Rebalancing vs Tactical Allocation: Same Trade, Not Same Act

Drift is the change in a portfolio's weights caused by prices moving while nobody trades, and nobody decided it. Rebalancing is the trade that puts the policy weights back and expresses no view about what comes next. A tactical move of the same direction and the same size expresses one. The instruction can be identical and the two are still not the same act.

Three things move a portfolio's weights. A tactical move is somebody deciding to depart from the written mix. Rebalancing is somebody deciding to put the written mix back. The third is nothing happening at all, and the weights move anyway. The other two cannot be told apart until the third can be seen, so the third comes first.

The working example is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore run for an invented charitable endowment, chaired by Rukmini Deshpande and managed by Faiz Ahmad Ansari. Its stated 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, and it permits equity between 50 and 70 per cent.

Two things settled earlier are carried rather than rebuilt. A policy weightThe weight a portfolio is designed to hold in a class. It is a stated intention written into the mandate, not a measurement of what is held today. and an actual weightThe weight a portfolio is holding right now: a sleeve's current value divided by the current total. It is a measurement, and it changes whenever any value changes. are different quantities answering different questions. And a tactical position is a departure from the first, taken inside limits granted in advance and covered separately, under strategic and tactical asset allocation.

What is drift, and who decided it?

Picture a household with Rs 3,00,000/- in a deposit and Rs 2,00,000/- in a chit that has been running well. A year passes and nobody adds, withdraws or moves a rupee. The deposit is now Rs 3,18,000/- and the chit Rs 2,40,000/-, so 60 to 40 has become 57 to 43. Ask that household when it decided to hold more of the chit and it will say, correctly, that it decided nothing.

DriftThe change in a portfolio's weights produced by prices moving, with no trade involved. It is arithmetic on ratios rather than a decision anybody takes. is that, at portfolio scale. A weight is a ratio: a sleeve's value over the total. Prices move, each sleeve changes by a different proportion, and all three ratios differ by the morning. Drift is the only change a portfolio undergoes that requires no decision from anybody, and it happens continuously whether or not anyone is watching. There was no meeting, so there is no minute to read.

Run it and see. The holder's own stated assumptions are equity 12.0 per cent, fixed income 7.5 and cash 6.0, chosen by the endowment for its own planning rather than forecast by anybody, and used here so that every class delivers precisely what was expected of it.

Equity at 12.0 per cent becomes Rs 336.00 crore, fixed income at 7.5 becomes Rs 161.25 crore and cash at 6.0 becomes Rs 53.00 crore, so the portfolio is Rs 550.25 crore. Now the weights: 336.00 over 550.25 is 61.06 per cent, 161.25 over 550.25 is 29.30 and 53.00 over 550.25 is 9.63, summing to 100.00.

Every class delivered what the holder assumed. Nobody took a view, nobody was wrong, nobody traded, and the equity weight is 1.06 points above policy. Drift is not a consequence of anybody being mistaken, so no person who caused it can be found.

The cash sleeve is where the arithmetic most often surprises people. Cash did not shrink: it grew from Rs 50.00 crore to Rs 53.00 crore, a return of 6.0 per cent, and its weight still fell from 10.00 to 9.63. Nothing bad happened to cash. Something merely happened faster somewhere else.

Seen as a ratio, the rule for which way a weight moves falls out immediately: a sleeve's weight rises if and only if that sleeve grew faster than the whole portfolio grew, and falls otherwise. The portfolio grew 10.05 per cent. Equity at 12.0 beat that and gained share; fixed income at 7.5 and cash at 6.0 fell short and lost share, both while making money. There is no exception.

So the weights can finish a day different in three ways, and only two have a person attached.

Where does an untouched portfolio end up?

Suppose nobody ever puts the weights back, not out of a view but because nobody does it. Most readers expect the portfolio to wander: equity ahead sometimes, fixed income ahead sometimes, evening out over a long enough run. The expectation of wandering is wrong, and the arithmetic says why.

The class that grows fastest gains share. Having gained share it has more of the portfolio riding on it, so the next period's growth of the whole is pulled further towards its own return, and it gains share again. Drift is self-reinforcing in one direction, so an untouched portfolio does not wander at random. The portfolio migrates steadily towards whichever class grows fastest, and in most mandates that class is also the most volatile.

Run the same assumptions forward twenty periods without a trade and the equity weight goes 60.00, 61.06, 62.12, 63.15, reaching 65.19 at period five, 70.03 at period ten and 78.39 at period twenty, with fixed income at 17.26 per cent and cash at 4.34. Nobody sold a rupee of fixed income and nobody bought a rupee of equity.

The household version is uncomfortably familiar. A couple buys a flat and keeps a savings account. They never sell any of the flat and it grows faster, so twenty years on the flat is most of what they have. Their exposure to one street in one city was arrived at by nobody, on no date, in no meeting.

What exactly is rebalancing, and what is it putting back?

Now the first of the two deliberate acts, defined in full before either is contrasted. RebalancingA trade that returns a portfolio's actual weights to the policy weights written in its mandate. Its whole content is restoration; it asserts nothing about what prices will do next. is a trade that returns the actual weights to the policy weights. The definition ends there. Everything else attached to rebalancing is a method for doing it or a consequence of having done it.

Notice what is missing. There is no expectation in it: no view about any class, no horizon, no target. A restoration trade asserts nothing whatever about what prices will do next: its entire content is that the document still holds and the portfolio has stopped matching it. In four words, the mandate still stands.

Whether a restoration is triggered by a date or by a weight crossing a band edge, and how far back it goes, are questions about how one is run, covered under rebalancing methods in the risk monitoring sequence. Writing a rebalancing policy into a document is covered by the mandate sequence.

The list of what a restoration leaves alone is longer than the list of what it changes, and reading it is the fastest way to see that no view is expressed. The policy weights are untouched: a restoration obeys them, it does not revise them.

What exactly is a tactical move, and what does it depart from?

Now the second side, defined as fully. A tactical deviationA deliberate departure from a policy weight, taken because somebody expects something, within limits the mandate already grants, and recorded before the trade is placed. is a deliberate departure from a policy weight, taken because somebody expects something. A tactical trade sits inside the same band, is sized inside the same limits and is placed through the same broker as any other trade. Its label comes from the reason behind it, not from where it goes.

Being a departure, it carries obligations a restoration does not. Somebody has to say what they expect, over what period, how large the departure will be and what would end it. A tactical move is the only one of the three weight changes that requires somebody to have an expectation, and the only one argued for in advance.

How can two different acts be the same instruction?

Here is what makes the distinction hard rather than obvious. Faiz Ahmad Ansari sells Rs 5.85 crore of equity on a Tuesday because the weight had drifted to 61.06 per cent against a policy 60.00. On the same Tuesday, in another version of the same portfolio, he sells Rs 5.85 crore because the committee expects equity to fall. The two contract notes read the same: same day, same amount, same sleeve, same broker, same price.

A trade is only an amount and a direction, and both acts produce the identical amount and the identical direction, so nothing inside the trade distinguishes a restoration from a departure. The ticket has no field for the reason, and neither does the custodian's statement or the performance record. The reason lives in one place only, the minute written before the order. Unwritten, it survives nowhere.

The consequence is the part with teeth. The evidence that would settle the question is created before the trade or not at all, so a portfolio that does not record the reason cannot recover the distinction afterwards. Memory reconstructs reasons to fit outcomes, so nobody can look at Rs 5.85 crore of equity leaving six years ago and say which act it was: not the custodian, not the auditor, and not the manager honestly.

Set the two acts against the same criteria and they separate on every one except the last. The last row is the only one observable without the record.

What does the direction of a restoration trade indicate?

The direction of a restoration trade explains more real behaviour than anything else about it. Restoring a weight means bringing the sleeve that grew back down and topping up the sleeves that lagged, and there is no other way to do it. A restoration always sells the class that rose and buys the class that fell, for arithmetic reasons unconnected to anybody's opinion of either.

Now sit with how that feels. Equity has just done well and the restoration sells equity; fixed income has done less well and the restoration buys it. To anybody without the record that is indistinguishable from a manager who thinks equity expensive and fixed income cheap. The trade looks exactly like a view, and it is not one. When the move runs on for three periods, the same sale looks wrong three times.

Take three constructed periods in which equity keeps outgrowing the other sleeves at the assumed rates, restoring at the end of each. Period one ends with equity at Rs 336.00 crore against a policy need of Rs 330.15 crore, selling Rs 5.85 crore. Period two, from Rs 550.25 crore, ends at Rs 369.77 crore against Rs 363.33 crore, selling Rs 6.44 crore. Period three, from Rs 605.55 crore, ends at Rs 406.93 crore against Rs 399.84 crore, selling Rs 7.08 crore.

Looking wrong for the whole of a sustained move is the most common reason a restoration process gets abandoned, and it is abandoned exactly where the drift has grown largest. Nobody stops rebalancing in a flat period, because there is nothing to feel bad about. Committees stop after the third year of selling the class that keeps going up, and that is exactly when leaving the portfolio alone has taken it furthest from the document.

What does one full restoration actually look like?

The drifted portfolio set out above, put back. The portfolio stands at Rs 550.25 crore with equity Rs 336.00 crore, fixed income Rs 161.25 crore and cash Rs 53.00 crore. The policy weights are unchanged at 60, 30 and 10, so the question is what they are worth on the new total: Rs 330.15 crore, Rs 165.075 crore and Rs 55.025 crore.

The trade column is the difference between the two before it, and the check is that sells equal buys. A restoration moves money between sleeves and never changes the total, so the sells and the buys tie by construction rather than by luck. Unrounded the tie is exact: Rs 3.825 crore plus Rs 2.025 crore is Rs 5.85 crore. Round each buy first and the extra Rs 1 lakh is rounding, not a missing rupee.

Note the direction once more: the sale is in equity at 12.0 per cent, and the buys are in the two classes that returned 7.5 and 6.0.

What does a restoration cost, and where does that cost appear?

Every restoration is a trade, and every trade takes something out: brokerage, the spread between what a buyer pays and a seller gets, the market moving while a large order is worked, and what the settlement chain charges. Where that cost shows up in the numbers is the question, and it does not. The cost is certain to exist and is reported nowhere in the mandate's record, so it can be named but never measured from the statements.

The record does carry one number that bears on it, turnoverThe share of a portfolio replaced by trading over a stated period. It counts how much was traded, and says nothing at all about why any of it was traded.. The Anantara portfolio ran 34 per cent over the stated twelve month period. On Rs 500 crore that is Rs 170 crore of trading. Turnover counts rupees traded and says nothing at all about why any of them were traded: the same 34 per cent could be all restoration, all view, or any mixture.

The two figures do not sit on the same base, so the comparison needs care. The restoration is Rs 5.85 crore against a drifted total of Rs 550.25 crore, or 1.06 per cent. The 34 per cent turnover is measured against Rs 500 crore instead. As percentages that is about one part in thirty two, and as rupee amounts about one part in twenty nine. Neither is wrong, and they differ only because the bases differ.

Cost behaves in one more way that surprises people. Restoring more often is not free. Less drift accumulates between restorations, so each one is smaller, but there are more of them and the cost attached to each does not vanish because the trade is small. No single frequency suits every mandate, and the trade-off is always the same: drift left standing against cost paid more often.

Can a portfolio simply never rebalance?

Never rebalancing is tempting to treat as the neutral option. Under a mandate with a band it is not available. The equity band of 50 to 70 per cent is not a preference the committee can decline to act on: it is a constraint, and the moment drift carries the weight across it a trade is forced whether anybody wanted one or not.

At the mandate's stated size that band puts equity between Rs 250 crore and Rs 350 crore. Reading it that way holds the total still while one weight moves, a device rather than something a live portfolio does. The drift arithmetic above lets the total move instead, and two readings of one band need not agree on a rupee amount.

So the choice is narrower than it looks. A committee can restore the weights on a basis of its own choosing, or wait until the constraint restores them on a date chosen by prices rather than by anybody in the room. Both are trades, and no version keeps the drifted weights indefinitely.

How can it be established afterwards which one a trade was?

The answer is the minuteThe written record a committee makes of a decision at the time it is taken. It holds the reason, which no contract note or custodian statement ever carries., and there is no second method. If it records only that the weights had moved, it was a restoration. If it records a view, a horizon and an exit, it was a departure. If it records the trade alone, the distinction is gone permanently.

A minute that records only what was traded has destroyed a distinction no later analysis can rebuild. The problem is one of documentation rather than analysis. Ownership of the fix changes with it. No skill with returns data recovers a reason never written, so the fix sits with whoever writes the minute, on the day, before the order goes out.

How does a practitioner use this distinction day to day?

Faiz Ahmad Ansari uses it as a discipline on his own writing. Before any order goes out, the line in the minute has to say which act is being taken, and the two forms of words are deliberately different. A restoration says the weights had moved and were put back. A departure says what is expected, over what period, how far and what closes it. The two forms look nothing alike, so nobody has to remember the rule.

Rukmini Deshpande, chairing the committee, uses it as a filter on what she approves. When a paper proposes moving equity she asks one question first: is this putting the document back, or departing from it? If the first, the discussion is about timing and cost. If the second, the paper needs the four lines a view requires. Papers that cannot answer have not finished being written.

An analyst reviewing a manager from outside tests the record rather than the returns. Does the turnover figure carry any account of what the trading was for, and do the minutes separate restorations from departures at the time. A manager whose record cannot separate the two has not necessarily done anything wrong, and has made it impossible for anybody to tell. The finding is about the record, and it stops there.

A household can use the same discipline without any of the machinery. When part of a holding that has grown large is sold, one line recording why is enough. Either the holding had become a bigger share of the household's savings than intended, and the sale is a restoration, or the expectation is that the price will fall, and the sale is a view that needs a date to look again. A year later that line is the difference between knowing what was done and telling a story that flatters the outcome.

When does the distinction stop mattering?

Four conditions leave the two acts placing the same trade, or no trade at all. On a day carrying any of them, which act a trade was decides nothing.

The first is a drift too small to act on. Let equity return 7.5 per cent for the period instead of 12.0, matching fixed income exactly. Equity ends at Rs 322.50 crore on a total of Rs 536.75 crore, a weight of 60.08 against a policy value of Rs 322.05 crore, so the restoration sells Rs 0.45 crore. Rs 0.45 crore is 0.08 per cent of the portfolio, and no view is sized at eight hundredths of a point either.

The second is a tolerance band wide enough that the rule almost never fires. On the untouched run above, a rule acting only at the 50 and 70 per cent mandate edges places nothing until period ten. For the nine periods before that the rule is silent, so any trade in the record came from somebody's view and the question answers itself.

The third is a view pointing where the rule already points. Equity sits at 61.06 per cent and the committee, on its own reasoning, expects a fall and takes equity back to 60.0. The order is Rs 5.85 crore, the same amount and direction the restoration would have placed, and the portfolio afterwards is identical to the rupee. Where the view and the rule agree, one order carries two reasons, and only the minute written before it went out says which one was acted on.

The fourth is cost large enough to swallow either. A trade of Rs 0.45 crore carries the same brokerage, spread and market impact whichever reason sits behind it, and where those exceed the value of the move, neither act is worth doing. The mandate's record carries no cost figure, so the condition has no threshold attached to it.

None of the four lasts, and none says when it has stopped. One class runs ahead of the others with nothing decided, and the drift that was 0.08 points last period is 1.06 this one. A band gets rewritten when the endowment reviews its objectives. A view that agreed with the rule stops agreeing the week the price moves. The condition goes quietly, the distinction is load bearing again, and what helps on that day is a record already kept in the way described above.

The Anantara Multi-Asset Portfolio, the charitable endowment that holds it, Rukmini Deshpande and Faiz Ahmad Ansari are invented.
Educational material. Not advice on any investment, tax, budget or market position.