Rebalancing: the trigger is the decision

The trigger is the decision

A calendar rebalance fires on a date and ignores how far the portfolio has drifted. A band fires on drift and ignores the date. Each is a rule for spending money now to move a portfolio nearer its target, and the two spend it at different moments.

What rebalancing is, why a portfolio drifts away from the split you chose, and how the trade itself is placed are set out in the rebalancing article. This one assumes all of that and takes up the part it leaves open: given that a correction is going to happen, what makes today the day it happens?

Strip both rules down and they are watching the same single number — the gap between the split you hold and the split you chose. A 60/40 target sitting at 64/36 has a gap of four points. Everything below is about when a rule is allowed to look at that number and what it is allowed to do about what it sees.

A calendar rule samples the gap at fixed times and acts on whatever it finds. A band rule watches the gap continuously and acts only once it exceeds a stated size. The first is a rule about when you look. The second is a rule about how big the gap is. They are not two flavours of the same discipline; they are answers to different questions, and each one leaves the other question unanswered.

Here is why that choice has stakes rather than being a matter of taste. Every correction is a trade, and every trade costs something immediate and certain — charges on the way through, possibly a load, possibly a tax bill on gains that were sitting untaxed. What it buys is a portfolio whose risk sits closer to the one you picked, which is real and is not a rupee amount you can put on the same line. Certain cost, uncertain benefit is the shape of every rebalancing decision, and it is why the answer is never to hold the portfolio perfectly on target.

What each rule cannot see

Take the calendar first. It is blind to everything that happens between two dates, and that single blindness produces two failures that are rarely connected to each other.

The first is the trade that did not need to happen. The date arrives, the split has moved by half a point, and the rule fires anyway. Charges are paid, a gain may be realised, and what has been bought is a correction nobody would have paid for if asked directly. The rule was built to remove the judgement about whether to act, and removing that judgement is precisely what it did.

The second is the move the rule never saw. Suppose the review runs each January. A market that rises hard in February and gives it all back by November leaves no trace at all in the January-to-January record — the gap opened and closed inside the sampling interval, and the rule never had an opportunity to act on it. The mirror image is worse: a large move that opens in February then sits uncorrected for eleven months, at a risk level nobody chose, while the rule waits for a date. The sampling interval is the design, and a year is a convenient unit rather than a fact about any portfolio.

Now the band. It is blind to time. It does not know whether the gap opened this morning or has been open for two years, and it has no view on whether it is about to close by itself. That produces its own pair of failures.

The first is frequency. A band is a threshold on a moving quantity, so how often it is crossed depends on how much the quantity moves. Through a quiet stretch it may not fire for years; through a turbulent one it can fire several times in a few months, and each firing is a full set of costs. The rule did not get worse. The market underneath it changed, and a threshold has no way of noticing that its own trigger rate just went up.

The second is the round trip. A band fires when the gap crosses out, and if the market reverses shortly afterwards it fires again to put back most of what it just moved. Two sets of costs, and a portfolio sitting roughly where it would have been had neither trade happened. Nothing has malfunctioned. The rule did exactly what it says, twice, on a series that did not stay where it went.

How wide a band actually is

A band is usually stated in weight points — act if a 60/40 target leaves the range 55/45 to 65/35. That reads as a tight tolerance. Measured in how far the market has to move, it is not, and the gap between those two readings is where most of the surprise in this topic lives.

Work it out on a target of 60/40 and a band of ±5 points, both chosen to make the arithmetic legible rather than because either is standard, and hold the debt side flat so the moving parts stay visible. Start with ₹60 in equity and ₹40 in debt per ₹100. Let the equity side rise by x. Its weight is 60(1+x) ÷ (100+60x); setting that equal to 0.65 gives 60 + 60x = 65 + 39x, so x = 5/21, about 23.8%.

Now the other side. Let equity fall by y. Setting 60(1−y) ÷ (100−60y) equal to 0.55 gives 60 − 60y = 55 − 33y, so y = 5/27, about 18.5%. Both figures assume the debt side stayed exactly where it was, which no real move obliges it to do — they are the arithmetic of a stated assumption, not a measurement of what a live portfolio needs.

Those two numbers are not the same, and the band that produced them was symmetric. A 5-point tolerance means roughly a 24% rise on one side and roughly a 19% fall on the other. The reason is that a weight is a ratio and the denominator moves with it. When equity rises the total rises with it, so each further rupee of gain shifts the weight less than the rupee before it; when equity falls the total shrinks, so each further rupee of loss shifts the weight more. Sensitivity grows as the sleeve shrinks.

State the consequence plainly and then stop. A band written symmetrically in weight points is an asymmetric rule in return terms, and it is the tighter trigger on the way down. That is arithmetic. Whether a rule that reacts sooner to falls than to rises suits a particular portfolio is a judgement about that portfolio and its holder, and this article is not in a position to make it.

A stated band hides a second thing: the same number means different things to sleeves of different sizes. Five points around a 60% sleeve is a tolerance of about 8% of the sleeve's own size. Five points around a 10% sleeve is 50% of it — that holding can halve, or grow by half, before anything fires. Writing the band relative to the sleeve reverses the distortion rather than removing it: a ±25% relative band gives the 60% sleeve a 45–75 range and the 10% sleeve a 7.5–12.5 range. One rule, two tolerances, and which distortion applies depends only on whether the band was written in points or in proportions.

Neither form is the correct one. They are different statements about which sleeves the rule is allowed to ignore, and a rule inherited without noticing which form it is in has made that statement by accident. What the sleeves are supposed to be doing for the portfolio in the first place is the subject of diversification.

Why a wider band is not laziness

The cost of correcting is not one number, and it does not scale the way the word ‘cost’ suggests. Three separate things are being paid, and each is charged on a different base.

• Transaction charges — brokerage where the units are listed, the exchange and statutory levies that accompany a trade, and the goods and services tax on those — scale with the size of the trade. This is the only one of the three that behaves the way intuition expects.
• An exit load, where it applies, is charged on the redemption value of units still inside their load window. It has nothing to do with the rest of the holding and nothing to do with how much the holding has gained.
• Tax is charged on the gain embedded in the units sold, which has nothing to do with the size of the trade at all.

That third base is the one that breaks the intuition, so put numbers on it with the figures chosen for contrast. Two switches of ₹1 lakh each. In the first, the units being sold cost ₹95,000, so the gain is ₹5,000. In the second, units bought long ago cost ₹40,000, so the gain is ₹60,000. Identical trade, twelve times the taxable gain. Trade size is the wrong base for estimating what a rebalance costs; the embedded gain is, and two portfolios running the same rule on the same day can face wildly different bills.

The rates attach to that gain rather than to the trade. Gains on equity-oriented units held at least 12 months are taxed at 12.5% above an annual threshold of ₹1.25 lakh; below that holding period the rate is 20%. Units of a fund holding more than 65% in debt and money-market instruments, acquired on or after 1 April 2023, are taxed at slab rates, with the gain always treated as short-term. Which units count as sold when a holding was built in instalments decides the embedded gain, and that question, along with the treatment in full, belongs to the capital gains article.

Two structural blocks sit on top of the arithmetic, and both are specific to Indian fund holdings. Units of an equity-linked savings scheme carry a statutory lock-in of three years from the date of each purchase, so a sleeve containing them cannot be trimmed on any trigger's schedule until they come free — and for anyone contributing monthly, that is a rolling series of lock-ins rather than one. Separately, a fund's weights update only once a day, when the single price at which units are struck — the net asset value, or NAV — is published, with the applicable cut-off for non-liquid schemes at 3:00 pm IST. A band watched during the trading day on fund holdings is measuring a price nobody can transact at.

Put the three bases and the two blocks together and the reason a wide band is a considered position rather than a lazy one becomes arithmetic. Correcting a two-point drift on a holding with a large embedded gain can produce a bill out of all proportion to the gap it closes, and no amount of discipline changes that, because the discipline is what generated the bill.

What both rules assume

Both triggers are control rules: they measure a gap against a set point and act to close it. Neither questions the set point. Everything either of them can tell you rests on a short list of assumptions, each imperfect in a known way, and naming them is what separates using a rule from believing it.

• The target is right, and stays right. Both rules treat the chosen split as fixed and the drift as error. That split came from judgements about horizon, income and tolerance for loss — if any of those changed, the rule is now defending a set point that no longer describes the person. A trigger cannot detect this, because it only ever sees weights. Where the set point comes from is asset allocation.
• A weight measures risk. Both rules act on weights, because a weight is the part that can be observed and corrected. What the target was chosen to express is an exposure, and the same weight carries a different exposure once the sleeve's own behaviour changes: a 60% equity sleeve through a quiet stretch and a 60% equity sleeve through a turbulent one are the same number and not the same position. A trigger restores the number. Whether that restores what the number stood for is a question neither rule is built to ask.
• Drift is noise, not information. Selling what has risen and buying what has fallen is structurally a position against continuation. That is a mechanical description of what the rule does, and it is not in dispute. Whether such a position has paid over any particular stretch is an empirical question about specific assets in a specific period, and answering it needs a stated dataset and window. This article makes the first claim and does not make the second.
• The band keeps its meaning. A tolerance in weight points is a threshold on a quantity whose movement depends on how much the underlying assets move. The identical band is a rule that almost never fires through a calm period and a rule that fires repeatedly through a turbulent one, with no edit in between and no notification that anything changed.
• Costs scale with turnover. They do not, for the reasons above — the tax component scales with embedded gain and the load component with the age of specific units. Any comparison of two triggers by turnover alone is measuring one of three costs and calling it the total.
• The weights are observable. For listed holdings that is nearly true during the trading day. For fund units it is not, because the weight only moves once the day's NAV is struck, so a continuously watched band on a fund portfolio is watching a step function that updates after the decision would have been made.
• There is no cashflow. Comparisons of triggers are usually drawn on a portfolio that neither receives contributions nor pays withdrawals. Real portfolios do one or both, and it changes the answer enough to be the next section.

None of that makes a trigger useless. It sets out what a trigger can be held responsible for. A rule that fires on a date cannot be blamed for the move it did not see, and a rule that fires on drift cannot be blamed for firing when there was a great deal of drift. Both did the job they were built to do; the open question is whether that job is the one you wanted done. What the summary statistics describing portfolio risk can and cannot capture about any of this is taken up in the risk measures article.

Two dials, and one that fires without a trade

The two conditions can also be required together: check on a fixed date, and act only if the gap is outside the band. That combination is described in the rebalancing article. What is worth adding here is what it fixes and what it does not.

It fixes the empty rebalance completely, because the date can arrive and nothing need happen. It fixes the round trip partially, because a rule can only fire as often as it looks — a quarterly check cannot fire twice in a month however violently the market moves in between. What it cannot fix is the move that opens and closes between two checks, which is the calendar's blind spot arriving intact. Sampling less often buys transaction economy and pays for it in responsiveness, and no setting of the two dials escapes that exchange.

A third dial gets less attention: where the correction stops. Going all the way back to target uses the largest trade and leaves the portfolio in the middle of the band, furthest from the next crossing. Stopping at the near edge of the band uses a smaller trade and leaves the portfolio sitting where the next ordinary move can push it straight out again. The first pays more per event and expects fewer events; the second pays less per event and expects more. Which total is smaller depends on how the assets actually move over the period in question, which is not knowable in advance.

One row in that table carries no cost of its own. A portfolio receiving contributions can absorb drift by directing new money entirely at the underweight side, correcting the gap without a sale — no load, no realised gain, only the cost of a purchase that was going to happen regardless. The arithmetic decides how far that reaches, on the same chosen 60/40 target and portfolio sizes picked for contrast. On a ₹10 lakh portfolio at 60/40, a ₹20,000 contribution sent entirely to the debt side moves the equity weight by about 1.2 points, so about four months of the whole contribution closes a 5-point gap. On a ₹1 crore portfolio the same ₹20,000 moves it by about 0.12 points, and closing the same gap would take more than three years.

The same logic runs backwards for a portfolio being drawn down: taking the withdrawal from the overweight side rebalances as a side effect of spending, and it costs only whatever the withdrawal was going to cost anyway. That route has the mirror-image limit — it works while the withdrawal is large relative to the gap, and stops when it is not.

Where trigger design goes wrong

Each of these is a rule working exactly as written, which is what makes them worth naming rather than warning about in general.

• The empty rebalance. The date fires on a gap of half a point. Every cost is paid and the risk position barely moves. What removes it structurally is a size condition the date has to clear; a shorter interval does the opposite, since the gap is smaller each time a more frequent rule looks.
• The round trip. The band fires out and fires back within weeks. Two sets of costs, no net change in position. A rule that only looks on scheduled dates cannot do this more often than it looks, which is the whole argument for pairing the two.
• One band, unequal sleeves. A single tolerance in weight points applied across a 60% sleeve and a 5% sleeve is a tight rule on one and almost no rule on the other. That may be exactly what was intended; the failure is inheriting it without noticing the choice was made.
• The band you cannot trade. Watching fund weights during the day and acting on what you see. The unit price that will actually apply is the one struck after the cut-off, so the crossing you reacted to is not necessarily the crossing that gets recorded.
• The locked sleeve. A trigger that assumes any holding can be trimmed will keep signalling on a sleeve holding units inside an ELSS lock-in or an exit load window. The signal is correct and the trade is either impossible or expensive, which is a different problem from drift and does not respond to a wider band.
• Contributions read as drift. New money arriving in one sleeve moves the weights without any market having moved. To a band, a gap created that way and a gap created by prices are the same number — and firing a sale to correct one of them undoes a purchase made days earlier, at full cost.

One more sits upstream of all of them. A portfolio with no written target has no gap to measure, so any trigger applied to it is answering a question that was never asked. The trigger is downstream of the allocation decision and of what the money is for, and it inherits every weakness in both. Tightening a band cannot repair a set point that was never chosen deliberately.

Where this sits in the app

The part of this that can be computed rather than argued about is the drift path: what the split actually did between two dates, and therefore which rules would have fired, and when.

FNOTrader's Mutual Funds app runs on the full AMFI NAV history — around 34 million NAV rows, appended nightly at 22:30 IST — and holds a saved portfolio, so a holding's value can be traced across the same dates on which a rule would have looked. That is enough to reconstruct how far a split moved over a chosen period and how often a stated band would have been crossed in it.

What the app does not do is compute a band, signal a rebalance, or carry a tolerance it treats as correct. That absence is deliberate rather than pending. The tolerance that matters is the one at which the cost of correcting stops being smaller than the cost of tolerating, and both sides of that comparison depend on the holder's own embedded gains, their own load windows, their own contributions and their own reason for choosing the target — none of which lives in a NAV series.

FNOTrader is not a SEBI-registered investment adviser or research analyst. Nothing here is a recommendation about any scheme, allocation, tolerance or rebalancing rule, and nothing here is a forecast — it is the arithmetic of two trigger designs and the costs each of them incurs, so the decision stays with the reader.

Common questions

Continue reading

More in Advanced Investing · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article