Calendar, Threshold and Cash Flow Rebalancing Compared

Rebalancing pulls a drifted portfolio back toward its policy weights. A calendar rule trades on stated dates whatever the weights are. A threshold rule trades only when a weight leaves a stated band. A cash flow rule steers money arriving or leaving toward whatever is underweight, placing no trade of its own. The Anantara Multi-Asset Portfolio permits equity between 50 and 70 per cent, the widest threshold its holder could write.

Three rules, one job, and the whole of the difference between them sits in one word: trigger. Everything else worth knowing about a rebalancing rule, who has to be watching, how much trading it creates, how far the weights are allowed to wander before somebody acts, follows from what fires it. So the three rules are best set side by side on the trigger first, and priced for what each one costs only afterwards.

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 stated shape 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 three sum to Rs 500 crore exactly. The mandate permits equity to sit anywhere between 50 and 70 per cent, and every weight and rupee amount that follows belongs to one stated twelve month period.

What is drift, and why does a portfolio need an answer to it?

A rebalancing rule is necessary rather than merely tidy, and the movement that makes it necessary has a name. Nobody moves the weights. The market moves them. The parts of a portfolio grow at different rates, and a share of a growing total is arithmetic rather than a decision. DriftThe gradual movement of a portfolio's actual shares away from the shares its holder decided on, caused by the parts growing at different rates rather than by anyone trading. is what that movement is called, and it happens to a portfolio nobody has touched.

Work it once with numbers so the shape is visible. Take the Anantara portfolio at its policy weightsThe share of a portfolio the holder decided each asset class should carry. It is a decision taken in advance, not an observation of what is there today. and give the three parts illustrative growth for a year: equity up 20 per cent, fixed income up 5 per cent, cash up 6 per cent. Equity goes from Rs 300 crore to Rs 360 crore. Fixed income goes from Rs 150 crore to Rs 157.50 crore. Cash goes from Rs 50 crore to Rs 53 crore. The total is now Rs 570.50 crore, and the actual weightsThe shares each asset class carries today, read off the current values. They move whenever prices move, whether or not anyone has traded. read 63.10 per cent, 27.61 per cent and 9.29 per cent.

Now say out loud what happened. Nobody traded. Nobody proposed anything. Nobody took a decision to hold 3.10 more points of equity than the holder had asked for, and yet the portfolio holds them. A portfolio left alone drifts toward whatever grew most. Its risk therefore rises exactly after the fastest growing holding has already grown, and no decision was ever taken to accept that risk. That is the sentence a rebalancing rule exists to answer.

The everyday version sits in most households. A household puts a fixed sum every month into a deposit and a fixed sum into an equity plan, in equal shares, and never changes the instruction. Five years later the equity side has grown faster, so the household is now running a savings pot that leans on the market far harder than the shares it originally set. Nobody moved anything. The shape moved on its own, and the first time anyone notices is usually the month the market falls.

What is calendar rebalancing, and what is its one real weakness?

A calendar rule states dates and nothing else. On the last business day of every quarter, or every month, or once a year, the portfolio is returned to its policy weights, whatever the weights happen to be on that morning. There is no test, no threshold and no judgement in the trigger. The date arrives and the work is done.

A date is not a matter of opinion, so the strength of a calendar rule is that its trigger cannot be argued with or postponed. This sounds like a small thing and it is not. Every discretionary trigger invites a conversation about whether now is the moment, and that conversation is where drift gets left alone for another quarter. A date closes the conversation before it opens. A stated date is the only trigger that survives a room in which everybody has a view.

The weakness comes out of exactly the same property. A rule that acts only on stated dates does two unhelpful things. The rule trades when nothing needed trading. On the appointed morning the weights may be sitting almost exactly on policy, and the small difference is dealt anyway. And the rule does nothing when something did need doing. A violent move three weeks after the last date waits patiently for the next one. The calendar rule's one strength and its one weakness are the same property seen from two sides: the trigger ignores the portfolio entirely.

What is threshold rebalancing, and what does it demand of somebody?

A threshold rule reverses the question. A threshold rule ignores the date and watches the portfolio. Around each policy weight sits a stated band, and while the actual weight stays inside that band nothing happens at all. The moment a weight leaves its band, the rule fires and the position is brought back, usually to the policy weight itself and sometimes only to the edge of the band.

The threshold rule's trigger is the portfolio's own state rather than the calendar, so it acts when the weights have actually moved and is silent when they have not. That is a genuine improvement on trading a portfolio that did not need trading. But it buys that improvement with a requirement most descriptions of the rule leave out, and the requirement is the whole practical difficulty.

Somebody has to be looking. A band cannot fire on its own. If the weights are computed once a quarter then the rule is a quarterly rule wearing a threshold's clothes. A weight that left the band in week two and came back in week nine will never be seen. A threshold rule is only as continuous as the person or the system computing the weights, and the trigger as written is only the one that can actually be observed. The honest comparison between the calendar and the threshold is not which is cleverer. The comparison is about who is responsible for noticing.

Absolute band or relative band: are they the same rule written twice?

The two forms are not the same rule, and the confusion between them is routine enough to be worth slowing down for. There are two ways to write the width of a band, and they produce different rules from identical looking numbers.

An absolute bandA band whose width is stated in percentage points of the whole portfolio, so the same number of points is allowed to every holding regardless of how large that holding is. is stated in percentage points of the portfolio. Plus or minus 10 points around a 60 per cent policy weight gives 50 to 70 per cent. Plus or minus 10 points around a 10 per cent policy weight gives 0 to 20 per cent. The small holding may vanish entirely or double, and the rule will not have fired.

A relative bandA band whose width is stated as a fraction of the policy weight itself, so a large holding is allowed to move more points than a small one before the rule fires. is stated as a fraction of the policy weight itself. Plus or minus one sixth of 60 is 10 points, so the equity band is again 50 to 70 per cent, and the two forms look identical. Apply the same one sixth to a 10 per cent policy weight and it is plus or minus 1.67 points, giving 8.33 to 11.67 per cent. The same band written the same way is a different rule for a 60 per cent holding and for a 10 per cent one, and a single relative band applied across every bucket is far tighter on small holdings than an absolute band is.

The last row repays a second reading. On the cash bucket the absolute form permits six times the movement the relative form permits, and the two rules were written with what looked like the same generosity. The base problem returns in a different costume: a number means nothing until what it is a fraction of has been stated.

What is cash flow rebalancing, and what can it not do?

Cash flow rebalancingSteering money that is already arriving or already leaving toward or away from particular holdings, so the shares move without any trade being placed for the purpose of moving them. is the third rule and it works from a different direction entirely. Money arrives into most portfolios and money leaves most portfolios. Contributions come in. Coupons and dividends land. Payments go out. Each of those events is a movement of cash that has to be handled anyway, and where it is directed changes the weights.

So the rule is this: send money coming in toward whatever is underweight, and take money going out from whatever is overweightCarrying a larger share of the portfolio than the policy weight says it should. Underweight is the mirror: a smaller share than the policy weight.. The weights move toward policy and not one trade was placed for the purpose of moving them. Cash flow rebalancing changes the shape of the portfolio as a by-product of transactions that had to happen regardless. Riding on transactions that were happening anyway makes it the cheapest of the three, when it works at all.

Two conditions decide whether it works. The flows have to be regular. A rule that can only act when money happens to move is silent in between, however far the weights have drifted. And the flows have to be large enough relative to the drift. Cash flow rebalancing cannot correct a drift bigger than the money that is moving, and past that point it is a partial measure that leaves the rest of the gap open. Put numbers on it: a drift of 5.0 points on the Anantara portfolio is Rs 25 crore of misplaced weight, and a payment of Rs 10 crore directed away from the overweight bucket closes 40.0 per cent of it. The remaining Rs 15 crore sits there until something else acts.

An endowment exists to pay money out, so an endowment is exactly the setting where cash flow rebalancing has something to work with. How much an endowment spends, and when, is settled by its spending policy and not by the rebalancing rule. The mechanism can therefore be stated exactly while the size of the flows it will meet stays open.

What fires the trade under each of the three rules?

Set them against each other on one criterion at a time and the shape of the choice appears. The trigger comes first because everything else falls out of it.

The calendar rule and the threshold rule are different answers to one question: who has the duty to notice. So the table sets out a choice about responsibility rather than a contest between two rules. The calendar rule puts the duty on nobody, which is why it survives an inattentive quarter. The threshold rule puts the duty on a person or a system, so it acts when it matters and fails silently when nobody is looking. In practice the two are frequently written together, with a date for the floor and a band for the exception, and that combination is a third answer rather than a compromise.

How Rebalancing Maintains Portfolio Discipline: what is the discipline actually doing?

Here is what a rebalancing rule does that nothing else in the process does. The rule systematically reduces whatever has grown and adds to whatever has not, and it does so without anybody forming a view about either. No forecast is made. No case is argued. The rule looks at a number, compares it with another number written down in advance, and acts.

Rebalancing is a discipline because it removes the decision, not because reducing what has grown is clever. The distinction matters more than it looks. A great deal of writing on this subject slides from the true statement, that the rule takes the judgement out of the moment, to the untested one, that the rule therefore produces a better result. Whether removing the judgement improves the result is unsettled, for the Anantara portfolio and for any other. The rule keeps the portfolio resembling the one that was designed. Whether the portfolio that was designed was a good one is a different question, settled earlier and not here.

The everyday version is a row of ten shops in one mall. At the start of every quarter the shopkeepers agree to move stock to whichever shelf has emptied fastest, whether or not anybody thinks that shelf deserves more stock. The agreement is not a claim that the emptiest shelf will sell best. The agreement is a way of ensuring the shop does not, quarter by quarter, end up selling only one thing because that one thing happened to move. The discipline is in the rule being made in advance and followed, not in the rule being wise.

What does one trip to the band edge actually cost in trading?

Time to work the rules against the Anantara mandate itself. Hold the portfolio total at Rs 500 crore while the arithmetic runs. The weight is then the only thing moving. In a live portfolio the total moves too, and the moving total changes the rupee figures without changing the shape of the argument.

The equity policy weight is 60.0 per cent of Rs 500 crore, or Rs 300 crore. The mandate permits 50 to 70 per cent. Read that as a band around the policy weight and it is plus or minus 10.0 percentage points in absolute terms, and 10.0 divided by 60.0 is 16.7 per cent in relative terms. The same single permission is a wide rule expressed one way and a moderate one expressed the other. The form of the band therefore has to be stated before anyone argues about the number.

Now price the trip. Equity at 70.0 per cent of Rs 500 crore is Rs 350 crore. Restoring the 60.0 per cent policy weight from there means selling Rs 50 crore, or Rs 50,00,00,000/- in whole rupees. The mirror case is identical in size: equity at 50.0 per cent is Rs 250 crore, and restoring policy means buying Rs 50 crore. Either way that is Rs 50 crore of dealing on a Rs 500 crore portfolio, or 10.0 per cent of the whole thing in one action.

Set that against what actually happened. The Anantara portfolio recorded turnoverA measure of how much of a portfolio was replaced over a period, expressed as a share of the portfolio. It counts trading, not the cost of trading. of 34 per cent over the stated twelve month period, meaning about a third of the portfolio was replaced, which on Rs 500 crore is Rs 170 crore of dealing. One drift out to the band edge and one trade back would be Rs 50 crore, or 29.4 per cent of the year's recorded trading. A single visit to the edge of a wide band would account for roughly three tenths of everything that was traded all year. Scaling one trade against a year of trading says nothing about what caused the recorded turnover, and a turnover figure on its own never can.

What happens when the rebalancing band is set at the mandate limit?

Both documents look sensible, and the sharpest point in the whole comparison hides between them. A mandate limit and a rebalancing band are written in the same units, percentage points of the portfolio, and they answer entirely different questions. The mandate limit answers: how far may this portfolio go before it is no longer the thing the holder agreed to hold? The rebalancing band answers: how far may the weights wander before the portfolio stops resembling the one that was designed?

Set the band at 50 and 70 and watch what happens to the second question. The rule now fires only when the portfolio is standing on the mandate boundary itself. Every drift short of the boundary is simply accepted. A rebalancing band and a mandate limit are two different instruments that happen to be written in the same units, and setting them equal quietly deletes one of them while both documents stay on file. What remains is a breach prevention device with the word discipline written on it.

Compare it with a band written inside the limit. Take plus or minus 3.0 points, purely to draw the contrast and not as a number anyone should adopt. The band sits at 57.0 to 63.0, and the drift accepted without action is 6.0 points, or Rs 30 crore of exposure on Rs 500 crore. At 50 and 70 the drift accepted without action is 20.0 points, or Rs 100 crore of exposure moving with nobody required to do anything about it. Same portfolio, same paperwork, and a difference of Rs 70 crore in how much exposure is allowed to move unattended.

What does rebalancing cost that no return figure shows?

Every rebalancing trade costs something, and the costs are of four kinds. Brokerage and dealing charges. Taxes on transacting, wherever they apply. The spread between what a buyer will pay and what a seller will take. And market impactThe amount an order moves the price against itself simply by being large enough to need more of the market than is standing there at the moment., the price an order pushes against itself once it is big enough to use up what was on offer.

Now look for those four in a return figure. The four costs are not there as a line. A return series is struck either after costs or before them depending on the basis, and either way the costs have been folded into the number rather than displayed beside it. Turnover is the visible trace of an invisible cost, and it is a count of trading rather than a measure of what the trading cost. That is why the amount of trading a rule creates is part of choosing the rule, and it is also why comparing two rules on the returns they produced without knowing the basis is comparing nothing.

How does anybody use this in a meeting, on a Tuesday?

Three numbers, prepared before the room sits down. The actual weight of each bucket today. How far each one sits from its policy weight, in points and in rupees. And what the rule as written would do about the gap, if anything. An investment committee chaired by someone like Rukmini Deshpande does not need a view on the market to run that check. Needing no view is the whole reason for having a rule.

A lender or an analyst reading a mandate from outside asks a narrower question and gets a lot from it: how wide is the band, and how does its width compare with the mandate limit? A band at the limit tells them the portfolio's shape is effectively unmanaged between boundaries. A narrow band tells them to expect trading, and to ask what that trading costs. Neither answer is a judgement about the manager; both are facts about how much the weights are permitted to move.

A household can run the same check on a savings pot in five minutes with a sheet of paper. The shares meant to be held go on one list, the shares actually held go on another. If the second list has quietly become a different plan from the first, the drift did that, and no decision did. And the cheapest correction available to a household is almost always the cash flow one: pointing the next few months of contributions at whatever has fallen behind, and letting the money that was moving anyway do the work.

What is rebalancing not?

Rebalancing is not a view. Nothing in any of the three rules contains an opinion about what will happen next, and a rule that starts taking such opinions has become something else that should be described honestly under its own name. Rebalancing is not a cure for a policy weight that was wrong to begin with: a rule that returns a portfolio to the wrong shape returns it faithfully. And rebalancing is not free, as the turnover arithmetic above makes concrete without needing a single cost figure.

Rebalancing is the one routine in the whole process that acts on the portfolio's shape without anyone having to be persuaded. The routine is worth having and worth describing accurately. The three rules set out here are three ways of deciding when that routine runs.