A price can be agreed while the trade remains difficult to make. The buyer must find the seller, establish what is being offered, arrange payment and obtain delivery. Someone may have to wait, carry the object or accept the risk that a promise will fail. The spread gives part of this work a visible form: the difference between what can be received on selling and what must be paid on buying. It makes immediacy costly to the person who wants it, while offering a receipt to the person prepared to supply it. Yet the gap between two quotations cannot contain every cost of exchange. A narrow spread can accompany an expensive search; a trade at one price can still require verification and settlement. The closing question is therefore what has actually been removed when the gap disappears. The answer determines whether a frictionless market is an attainable arrangement, a useful assumption, or a price that leaves some of its own conditions uncounted.
The simplest spread calculation already requires a description of the trade. Suppose a dealer stands ready to buy one unit at ninety-nine and sell it at one hundred and one. A customer who buys and immediately sells back, with the quotations unchanged and no other charges, loses two. Relative to the midpoint of one hundred, each leg costs one; calling each leg a two-unit cost would count the spread twice. If there is also a commission of one quarter of a unit on each leg, the round trip costs two and a half. These are authored numbers, and the fixed quotation matters: a later price movement would add a gain or loss that the initial spread alone did not measure. Size matters too. A quotation for one unit does not promise the same execution price for an arbitrarily large order. Demsetz's account of the cost of transacting isolates the service of prompt exchange, distinguishing it from the wider work of deciding what to trade and from separately charged brokerage. The quoted gap is a measure with a defined object, not a complete account of everything the customer gives up.1
Two customers with compatible orders can instead meet at one hundred. Their exchange need not pass through a dealer who first buys at ninety-nine and then sells at one hundred and one. Matching the orders can avoid that particular demand for dealer inventory, while leaving a venue, broker or the customers themselves to do the matching and complete the transaction. If the counterpart is not already present, waiting for one has an uncertain duration and can expose the trader to a missed opportunity. Paying for immediacy and accepting that uncertainty are different ways of arranging the exchange. Demsetz makes the distinction between prompt execution and a trade whose opposing orders arrive together. It follows that a transaction at a common price is perfectly intelligible without establishing that the market has become costless. Even the absence of a separate fee would tell us only that the relevant work is not being charged in that particular form. Its resources might be supplied by the parties, paid through another service, or reduced by a better method.2
This movement from a visible charge to the work behind it is also the movement in Coase's explanation of the firm. Using a market requires discovering relevant prices and negotiating and concluding agreements. An organization can bring some of that coordination inside a continuing relationship, replacing repeated bargains with directions issued within agreed limits. Its advantage depends on the costs it saves and the costs of directing, monitoring and making mistakes that it creates. The comparison must therefore include both arrangements. A firm that avoids an outside commission can still use more resources administering the transaction itself; a specialist paid to stand between two parties can make their exchange cheaper in total. Technology changes the comparison when it changes either set of costs, and its effect on the boundary of the firm depends on their relative movement. The middleman's removal, the broker's fee and the dealer's inventory all become instances of a larger question: through which feasible arrangement can this particular exchange be carried out on the best terms?3
The frictionless model makes a different move. Within its specified domain, it removes the costs or restrictions that would otherwise stand between a desired exchange and its execution. An asset can then be bought or sold at the common price the model supplies, without the bid–ask wedge or trading charge that the abstraction has excluded. This simplification can reveal how values and allocations are connected when those obstacles are absent. It cannot establish that the real institutions enabling exchange are dispensable. A rule identifying the asset, a procedure for settling the payment, or an inspection that reduces the risk of receiving something else can cost resources while permitting trades that would otherwise fail. Coase's later account emphasizes the institutional setting of the market itself, including the rules of an exchange. Removing a costly safeguard is an improvement only if the resulting arrangement is better after the consequences are included. A model that assumes the problem solved does not tell us which actual method of solving it is least costly.4
The distinction between a payment and a resource cost matters when the account moves from the trader to the economy. A fee is an expense to the customer and income to its recipient. Part of it may pay for labour, equipment, finance or the bearing of risk; part may reflect control over access or a scarcity of competing providers. Counting the whole fee as resources destroyed would erase the recipient without explaining the work performed. Treating the whole fee as proof of useful work would excuse every charge a gatekeeper could sustain. A better account asks what the service makes possible, what resources it uses, what alternative is available, and how the gains are divided. A lower charge can improve the customer's position even where its immediate effect is a transfer from the provider. A better method can also reduce the resources consumed and make previously uneconomic trades worthwhile. The two improvements are related, but their measures differ. If the benefit of an exchange is smaller than the cost of arranging it, the transaction may never appear in the record at all.5
The spread has carried the inquiry from the dealer's two prices to the conditions under which a price can be acted upon. Search, waiting, inventory, information, credit, access and representation each supplied a reason why standing between might be useful, and a place where control might become profitable beyond the service it furnished. No single width can decide the entire question. What can be demanded is an account of the exchange: the promised service, the feasible alternative, the resources and risks involved, and the people left with the gains. This closes the family of prices without making any of its members innocent by definition. The wage, rent, rate, premium and spread each name a payment whose explanation depends on the relationship that sustains it. The next volume, The Margin, turns from the names of those payments to the point at which a decision changes: the additional unit, the next cost, the benefit that remains just sufficient. Once the cost of reaching a market is admitted, even that small step has somewhere to begin.