Contestable Markets: An Uprising in the Theory of Industry Structure: Comment
批评了完美可竞争市场理论，指出其通过假设无沉没成本来回避规模报酬递增问题，并论证了没有沉没成本就不可能出现平均成本递减，因此完美可竞争性仅在成本近似非递减时近似成立。
A situation traditionally identified conducive to imperfect competition is when average costs decline and the cheapest scale of production is large relative to the size of the market. However, some of the more recent literature on industrial organization has emphasized supposed distinction between fixed costs of production and sunk costs of production. A point frequently made is that while the possibility of sunk costs can create genuine barriers to entry, fixed costs per se do not inhibit competitive market performance. This approach finds its logical extreme in the norm or abstraction of perfect contestability. Perfect contestability is essentially theory of the polar case of frictionless entry and exit. As William Baumol stated, a contestable market is one into which entry is absolutely free and exit is absolutely costless (p. 3). Certainly this is the most reasonable interpretation of situation where potential entrants feel free to disregard an incumbent's price response. Suppose every potential and actual producer has access to the same technology. In perfectly contestable market, there are no set-up or shut-down losses. The main result is that even with declining average costs of production, the incumbent firm does not dare to post price higher than average cost. Price above cost invites being undercut by hit-and-run shadow entrant capable of producing at the same flow rate and at identical unit cost during just the briefest instant of time. I argue that this line of reasoning is misleading. Perfect contestability gets around the problem of increasing returns only by, in effect, assuming it away. A hit-and-run technology makes the firm behave as if it is competitive in market precisely because the convexity preconditions for competition are de facto being met in that market. This comment shows that, strictly speaking, there is no such thing pure fixed cost. Unless there are sunk costs located somewhere in the relevant production technology, all costs are variable. As matter of formal theory, you cannot have range of decreasing average cost without sunk costs. Presumably there is also an approximation theorem which states that when sunk costs are close to being negligible then average production costs are practically nondecreasing. Perfect contestability holds approximately in market only to the extent that production costs for the market are approximately nondecreasing. My discussion will be restricted to the familiar case of single, well-defined, homogeneous, fully divisible commodity. While more general approaches are possible, for the sake of simplicity the following definition is used. The average cost function A C(y, t) describes, say, the minimum cost per unit of output produced at uniform flow rate y throughout the time interval (0, t]. Especially in applications to industrial organization theory it is essential to remember that cost functions are not generally timeless, and that writing AC function of y alone can be dangerously vague. Behind the free entry and exit of hitand-run technology in perfectly contestable market is the abstraction of no sunk costs-investments are fully reversible because nothing is lost in setting up or shutting down production.