Layered FX hedging, explained

One decision, spread across many days.

Instead of covering a June exposure in a single trade, you cover a portion of it each month from the previous summer onward. By the time June arrives the position is fully hedged, but at a blend of every rate you dealt at along the way rather than at whatever the screen happened to show on one particular morning.

The mechanism is worth stating plainly, because it is the part that gets lost. What smooths the rate is not that you traded more often. It is that consecutive value dates come to share most of their trade dates. June and July are both built from the same run of monthly executions, so the two months cannot end up far apart, and neither can they diverge sharply from the months either side.

A useful side effect follows from that: you no longer need the forecast to be right. A programme that hedges a rolling proportion of a rolling forecast tolerates the forecast moving, because each month only ever commits a slice.

What the textbook ladder actually looks like.

The Association of Corporate Treasurers describes the standard form as a set of coverage targets by lead time. More of the near months are hedged, less of the far ones, and each rung is topped up as the one in front of it matures. The proportions below are the ACT illustration, not a recommendation, and the right numbers for any given book depend on how firm its forecast is.

Note what the shape encodes. Near months are forecast with confidence, so they carry high coverage. Far months are guesses, so they carry a little. The ladder is a statement about forecast quality as much as about the market.

It also spreads the mark to market. A book hedged in one trade carries one large valuation swing. A book hedged across eighteen executions at eighteen different rates carries eighteen small ones that substantially offset, which is usually the difference between a quiet month end and an awkward one.

It narrows the range. It does not move the middle.

This is where most explanations of layering quietly overreach, so it is worth being exact. If you assume the market is as likely to move up as down, the expected rate from layering is the same as the expected rate from dealing once. Not slightly better. The same, to the decimal.

The research on averaging into a position points the same way, and if anything points against it: deploying in one go beats averaging in across roughly two thirds of historical periods, simply because markets drift upward more often than not. Layering is not a way of getting a better price. It is a way of being less exposed to which morning you happened to pick.

The benefit of layering is the narrowness of the range you land in, not the position of its centre. A programme sold on a better average rate is being sold on the one thing it does not do.

That distinction matters commercially as well as intellectually. A treasury policy justified on rate improvement will eventually be measured on rate improvement, and over any single year the result will be dominated by which way the market went. A policy justified on reduced variance can be measured on reduced variance, which is a thing it genuinely delivers and a thing a board can be shown.

Two different things are both called layering.

Anyone reading around this subject will meet the word attached to two different mechanisms, and they behave differently enough that treating them as one is a real source of error.

The first is the ACT ladder above. It is triggered by the calendar: a rung comes due on a date, and it is executed on that date at whatever the market offers. The second sets a target level on each rung and fills it when the market reaches that level, with whatever remains booked at the deadline regardless.

They are not interchangeable. The calendar version is fully deterministic and will complete. The price version may not fill every rung, so it needs a rule for what happens at the deadline, and the portion that ends up filled by that rule is adversely selected by construction: it fills precisely when the market did not come to you. Any honest account of a price ladder has to include that residual, which is also why it cannot be presented as a way of getting a better rate.

A risk decision, not a return one.

Layering is worth doing for reasons that survive scrutiny. It removes the need to be right about timing. It removes the discretion that makes a single dealer responsible for a whole year. It reduces the size of individual valuation swings. It tolerates a forecast that moves. None of those is a claim about the average rate, and none of them needs to be.

What it requires in return is record keeping that most spreadsheets do not survive. A programme is only measurable if every rung carries the date it was due, the date it was actually dealt, the rate achieved and the reason for any difference. Without that, a ladder is an intention rather than a policy, and there is no way to demonstrate at year end that it was followed.