Tactical Asset Allocation: Deviating Within the Limits

Tactical asset allocation is a deliberate, temporary departure from a written policy mix, taken because somebody holds a view and bounded by limits agreed before that view existed. In the Anantara Multi-Asset Portfolio the mandate's equity band of 50 to 70 per cent sets the size: against a policy weight of 60 per cent the deviation may run ten points either way, and no further.

A departure from a written mix is arithmetic about a shape, and the arithmetic runs the same way whether the view behind the departure turns out well or badly. Which way anybody should lean is a different question with a different kind of answer, and no amount of computing the shape reaches it.

The running example is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore run by Faiz Ahmad Ansari for an invented charitable endowment whose investment committee is chaired by Rukmini Deshpande. Its policy mix 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 mandate permits equity to sit anywhere between 50 and 70 per cent. The permitted range, together with the return and volatility assumptions the holder chose, fixes everything a deviation can do to this portfolio.

What is tactical asset allocation, and which three words carry it?

Three words do all the work: deliberate, temporary and bounded. A tactical deviationThe gap between what a portfolio actually holds in a class and what its written policy says it should hold. Measured in percentage points of the whole portfolio. is deliberate because somebody decided it, temporary because it is expected to be reversed, and bounded because the limits that contain it were written down before anybody had the view that now wants to use them. Remove any one of the three and the activity turns into something else with a different name and a different owner: without deliberate it is drift, without temporary it is a change of policy, and without bounded it is simply an unconstrained mandate.

The first condition is the one people skip. Take it seriously. Suppose equity prices rise through a quarter and nobody trades anything. The Anantara portfolio's equity weight climbs from 60.0 per cent towards 63 per cent purely because the equity sleeve grew faster than the rest. Nothing about that is tactical. Nobody formed a view, nobody sized anything, nobody wrote a sentence. The climb to 63 per cent is policy driftThe movement of actual weights away from policy weights caused by prices alone, with no decision behind it. It happens between one rebalancing and the next in every portfolio that holds anything that moves., and it is arithmetic rather than judgement. The distinction sounds pedantic until one counts how many portfolios are described as tactically positioned when what actually happened is that nobody traded for eighteen months.

The second condition, temporary, is what distinguishes a tactical position from a change of policy. Both move the portfolio. Only one of them expects to move it back. Suppose a committee raises equity and has no intention of ever lowering it again. The committee has amended the policy mix and should say so in the document. The holder is answerable against the document. Calling a permanent change tactical is not a naming quibble: it means the written policy no longer describes the portfolio, and every later review is measuring against a shape that has quietly stopped being real.

Bounded, the third condition, carries most of the arithmetic below. A limit is the only part of the arrangement anybody can compute. The Anantara mandate says equity between 50 and 70 per cent. The band was written when nobody had any particular view, and its authority comes from exactly that. A limit set in ignorance can hold against a view somebody now holds very strongly indeed.

How Tactical Asset Allocation Works

Run as a sequence, it is six steps, and they happen in this order because each one needs the answer to the one before it. Form a view. State what evidence would show the view to be wrong. Size the position inside the band. Decide which class funds it. Take it. Write down the condition under which it closes. Steps two and six are the ones that get skipped, and they are precisely the two that make everything else reviewable, so a process that drops them has kept the trading and thrown away the discipline.

Notice what each step leaves behind. Step one leaves a sentence. Step two leaves a test. Step three leaves a number in percentage points and in rupees. Step four leaves the name of a class. Step five leaves a trade record. Step six leaves a condition. Six written artefacts, none of them long, and together they are the only thing that lets anybody twelve months later say what was intended and whether it happened. Without them there is a portfolio that is shaped a certain way and nobody who can explain why.

The sequence contains no method for forming the view. Deciding whether one class is more attractive than another is not arithmetic, and no step says how to do it. The sequence takes a view as given, from wherever it came, and imposes enough structure on it that the decision can be examined afterwards by somebody who was not in the room.

What sets the size of the deviation?

The mandate bandThe range of weights a written mandate permits for a class, agreed in advance between the holder and the manager. It is a permission, not a target, and the policy weight is a separate decision taken inside it., and nothing else. In the Anantara mandate equity may run from 50 to 70 per cent against a policy weight of 60 per cent, so a deviation may be ten points in either direction and not one point more. On a Rs 500 crore portfolio ten points is Rs 50 crore, the largest cheque any tactical view in this mandate can ever write. The size of a position is set by a document written before the view existed, not by how convincing the view is, and that ordering is the entire source of the constraint's authority.

The interesting property of a band is that it was agreed in ignorance. When Rukmini Deshpande's committee wrote 50 to 70, nobody in the room had any particular opinion about what equity would do. Agreement reached in ignorance is what makes the number credible later. A limit set by people who already know which way they want to lean is not a limit, it is a rationalisation with a number on it.

Which is why a proposal to widen the band in order to fit a position is worth naming as the specific event it is. The committee has not adjusted a technicality. The committee has discovered that the constraint binds. A binding constraint is the only kind that is doing anything at all, and the committee has responded by removing it. A band widened to accommodate the position somebody wants to take has stopped being a constraint and become a formality, and the honest description of what happened is that the mandate changed, not that the position was permitted.

Why is raise equity to 70 per cent not a complete instruction?

Because ten points of the portfolio have to come from somewhere, and there are two somewheres. The Anantara mandate holds fixed income and cash alongside equity, so an instruction to move Rs 50 crore into equity is only half an instruction until it names which sleeve writes the cheque. The same tactical call funded two different ways produces two different portfolios with two different expected returns and two different volatilities, so an instruction that names only the destination has left half the decision to whoever happens to execute it.

The holder's own assumptions are equity at 12.0 per cent expected return and 18.0 per cent volatility, fixed income at 7.5 and 5.0, cash at 6.0 and 0.5, with a correlation of 0.20 between equity and fixed income and cash treated as uncorrelated with either. Work both branches on those. The endowment chose the figures for its own planning, and planning inputs are not forecasts, not market expectations and not anybody's published estimates. A different set of assumptions gives a different answer to everything below.

Funded from cash, the mix becomes 70 per cent equity, 30 per cent fixed income and nothing in cash. Expected return is 0.70 times 12.0 plus 0.30 times 7.5. Add 8.40 and 2.25 and the answer is 10.65 per cent. Funded from fixed income instead, the mix becomes 70, 20 and 10. Expected return is 8.40 plus 1.50 plus 0.60, or 10.50 per cent. The funding classThe sleeve that pays for a tactical increase in another sleeve. Naming it is part of the instruction, because the portfolio that results depends on which one is sold. is worth 0.15 percentage points of expected return before anybody has been right or wrong about anything.

The volatility half takes more care because it does not add up in a straight line. Funded from cash, the variance is 0.70 squared times 324, plus 0.30 squared times 25, plus twice 0.70 times 0.30 times 18 times 5 times 0.20. The three terms are 158.76, 2.25 and 7.56, adding to 168.57, and the square root of 168.57 is 12.98 per cent. Funded from fixed income, the variance is 158.76 plus 0.20 squared times 25, plus 0.10 squared times 0.25, plus twice 0.70 times 0.20 times 18 times 5 times 0.20. The four terms are 158.76, 1.00, 0.0025 and 5.04, adding to 164.8025, and the square root of 164.8025 is 12.84 per cent.

Set out in full the subtraction is 12.983 less 12.838, or 0.146 points. Name the two figures simply as 12.98 and 12.84 and their difference reads as 0.14. Both are the same quantity at two roundings, so the rounding has to be named every time the figure is used. Naming the rounding sounds fussy for a seventh of a point until a committee minute records one and a performance report records the other.

What can the corridor actually do to the portfolio?

Take the whole span, funded from cash so there is one variable rather than two. At the floor the mix is 50, 30 and 20. Expected return is 6.00 plus 2.25 plus 1.20, or 9.45 per cent. Variance is 0.50 squared times 324, plus 2.25, plus 0.20 squared times 0.25, plus twice 0.50 times 0.30 times 18 times 5 times 0.20. The four terms are 81.00, 2.25, 0.01 and 5.40, adding to 88.66, and the square root of 88.66 is 9.42 per cent. At the policy point the pair is 10.05 and 11.20, computed under the policy mix and used here rather than rebuilt. At the ceiling the pair is 10.65 and 12.98.

Put on one scale, the two spans give the shape worth carrying away. Expected return travels from 9.45 to 10.65, a range of 1.20 points. Volatility travels from 9.42 to 12.98, a range of 3.56 points. The second divided by the first is 2.97. Across exactly the range this mandate permits, risk moves almost precisely three times as far as expected return, and that ratio is a property of the assumption set rather than a comment on anybody's judgement.

The three to one ratio carries a small correction worth stating plainly. Volatility across this corridor is very nearly a straight line. Each point of equity weight adds 0.1778 points of volatility at the floor and 0.1788 points at the ceiling, so the curvature is in the fourth decimal place and invisible at any scale a committee would look at. The consequence is that the policy point sits at the midpoint of both spans, not just of the return span: 11.20 per cent is 1.781 points above the floor's 9.42 and 1.786 points below the ceiling's 12.98. The upper and lower halves of the corridor are almost identical, so the asymmetry in this mandate is not between them but between the two axes. The same twenty points of permission buys three times as much movement in risk as in expected return.

What has to be written down before the position is taken?

Six things, and all six before the trade rather than after it. The view, in a sentence somebody else could disagree with. The evidence that would show the view to be wrong. The size, in points and in rupees. The funding class. The horizonThe period over which the view is expected to resolve. Stating it converts an open-ended position into one that has a date by which somebody has to look at it again.. And the exit condition. Six lines written before the trade are pre-commitmentsStatements written before a decision is acted on, so that the standard the decision will later be judged against cannot be adjusted once the outcome is known. rather than notes, and the difference is entirely in the timing. Written before, they are a standard the decision can be held to. Written after, they are a description of what happened.

The second line is the one that carries the weight, and it is the one most often blank. Falsifying evidenceA statement, written in advance, of what would have to be observed for a view to count as wrong. Without it a view can absorb any outcome and still appear to have been correct. is a statement of what would have to happen for the view to count as mistaken. Write it and the position becomes reviewable by anybody. Leave it blank and the position is unreviewable by everybody, including the person who took it. Twelve months later any outcome at all can be read as partial confirmation. None of that is a criticism of anybody's honesty. Memory works that way once the answer is already known.

Run the list against a position taken at the ceiling of the Anantara corridor. The view sits in a sentence. The falsifier sits in a second sentence. The size is ten points, or Rs 50 crore. The funding class is named, and naming it fixes the destination at 10.65 per cent and 12.98 per cent rather than 10.50 and 12.84. The horizon is stated. The exit condition is stated. The arithmetic tells a committee exactly how the position changes the portfolio and tells it nothing whatever about whether the view behind the position is any good.

How is a tactical position closed, and why is that the hard part?

An exit conditionThe written statement of what has to be true for a temporary position to be unwound. It can be a date, an observation or a level, but it has to be checkable by somebody who was not in the room when the position was taken. takes one of three forms and is worth choosing between explicitly. One form is a date, in which case the position closes when the horizon arrives whatever else is true. A second is evidence, in which case it closes when the falsifier arrives or when the view is confirmed and has been captured. The third is a level, in which case it closes when whatever the view was about reaches a stated point. All three are checkable by a person who was not in the room. An exit condition that says the position closes when it is no longer attractive is checkable by nobody. The sentence contains no observation anybody could disagree with.

Now the part that makes closing hard. Taking a position is exciting and closing one is admitting something: either that the view was wrong, or that it was right and the reason for holding the position has gone. Neither admission is comfortable, and neither has a deadline unless somebody wrote one. So positions stay on. A tactical position that stays on long enough is no longer temporary in any sense the document would recognise, so it stops being tactical by the second of the three conditions.

A tactical position that is never closed has quietly become the policy while the document still says something else. Preventing exactly that is why the whole discipline exists. Notice what makes it so hard to catch. Nothing is breached. Equity at 70 per cent is permitted, every report is clean, every limit check passes. The only evidence that anything has gone wrong is a comparison between the written policy weight and where the portfolio has actually sat for three years, and almost nobody runs that comparison because no system produces it automatically.

What does a tactical position cost even when the view is right?

Two things, and the first is easier to name than to measure. It trades. Moving ten points of a Rs 500 crore portfolio is Rs 50 crore of buying and, when the position closes, Rs 50 crore of selling. Trading carries a cost, and no return or volatility figure above shows it. Every figure above is computed from expected returns rather than from realised cash. The cost exists, and the record for this mandate carries no trading cost figure, so no number is attached to it.

The record does carry one adjacent figure, and it is worth being careful about what it can and cannot say. Portfolio turnover for the stated twelve month period was 34 per cent, meaning about a third of the portfolio was replaced during the year. The record does not split that figure. The record does not say how much of it came from tactical positions, how much from ordinary rebalancing back towards policy weights, and how much from changing what is held inside a sleeve without changing the sleeve. So the honest picture is one bar with a total and three parts that are simply not supplied.

The second cost is attention. A tactical position has to be monitored, reviewed and eventually argued about, and the people doing that are the same people who would otherwise be looking at whether the policy mix still matches what the endowment needs. Attention is a real cost with no line in any report, and it falls hardest on small committees. Most committees are small.

How does a committee actually run this in a room?

Rukmini Deshpande's committee meets, and the tactical part of the agenda runs to one printed sheet. The sheet carries four things. Where the portfolio actually sits against the policy weight, in points and in rupees. Which of those points are there because somebody decided and which are there because prices moved. For each decided position, the six pre-commitments as they were written, not as they are now remembered. And for each one, whether the exit condition has been met.

The same discipline scales all the way down, and scaling down is the test of whether it is a real idea or an institutional habit. A household that has settled on putting a fixed share of each month's income into a long term savings plan and the rest into a sweep account has a policy mix. Suppose somebody in that household holds back two months of contributions for an expected expense. The departure is deliberate, temporary and bounded, and it needs exactly the same three sentences: the expense, the amount held back, and the date the normal contribution resumes. The third sentence is the one nobody writes, and six months later the reduced contribution has become the household's actual savings rate.

A street vendor does the same arithmetic without writing anything at all. Extra stock is bought before a festival week, funded by holding back the usual payment to the wholesaler. Holding back the payment is deliberate, temporary and bounded by how much credit the wholesaler allows, a mandate corridor with a person's name on it. The festival ends on a fixed date, so the vendor knows exactly when the position closes. The discipline is not an institutional invention. Institutions add the writing down. Institutions forget in a way one person watching one stall does not.

What does none of this arithmetic settle?

The direction. A position of a given size, funded a given way, has an exact effect on expected return and volatility once the holder's assumptions are fixed. Whether equity at 70 per cent is a better place to be than equity at 50 per cent depends on a view about the future, and no arithmetic of any kind reaches a view about the future.

There is a second thing the arithmetic does not settle, and it sits closer to the surface than people expect. The corridor's shape, that ratio of about three to one between risk and expected return, is a property of the assumption set the endowment chose. Change equity's assumed volatility from 18.0 per cent to something else and the ratio moves. The numbers here are consequences of assumptions rather than facts about markets, and treating them otherwise would quietly turn an invented planning input into a claim about the world.