What a Corridor Actually Is
Two things share the name
A corridor is a floor and a ceiling with a promise attached: inside the boundaries, nothing is supposed to require action. That promise is what makes it useful, and it is also where the word gets abused.
There are two kinds of corridor and they have different obligations. A policy corridor is an operating band a central bank enforces with its balance sheet — it holds because somebody will transact at the edges in unlimited size. A trading corridor is a range that someone believes price will stay inside — it holds because market participants believe it, and it breaks the moment they stop.
Same word, opposite guarantees. Always ask which one you are looking at.
The enforced version: the Fed runs a corridor, not a rate
The Federal Reserve does not set one interest rate. It sets a band. At the top sits the rate at which banks can always borrow against collateral from the central bank; at the bottom sits the rate at which eligible counterparties can always park cash overnight and be paid for it. Inside the band sits the administered rate the Fed pays on reserves, which acts as the anchor.
The design is deliberately simple: a hard ceiling and a hard floor mean the market has no reason to trade outside them. Our own board as of 21 September shows it working — the administered rate at 3.90%, the effective federal funds rate at 3.88%, and SOFR at 3.85%. Both market rates sit below the administered rate, which tells you reserves are still abundant: nobody is pressing against the ceiling and nobody needs the floor.
But a floor is only a floor if there is a balance standing behind it. Reverse repo take-up printed at $0.6bn on 21 September — effectively nothing parked at the facility. A boundary that anyone can reach is not the same as a boundary with capacity behind it, and that distinction is the entire subject of our 22 September plumbing piece.
The judged version: a trading corridor, tested
The second kind of corridor is a claim about price. This column made one in early September: a 40 to 60 basis point band for the 2s10s spread.
Here is the whole test, from Treasury's own par yield curve. The spread printed 40.0 basis points on 1 September, 43.0 on the 3rd, 41.0 on the 4th, 41.0 on the 8th, 40.0 on the 9th — six consecutive sessions inside the band. Then 39.0 on 10 September: the first close below the floor. From there it did not bounce. 33.0 on the 11th, 27.0 on the 16th, 25.0 on the 18th, and 20.0 on 21 September — 20 basis points below the floor, and the lowest close of the year.
The value of that exercise was never the prediction. It was that the corridor defined failure in advance. A range with a stated floor tells you the day you are wrong — 10 September — rather than letting a thesis dissolve gradually into a rewording.
Corridors built out of percentiles
The judgement version has an obvious flaw: where do 40 and 60 come from? The mechanical answer is to take the same series and set the floor and ceiling at low and high percentiles of its own history — a 10th and 90th, say — which deliberately excludes the tails and leaves what practitioners call the core trading range.
This is a real improvement in one respect: a percentile-built corridor is mechanical, which means it can be measured. You can ask how often it has been breached, how long breaches persisted, and whether the breach preceded anything. A judgement corridor can only be argued about.
It also imports a weakness from the percentile itself: the band is only as meaningful as the window you built it from. A 10th percentile of 3 years and a 10th percentile of 30 years are different floors for the same price — which is why the percentile lesson on this site is worth reading before you build one of these.
What actually breaks a corridor
A corridor does not break because price wandered. It breaks because the thing enforcing it changed.
For the policy corridor, that means the balance behind the floor drains — the boundary is still printed in the operating framework while the capacity that made it binding has gone. For a trading corridor, it means the driver moved: in our case the front end repriced the policy path while the long end refused to follow, so the spread was not drifting through a floor, it was being pushed through one.
That is the useful habit. When a corridor gives way, look at the mechanism, not the line. The breach is information about the driver; the boundary was only ever a description.
Next: What a Percentile Hides