What interest rates do to a currency

The one rate a committee actually sets 

A central bank does not set the rate on a mortgage, on a corporate bond, on a savings account or on the financing line attached to a currency position. It sets one narrow price: the cost of very short term money, usually overnight, between the banks that hold accounts with it. It does that either by declaring a target and dealing in the market until the overnight rate sits at it, or by announcing what it pays on reserves parked with it overnight, which puts a floor under what any bank will accept elsewhere. That single number is the policy rate, and it is the only rate in this lesson that is set by decision.

It has an unusual property for a price. It is administered rather than discovered. It changes on scheduled dates, by a vote taken in a named committee, in steps that committee chooses, and it then stays exactly where it was put until the next scheduled meeting or an unscheduled one. Nothing else in a financial system behaves that way. Every other rate is a market rate: the residue of transactions between parties free to disagree, moving continuously, quoted to more decimal places, and set by nobody in particular.

The distinction is worth holding because almost everything a trader watches sits in the second category. A government bond's yield is not announced by anyone. It is arithmetic performed on a price: the return implied for a holder who pays today's price and receives the bond's fixed payments through to maturity. When the price rises the yield falls, mechanically, with no decision taken by anybody. Interbank term lending, corporate borrowing and mortgage rates are then negotiated off those market rates, which is why a policy rate can sit unchanged for a year while the rates an economy actually borrows at move every day of it.

How one administered rate reaches every other rate 

The link runs through time. A bank holding a surplus can lend it overnight at something very close to the policy rate, so no bank will lend that surplus for three months at a rate meaningfully below what it expects to earn by rolling an overnight loan across those three months. That is the transmission mechanism in one sentence, and it generalises up the maturity ladder: a term rate is approximately the average of the overnight rates expected to prevail across that term, plus compensation for the risk of being wrong about them and for tying the money up.

That approximation carries a consequence which is easy to state and easy to forget. A term rate contains a forecast. The policy rate contains only a decision already taken. A two year government yield moves on most days of the two years between its issue and its maturity, and the bulk of what it moves on is a revised opinion about decisions that have not been taken yet. This is why market rates routinely move ahead of the committee, sometimes by months, and it is why a committee that does precisely what everybody assumed it would do changes very little about the rate at which anybody actually borrows or lends.

What the currency is priced off 

Money denominated in different currencies competes for the same holders. A deposit in one currency and a deposit in another are, to a holder otherwise indifferent between them, two offers with different compensation attached to them. The conventional account of an exchange rate, and it is a convention rather than a mechanism, is that capital is described as moving toward currencies where short dated instruments of low credit risk pay more, and away from those where they pay less.

Stated that baldly the account is wrong, and it is wrong in an instructive way. The compensation currently on offer is public information. It is already inside the price of every instrument through which that compensation could be obtained, and the exchange rate is one of those instruments. What is not already inside the price is a change in what the compensation is expected to be. So the conventional relationship is always stated in expectations rather than in levels: the currency of an economy whose expected path of policy rates has risen relative to another's is conventionally associated with appreciation against it, and the association is with the revision, never with the level.

A currency carrying the higher rate can therefore weaken for a year while it goes on carrying the higher rate, and nothing has been contradicted. The level was known throughout. Only revisions were news.

The two rates inside a pair 

The costs module established the other end of this chain. A currency pair is two interest rates carried inside one price, so a position in a pair holds one currency and owes the other, and a position left open past the daily rollover generates a financing entry whose size comes from the difference between two policy rates and whose sign comes from the direction held. That entry and the price of the pair are fed by the same differential. They are not two separate facts about a pair. They are one fact surfacing in two places on the same statement.

The same differential is experienced very differently at the two ends. The nightly amount is small and accrues once per night. A revision to expectations arrives at a single instant and prices in full across the whole position. One input, paid out in instalments at one end and settled in a single movement at the other.

Why a decision can move the price the other way 

Ahead of a scheduled decision the market is not waiting to discover what the rate will be. It has already formed a view, expressed that view in the prices of rate sensitive instruments, and traded accordingly. By the moment of the announcement, the expected outcome is embedded in the exchange rate. What is left to reprice is the distance between what arrived and what was expected, and then, usually far more consequentially, the revision to the expected path beyond this particular decision.

Case one is the mechanism behind the result readers find hardest to accept. The rate rose. The expected path fell. Those are two different numbers in one release, they moved in opposite directions, and only the second of them was news. Case two holds the delivered level identical and moves the path the other way, which isolates the point: the same published rate sits alongside opposite revisions, so the published rate cannot be what the repricing was about.

The decision is also rarely the only thing published. A statement, the split of the vote, a set of projections and a press conference commonly arrive alongside it, and every one of those is information about the path rather than about the level. The vocabulary practitioners use for a release whose level and whose tone point in opposite directions is the subject of a later lesson in this module. What a full release contains, and the order in which its parts land, is walked through in the guide to a rate decision.

Where the link breaks 

The chain from a policy decision to a currency's price runs through several links, and each link is a convention that has held in some periods and failed in others. The cases below are the ones that break it most often. They are described rather than ranked, because their relative importance is itself disputed.

• Inflation. A policy rate is a nominal number, while what the holder of a currency is compensated in is purchasing power. The comparison practitioners usually reach for is therefore the rate net of expected inflation. A rise in the policy rate accompanied by a larger rise in expected inflation leaves the inflation adjusted rate lower than it started, and the conventional relationship, stated in nominal terms, points the wrong way.
• Credit and fiscal stress. A yield is compensation for risk as well as for time. When a government's borrowing cost rises because lenders are demanding more for the risk of lending to it, the currency and the yield frequently move in opposite directions, which is the exact inverse of the conventional relationship.
• Risk aversion. In periods of market stress, currencies whose government debt markets are the deepest and most liquid are conventionally described as receiving flows for reasons unconnected to their rates, and differentials that appeared to describe the market a week earlier stop describing it at all.
• Managed and pegged exchange rates. Where an authority holds a currency at or near a fixed rate against an anchor currency, its rate decisions conventionally follow the anchor's rather than domestic conditions, and the exchange rate does not respond to either. Everything in this lesson describes floating rates. The managed case is taken on its own terms in a later lesson in this module, and it is the arrangement several currencies in this region operate under.
• Convertibility. Where the movement of capital is restricted, the flows the conventional account depends upon cannot occur at the size that account assumes, so the relationship it describes has nothing to work with.

Where practitioners disagree 

The first argument is old, and it is about whether an interest differential is compensation for anything at all. The proposition named uncovered interest parity holds that a currency carrying the higher rate should be expected to depreciate against the lower one by approximately the differential, leaving a holder indifferent between the two, so that the differential is payment for an expected depreciation rather than a return. Forward exchange rates are quoted on exactly that arithmetic. Whether observed spot behaviour matches it has been argued over for decades without resolution, and the lack of resolution is the useful part: one tradition treats the differential as an amount that arrives nightly, another treats it as compensation for a move that arrives all at once, and both are reading the same two numbers.

The second argument concerns the phrase priced in, which this lesson has used repeatedly and which is harder to pin down than it sounds. One tradition reads an expected policy path directly off rate sensitive instruments and treats the resulting numbers as the market's expectation. Critics of that reading point out that those prices contain compensation for uncertainty as well as an expectation, so what is extracted is an expectation plus an unobservable premium, and the two cannot be separated from the price alone. The disagreement matters directly here, because a surprise is measured against whichever estimate is used, and two desks working from different estimates can describe the same release as surprises of different sizes.