure of Rs 4,00,00,00,000, being 80.0 per cent of its net assets. Move the market on that book and 80.0 per cent of the move reaches the answer, every time, in proportion. How that book behaves is covered separately and is not drawn here. The same control, the same variable, and two opposite results, and the only thing that differs between them is whether long less short came to nil.

Can the spread go the other way?

The spread can go either way, by identical arithmetic, and working the favourable case alone would teach half a mechanism. With the same market rise of 10.0 per cent, suppose the long side does 2.0 percentage points worse than the short side rather than better, so it moves 8.0 per cent while the short side moves 10.0 per cent. The long side gains Rs 40,00,00,000. The short side loses Rs 50,00,00,000. The result is minus Rs 10,00,00,000, being 2.0 per cent of net assets in the other direction.

Look at what did not change between that case and the earlier one. The book did not change. The market move did not change. The construction did not change. Only the sign of the spread changed, and the answer flipped by exactly the same amount it had gained before. Removing the market's direction from a result does not make the remaining difference more likely to fall one way than the other, and nothing in the construction has an opinion about which side does better. The identity from earlier says the same thing in one line: the size of one side multiplied by a spread of minus 2.0 per cent is minus Rs 10,00,00,000, and a multiplication does not care about signs.

What if the two sides stop moving together?

Then nothing cancels, and this is the assumption the entire construction has been resting on since the first paragraph. Every worked case above quietly assumed that a market move of 10.0 per cent reaches both sides as 10.0 per cent. The shared move is what made the two rupee figures equal, and equal figures with opposite signs are the only reason anything cancelled. A shared move was never a rule. A shared move is a statement about how two collections of shares behave, and statements about behaviour stop holding.

Work the case where it stops. Suppose the market does nothing at all: no rise, no fall, nil. Suppose the long side falls 6.0 per cent anyway and the short side rises 6.0 per cent at the same time. The long side loses Rs 30,00,00,000. A short position that rises is a loss, so the short side loses Rs 30,00,00,000 as well. The result is minus Rs 60,00,00,000, being 12.0 per cent of the fund's net assets, on a day the market did nothing whatever.

Sit with the shape of that. The same shape belongs to every bad day this construction can have. An adverse move here is not a market falling; it is the two sides moving apart, with the longs going down and the shorts going up at the same time, and there is nothing in the book left to offset either of them. The identity holds even here: Rs 5,00,00,00,000 multiplied by a spread of minus 12.0 percentage points is minus Rs 60,00,00,000. The construction did not fail. The construction delivered the spread and nothing else, and the spread came in badly. How far a loss of that shape can run, and what a loss large enough to end a fund looks like, are covered separately by the treatment of a fund's worst outcomes.

Who sets the margin on a book twice the size of the fund?

The broker does, and it sets the margin on the gross figure rather than the net one. A construction that looks like nothing on paper starts costing real cash right there. Marudhar Securities Private Limited, invented, is the prime broker to Nilgiri Absolute Return Fund, invented. Marudhar holds the positions, lends the stock that is sold short and lends the cash that funds the long side. Everything that counterparty does and the concentration that comes from one firm doing all of it are covered separately. One of its contracted rates matters here.

Under this invented arrangement's own contracted terms, the maintenance marginCash and collateral the broker requires to be posted and kept against the positions it holds and finances. is 15.0 per cent of the gross value of a position. The 15.0 per cent is this fund's own negotiated rate, invented for the worked case. Rates of this kind are negotiated fund by fund and differ widely between one broker and the next. Apply it to the constructed neutral book. Gross positions of Rs 10,00,00,00,000 at 15.0 per cent means Rs 1,50,00,00,000 of margin posted against a fund whose net assets are Rs 5,00,00,00,000.

Now move the market 10.0 per cent up and let both sides move with it. The result from that move is nil. The long side is now worth Rs 5,50,00,00,000 and the short side is now worth Rs 5,50,00,00,000, so the gross value of positions is Rs 11,00,00,00,000. At 15.0 per cent that is Rs 1,65,00,00,000 of margin. The requirement is measured on a figure the cancellation never touched, so the book made nothing, lost nothing, and Rs 15,00,00,000 of cash still has to be found and posted. That is the third risk in plain form: neutrality is an arithmetic property of the result, and the machinery around the book is sized on the positions.

What is this book still exposed to once the market has been stripped out?

The list of what survived the cancellation is longer than most readers expect, and every item on it is something the cancellation was never built to touch. Take them in order, and note that not one of them is a market move.

Read that last row against the first. The pair is the honest summary of the approach. One thing was removed with complete success, and it was removed in both directions, so the fall and the rise went together. Everything else stayed exactly where it was. A construction that removes one risk perfectly is not a construction that has removed risk, and the difference between those two sentences is the whole of what market neutral offers.

How would somebody reading a factsheet actually use this?

Most people who meet this idea will never build such a book. Most will read about one instead, in a monthly sheet of figures or in a summary written by somebody else, and the whole value of the mechanism to them is that it turns a single number into a question. Four different readers meet the same two figures and do different work with them.

An analyst reading a document that says net exposure is nil now knows that the sentence is arithmetically complete and informationally empty, and asks for the gross figure next. Two numbers, nil and Rs 10,00,00,00,000, describe a book. One of them on its own describes nothing. The analyst also knows what to ask third: what the two sides have in common. The answer to that question is the assumption everything else rests on.

A risk or margin desk at the broker's end computes the cash it wants posted on the gross value of the positions, so it never looks at the net figure at all for its own purposes. The earlier arithmetic gave a call of Rs 15,00,00,000 on a move that produced no gain and no loss for exactly that reason. Somebody who understands that will not be surprised by a demand for cash on a flat month.

An investor in such a fund reads the two figures together and reads the spread separately from the market. In a month where everything rose, a flat result is what the construction says should happen and is not a sign that anything went wrong. In a month where nothing much happened at all and the fund lost a great deal, the construction also says that is possible, and the explanation is the two sides rather than the market. Neither reading tells anybody whether to be in such a fund.

A household meets the same shape without any of the vocabulary. Somebody running a wedding budget sets aside money for the caterer and, in the same breath, negotiates a discount that grows if food prices rise. Prices rising costs them on one line and pays them on the other, so the monsoon has been taken out of their budget. Whether their own caterer did better or worse than caterers in general is what is left, and that is the only thing that can still surprise them. The size of the wedding is the gross figure, and it has not changed at all. Nobody would call the wedding cheap merely because two lines cancel.