The Marginal Cost Controversy

How do we inform government interventions with prices?

Imagine a factory which, once set up, is able to produce additional units for just the cost of the materials. Since they pay a fixed cost for entry, and then constant marginal costs, they necessarily have falling average costs. In order to break even in a competitive market, a firm must price above its marginal cost, which necessitates producing less than is socially optimal. There will be people who value the good in excess of the cost to produce it, but less than the price which must be charged for the good to be produced at all. If the government provides a lump sum subsidy equal to the fixed cost, then the firm will produce at marginal cost, and social welfare is maximized.

Such is the argument of Harold Hotelling, in his paper “The General Welfare in Relation to Problems of Taxation and of Railway and Utility Rates” (1938), given as an address to the Econometric Society in that same year. It kicked off a flurry of discussion, culminating with Ronald Coase’s 1946 paper, “The Marginal Cost Controversy”. He imagines a world where consumers are on the end of roads, and are ordering goods to be shipped from the central hub. It is correct that equating marginal costs and benefits leads productive efficiency, but this is true if and only if consumers would have been willing to pay for the transportation. A subsidy assumes that they are, but we cannot know this in the absence of any information about people’s true desires. Coase argues, in reference to the possibility of forecasting whether the bridge be used, “If it were possible to make such estimates, at low cost and with considerable accuracy and without knowledge of what had happened in the past when consumers had been required to pay the total cost, this would be likely to lead, in my opinion, not to a modification of the pricing system but rather to its abolition. The pricing system, as I pointed out earlier, is a particular method of allocating the use of factors of production between consumers and the arguments for its adoption derive their main force from the view that such estimates of individual demand by a Government would be very inaccurate.”

These arguments echo that of Hayek, in “The Use of Knowledge in Society”. The possible uses of resources are scarcely imaginable, and we need a price system to provide accurate information. It certainly doesn’t help that demand curves are not directly measurable without plausibly exogenous shocks. To know the demand curves, we need unsubsidized markets, and even then we cannot expect them to remain the same over time. Vickrey, in his 1962 paper now more famous for proving the equivalence of a second-price auction and an ascending price auction, begins by exploring the possibility of the government intervening to “counterspeculate” and bring cost and benefit into perfect equality. He rejects this as requiring far more information than the government could possibly know.

I would argue that Coase is, if anything, understating the difficulty of knowing the optimal uses of resources. Coase is granting that we know the possible uses of factors, and lack only the knowledge of which use we should do. However, technology is not static. There are myriads of possible goods, the nature of which cannot be described ex ante. To equate marginal cost and benefit by subsidy on a static technology is to give up on progress. I am hence skeptical of things like the EU mandating USB-C usage. It certainly will make life more convenient – I would not deny that for a second – but it would also make it much harder for any of the big companies to make an advance in charging technology. We will certainly not be getting any iterative improvements over it. It may be justifiable in this particular case, but we are surely biased against the things which we cannot see than those we can.

We today want to subsidize the discovery of technologies whose costs and benefits we do not know in advance. Coase proposes a solution in the case of the radial roads of a multi-part price; a price which consists of a fixed fee to have the good shipped at all, and then a price for the good. (This is, to be clear, simply the equivalent of paying for shipping). This is not adequate, however, when the benefits of the infrastructure or the idea are not clearly divisible by person. The electricity generated by a dam, for example, can be charged for, but the cost incurred by the dam is not easily divisible. I would not go so far as to claim there should be no government involvement in solving public goods problems. I do believe there are ways for us to use market information to make better decisions.

One of the more elegant methods of subsidizing innovation with the information gained from prices is the Kremer patent auctions, which I have covered earlier. The government holds an auction for patents, but only sells it some of the time. In most cases, the government buys it from the patent filer and puts it into the public domain. In this way, the patent holder is rewarded to the same extent they would have been, but without the need for monopoly distortions. So long as the distortions from taxes are less than the distortions from patents, we can improve outcomes this way.

Another way are assurance contracts. Suppose there’s something basically everyone wants done, though not at any cost, and which everyone would benefit from whether they paid or not. The standard example is flood control from a dam. Now, the government could levy taxes and pay for it, but we can do better by announcing a contract in which the dam gets built if and only if everyone agrees to it. If there is a small cost to agreeing to the contract, then it is vulnerable to negative beliefs about failing, so we can do one better by having an entrepreneur offer a contract in which he pays those who enter a fee if the contract falls through. (In return, a fee for the entrepreneur is set aside if the contract succeeds). This is called a dominant assurance contract by Alex Tabarrok, who proposed it.

I favor quite a lot of government interventions. The presence of fixed costs means that market outcomes can substantially deviate from what is socially optimal. Nevertheless, perfect allocative efficiency is an impossible dream, and the government should stick to the obviously good and useful projects. When they do so, they should try and incorporate the power of the price system in finding information.