Where a price comes from

The two numbers on the screen 

A currency pair on a trading screen carries two numbers beside it, and both of them move several times a second whether or not anything is visibly happening in the wider world. Nothing in the arrangement says where either number came from. There is no ministry of exchange rates, no institution whose published figure every other institution copies, and for a pair like that one there is no official rate at all. What the screen shows is the visible surface of an argument between the participants named in the previous lesson, settled and re-settled faster than the display can redraw.

The bid and the offer are already familiar from the costs module. What is worth taking apart now is what each of them actually is underneath the label. A bid is a conditional commitment by somebody to buy a stated amount at a stated price for as long as the commitment stands. An offer is the same kind of commitment in the other direction. Neither is a description of the market and neither is a forecast about it. Each is an obligation a participant has voluntarily taken on, which is why each disappears the moment its maker withdraws it.

Nobody publishes the price 

The contrast with an exchange makes the point. When an instrument trades on an exchange, matching is centralised: orders arrive at one venue, the venue matches them by a published set of rules, and every completed trade prints to a tape all participants can see. Even there the price is not decreed by anybody. It is the outcome of orders meeting, and the last printed trade is a record of something that has already finished happening rather than a price any participant is currently obliged to honour.

Spot currency has no such centre. Dealing happens bilaterally, between banks, market makers, funds and brokers, across many venues at once, and no tape consolidates the result. There is therefore no number that could honestly be called the price of the pair, in the way there is a last traded price for a listed share. There are only the prices individual participants are willing to deal at right now, each posted for that participant's own reasons: the inventory it is carrying, the risk it is prepared to hold, the flow it has just seen, and what it believes about the next few seconds. A firm that wants to show a client a quote has to go and collect them.

A price that describes, and a price that commits 

Two quite different things get called a price, and confusing them is the most common reason a reader concludes a screen must be wrong. A reference price is a number that describes a market without obliging anybody to transact at it: an index level, a central bank fixing, a previous close, a benchmark averaged across a stated window, the figure printed beside a pair on a news page. A dealable price is one a named counterparty will actually transact at, in a stated size, for a stated moment.

Dealability has three parts, and all three have to be present. Who: a specific counterparty that has taken the obligation on, since an unattributed number obliges nobody. How much: an amount for which the obligation holds, because a price good for a small amount is not the same price as one good for a hundred times that amount. How long: an interval, often measured in milliseconds, after which the commitment lapses. A reference price has none of the three. It is a photograph of something that has already happened, or an arithmetic summary of several such things, and a photograph cannot be transacted against.

The midpoint between a bid and an offer is the clearest case of the difference. It is a genuinely useful number, because it describes where a market is without taking a side. Nothing trades there. No participant has committed to it, and a position opened at a midpoint would be a position opened at a price nobody had offered.

Assembling the quote 

A firm that quotes a client price subscribes to a set of sources. Each source streams its own two sided prices, each carrying the size behind it, and each stream updates continuously and independently of the others. The firm's systems read all of them at once and take the highest bid available and the lowest offer available. That pair is the top of the book, and some version of it is what a client screen shows.

Two features of that arithmetic are easy to read past. The composite is a hybrid: the bid came from one participant and the offer from another, so the pair as a whole never existed at any single source and nobody is standing behind both halves of it. And it is only true for the size sitting behind it. An amount larger than the size at the top reaches into the prices underneath, which are worse by construction, because if they were better they would have been the top. How a firm chooses its sources, ranks them, filters them and layers them is the subject of the next lessons in this module.

Why two firms show different prices at the same moment 

If every firm runs the same arithmetic over the same market, it is reasonable to expect the same answer everywhere. Firms do not produce one, and the reasons are structural rather than suspicious.

• Source sets differ. No firm is connected to every participant in the market, and the set it is connected to sets a ceiling on the best price it can possibly see.
• Filtering differs. Firms drop sources whose prices are frequently unavailable by the time an order reaches them, and they differ in how fast they drop them and on what evidence.
• The size assumed at the top differs. A quote shown for a small amount and a quote shown for a large one are different quotes, even when they are assembled from identical sources at the same instant.
• Displayed precision differs. A price carried internally to five decimal places and shown to four has been rounded, and rounding conventions are not uniform across firms or across instruments.
• Distance differs. A price takes time to travel from the source to the firm and from the firm to a screen, so two screens are always showing two slightly different moments.
• Any adjustment the firm applies differs, and whether one is applied at all is a property of its execution model, which is a later lesson in this module.

Neither quote there is wrong, and neither firm has done anything to the price. Each took the best of what reached it, and what reached them was not the same. An instrument that trades on an exchange narrows the gap, because a central print exists for a derived quote to reference, but it does not close it. Derived quotes still differ by the moment the reference was taken, by the conversion applied when the underlying is denominated in another currency, and by whether the reference is the cash market or the futures contract written on the same underlying.

What happens when sources withdraw 

A quote is not a fixture of the world. It exists because participants are currently willing to stand behind prices, and that willingness is conditional on everything they know at the time. Sources widen their prices or step away entirely around scheduled announcements, at the daily rollover when desks hand over, on public holidays when a whole region's participants are absent, and whenever volatility passes the point at which the risk of facing a better informed order outweighs whatever quoting earns them.

The effect on the assembled quote is mechanical rather than mysterious. Withdrawal removes the best prices from one or both sides, so the top of the book falls back to the next price still standing, which is by definition further away. The size behind the top falls at the same moment, because that size belonged to the sources that left. Nothing has to trade for any of it to happen. A price can move a long way while no transaction occurs at all, simply because the commitments that used to sit in between have been withdrawn.

The costs module described this event from the point of view of an order: an instruction arrives, and the price available when it arrives is not the price that was on the screen when it was written. This is the same event seen from the other side of the wire. The order did not miss the price. The price stopped existing, and the nearest one still standing was somewhere else.

Where practitioners disagree 

Whether the midpoint deserves to be called the price is genuinely contested, and both positions are held by serious people. One tradition treats the midpoint as the market's fair value and the two quoted sides as the cost of transacting around it, which is how most valuation and most measurement of execution is done. The other holds that a number nobody will deal at is not a price in any operational sense, and that measuring against a midpoint understates what changing a position involves. The disagreement survives because the two camps answer different questions, valuation in one case and transaction in the other, and neither answer is available to the other.

The second argument is about whether a source that has shown a price may decline the order arriving against it. The practice, which the industry calls last look, gives the source a brief window to check the order against its own current view before accepting it. Its defenders argue the window is what allows a source to show tight prices to counterparties whose intentions it cannot see, and that removing it would widen quotes for everybody. Its critics argue it hands the source an option it did not pay for, settled by the firm and ultimately by clients in rejected orders. Industry conduct codes have narrowed the practice and required it to be disclosed rather than abolishing it, which is a fair description of an argument that has not been settled.

A third disagreement is quieter. When no dealable price is available, some firms display an indicative one anyway, on the reasoning that a stale number informs more than an empty field. Others hold that a number nobody will honour is worse than nothing, because it looks identical to a price that can be dealt on and no convention distinguishes the two on a screen. Both practices exist, which is the strongest available argument for reading the label on a price rather than only the number.