Calendar or Threshold Rebalancing: What Changes?

Joseph has to decide what he is even rebalancing

Joseph has a flat beyond the one he lives in, a portfolio of funds, and some gold. He has read that he should rebalance, and he has hit a problem before reaching the question this page is named after: he cannot rebalance the flat. It cannot be trimmed by a few per cent. It is sold entire or not at all.

That is worth settling first, because it changes what a rebalancing rule is for. Joseph's rule applies to the part of his wealth he can actually trade, and the property sits outside it as a fixed block that the tradeable part has to be sized around. Once he accepts that, the question becomes manageable: within the funds and the gold, what triggers an adjustment — a date, or a distance?

The two rules

A calendar rule sets a date. Every year, or every six months, Joseph looks at the split, compares it with his target, and corrects whatever has drifted. Nothing happens in between, and nothing is supposed to.

A threshold rule sets a distance instead. He picks a band around each target and does nothing while the portfolio stays inside it. When something moves outside, he acts, whenever that happens to be.

The difference is what the rule is watching. The calendar watches the clock and is indifferent to the portfolio; the threshold watches the portfolio and is indifferent to the clock. Each therefore fails in a way the other does not.

Where each one fails

The calendar's failure is that it is not connected to the thing it exists to control. In a quiet year it will have Joseph making small trades that correct a drift nobody would have worried about. In a violent one it will leave a large drift sitting there for eleven months because the movement happened just after he last looked. A rule that reviews annually accepts, by construction, up to a year of whatever drift the market produces.

The threshold's failure is the opposite. It is always connected, which means it demands attention continuously and can fire repeatedly in a turbulent stretch, generating trades and tax exactly when markets are moving most. It also has a subtler problem: the band is a number Joseph has to choose, and unlike a date, there is no obviously reasonable default. Set it narrow and he has invented a trading strategy. Set it wide and he has a target he never enforces.

Most people who think carefully about this end up combining them — look on a schedule, act only if the drift is beyond a band. That keeps the monitoring burden of a calendar and the risk-connection of a threshold, and it is the version worth recommending if Joseph wants one recommendation.

Choosing the band

If he goes down the threshold route, or the hybrid, the band is the only real decision and it should come from consequences rather than from convention.

The question is how much extra risk actually matters to him. A drift of a few points in a portfolio he will not touch for twenty years changes very little; the same drift in money he needs in three years changes a great deal. A band should be wide enough that ordinary market movement does not trigger it and narrow enough that a genuinely different risk profile does. Costs push the same way: where a trade is expensive or taxable, a wider band is not laziness but arithmetic.

The one thing a band must not be is negotiable in the moment. A threshold that gets widened whenever crossing it would mean selling something that has been doing well is not a rule; it is a preference with a number written next to it. If Joseph finds himself revising the band, the honest move is to admit the original target was not one he believed in and change that instead.

Why there is no answer from a backtest

It is tempting to settle this by testing both rules on history and picking the winner. That approach is more misleading here than in most places, and it is worth saying why rather than simply asserting it.

Any single sequence of returns will favour one rule over the other, and the margin will look convincing. But the mechanism producing that margin is the accident of when the peaks and troughs fell relative to the review dates, and there is no reason for that alignment to repeat. Rebalancing is not a strategy with an expected return to be measured; it is a control that keeps exposure near a chosen level. Testing it as though it were the former is the error described in data mining and backtest overfitting, applied to a question where the pay-off is not even the objective.

So this page carries no comparison table. The honest position is that the choice of rule is a second-order decision, and that we would be dressing up an arbitrary result as evidence by quantifying it. What is first-order is having a rule and following it.

What this looks like in practice

Write the target down, including what the property counts as and whether the gold is part of the stability allocation or its own thing. Untargeted assets cannot drift, because there is nothing to drift from.

Pick a review date and keep it. Once a year is enough for a long-horizon portfolio and the case for more frequent review is weak. Set a band so that small movements do not generate trades, and act when something is outside it.

Use incoming money first, which for Joseph includes the rent — directing it at whatever is underweight will do a surprising amount of the work without a single sale. And when the rule tells him to reduce the thing that has been performing best, that is not a signal that the rule has malfunctioned. It is the only moment the rule ever does anything, and the discomfort is the price of holding the risk he chose rather than the risk the market handed him.

What to take away

Calendar and threshold rebalancing are two ways of answering the same question, and the evidence does not crown a winner because the difference between them is smaller than the difference between having a rule and not having one.

Look on a schedule, act on a band, decide both in advance, and use new money before selling anything. Then leave it alone. The point of a rebalancing rule is to remove a decision, and a rule that gets reconsidered every time it fires has not removed anything.