A price held above the full cost of supplying a market is an advertisement addressed to every would-be rival. Here, it says, is something worth making; here is a seller who can be undercut. Yet some advertisements remain unanswered for years. The rival sees the same price that the customer sees, but must ask a different question. The customer asks whether the purchase is worth making today. The rival asks what the market will look like after it has spent the money, assembled the staff, won the customers and announced its arrival. A price that supports an established business may disappear before a newcomer has recovered the cost of becoming one. The barrier is whatever makes that prospect sufficiently unequal, costly or uncertain to keep the profitable position safe. To identify it, we must follow the entrant through the gate, rather than merely count the sellers already inside.
The cost of entering matters through what it buys, what it risks and what can be recovered on leaving. A building that can be sold for what it cost ties up funds but need not put the whole purchase price at hazard. A campaign to teach customers an unfamiliar name may leave little to sell if the venture fails. Both require finance; only the second is wholly lost on exit. Even this distinction does not settle the question. The incumbent may once have borne the same irrecoverable expense, yet no longer need to recover it in deciding how low to price tomorrow. The entrant must decide before spending; the incumbent responds after spending. Their physical technologies may be identical while their decisions are made from different positions. That difference can make an otherwise efficient newcomer hesitate without proving that every large capital requirement is an exclusionary obstacle.
Scale can shelter the established firm because the market may genuinely be cheaper to serve through a single set of works. Consider an invented service with one hundred customers, a fixed cost of one hundred units of money per operating period and a cost of two for each customer served. One supplier spends three hundred. Two suppliers dividing the customers equally spend four hundred, because they duplicate the fixed installation. A price of three covers the single supplier's full cost; a price of two, equal to the cost of one more customer, leaves one hundred unpaid. These figures include the normal return to the funds employed among the costs. The margin between two and three therefore does not establish an exceptional profit. Nor does the saving from a single installation authorize any price the supplier chooses. The same calculation that explains why rivalry may be expensive leaves open the question of how the cheaper arrangement is to be held answerable.
The entrant must also acquire a place in the customer's calculations. Familiarity can be earned by reliable service; it can also become protection against a service that would now be better. A buyer who must learn a new system, move records, persuade colleagues or risk a lost connection needs more than a slightly lower posted price. The promised saving must cover the cost and uncertainty of moving. Where the service joins people to one another, one person's willingness to move may depend on others moving at the same time. The incumbent then possesses something a rival cannot produce merely by building an equally good machine: an assembled body of users already expecting to find each other there. This is a possible source of protection, not a law that every network must have only one owner. Customers who can use several services, or carry their connections between them, give the rival a different opening.
Some gates depend on permission rather than on production. An exclusive right, a franchise or a scarce connection may keep an entrant out even when it could supply at the same cost. These arrangements can serve purposes that must be examined on their own terms: funding an invention, coordinating a shared facility, ensuring that someone undertakes a costly obligation. Their existence still changes the contest. The incumbent's price may reflect the terms on which a permission was granted as well as the work of supplying the service. An inquiry into the barrier must therefore ask who controls the permission, how long it lasts, whether rivals can obtain an equivalent right and what obligation its holder accepts. Calling the resulting position either natural or artificial, without answering those questions, tells us remarkably little about whether it should endure.
The contrary possibility is that a gate which can be crossed cheaply need never be crossed at all. This is the insight of contestability. A single supplier may face a threat credible enough to restrain its terms without acquiring a permanent neighbor. For that threat to have the force assigned to it in the ideal model, the newcomer must have access to comparable technology and product quality, must be able to enter without an incumbent advantage defeating it, and must be able to recover its capital on leaving apart from ordinary use and depreciation. Timing matters as much as cost. The entrant must be able to win business at a price based on the incumbent's existing terms before a response wipes out the opportunity. These conditions describe a demanding benchmark. A warehouse that can be hired and released quickly resembles it more closely than a specialized installation whose only buyer on exit would be the firm it challenged.
A credible threat can eliminate exceptional profit while leaving the single supplier's price above marginal cost. Return to the invented hundred-customer service. At a price of four the incumbent receives four hundred against a full cost of three hundred. Suppose a newcomer can supply the whole demand with the same costs, win all the customers at 3.90, and complete that business before the incumbent changes its terms. It can earn ninety. If the capital required for such an attempt is recoverable, the incumbent's hundred-unit surplus is exposed to attack. At a price of three the full cost is already covered without an exceptional return, and this particular opportunity has vanished. Price is still one above marginal cost. The threat has disciplined the division of receipts without making the fixed installation free. Confusing those two results would turn an entry test into a demand that the service finance itself with money the example does not provide.
The absence of an entrant, accordingly, can mean either that the threat has worked or that the threat has been defeated. The observation alone will not distinguish them. A lone seller charging little may be protecting itself against a rival that could arrive next month; a lone seller charging much may be protected by a cost that no rival can recover. Equally, a market full of firms can contain customers who cannot readily move between them. Counting establishments gives a useful description of the scene, but the decisive evidence concerns alternatives: what a new supplier would have to spend, which customers it could reach, how soon it could supply, how the incumbent could respond and what would remain if the attempt failed. Competition is present in those possibilities before it is visible in a second sign above a second shop.
Uncertainty can keep an entrant waiting even when entry has a positive expected return. The choice may include an opportunity to learn before making an irreversible commitment. Waiting has a cost—the business forgone in the meantime—but it can spare the firm an investment that later evidence would have warned against. An incumbent's ability to alter terms after entry is one part of that uncertainty; changing demand and technology are others. A barrier analysis that treats the observed margin as a guaranteed stream of future receipts leaves all this out. It also leaves out the distinction between delay and permanent exclusion. A new technology, a larger market or a way to lease what previously had to be built may change the calculation. The protection is an arrangement of circumstances, capable of strengthening or dissolving, rather than a permanent quality stamped on the firm.
Removing a barrier is therefore an exercise in distinguishing the protection from the thing it protects. A demand that entrants duplicate an essential facility may preserve the incumbent when access to the facility could admit rivalry. A rule admitting rivals without requiring them to bear an associated service obligation may instead unravel the means by which the obligation is financed. Shared access, portable records, a different franchise or a narrower exclusive right each changes a different part of the problem. None is a general substitute for asking where the actual constraint lies. The useful comparison is between workable arrangements, including their costs and their opportunities for evasion. If one supplier remains the economical structure, accountability must be found somewhere in that arrangement; multiplying costly installations is not the only way to make its terms contestable.
The barrier also determines how much of a successful innovation its maker can keep. An improvement that every rival can reproduce tomorrow confers a different reward from one that customers cannot leave for a decade. Some prospect of keeping a return can support the initial experiment; a durable shield can also reduce the pressure to undertake the next one. The balance cannot be read off the existence of a large firm, or a large profit, without asking what created it and what can now challenge it. This returns us to the distinction the first plates established. The single seller's position may have a defensible origin and still require scrutiny in its continuation. The customer meets the terms being offered now. The historical story matters because it explains those terms, not because it settles every future claim to them.
The rival need not arrive to matter. It must be able to arrive on terms the incumbent cannot safely ignore.
What keeps the single seller safe is finally a question about the choices available to everyone else. Can another supplier reach the buyer? Can the buyer move without abandoning what makes the service useful? Can capital be withdrawn if the attempt fails? Can permission be obtained without the incumbent's consent? The answers tell us whether an unoccupied place in the market is an open invitation or a place deliberately made uninhabitable. Contestability reminds us that an invitation may discipline the host even when nobody accepts it. The barriers remind us how much has to be true for the invitation to be real. Between them lies the task of reading a price as the outcome of a contest whose most important participant may be the one we never see.