It seems like Craig Westover and I speak more often than I speak with other bloggers, because we have a good deal in common. This includes a relatively libertarian view on social issues and in the main on economic issues as well. So when Craig first sent me an email over the weekend outlining the positions he takes in this post, I realized here was another chance at some education.

My own education on public goods comes from a series of people who studied at the University of Virginia in the 1960s. Over time three of them -- Craig Stubblebine, Tom Borcherding, and Tom Willett -- instructed me in public finance, public choice and public policy; I read a great deal of James Buchanan, Gordon Tullock, Geoffrey Brennan and other writers of the school. Oddly enough, the assistantship I turned down last to take my fellowship at Claremont was from UVa. So it seems to have been predestined that I would fall into that camp.

Stubblebine was my public finance professor, and he first got me to think about the nature of public goods. When I said what I said on the air, it was his teaching that was in my mind. He's of course not to blame for my misunderstanding of his teachings, if I have. But I think I have this right.

King did however, make one statement that surprised me; that is, King stated in passing that there is no such thing as a public good. To the pure libertarian, that's true, but sticking to that "truth" as a matter of policy leaves the door open for politicians to step into the vacuum and define �public good� anyway they want to -- with usually disastrous political and economic repercussions.

Now part of me simply doesn't care about shading a "truth" in order to be sure that public policy isn't hosed up by political repercussions, but since Buchanan and Richard Wagner wrote Democracy in Deficit -- a wonderful, slender book that argues that Keynes let the genie out of the bottle by providing intellectual cover for deficit spending -- I can't let myself off so easily.

Let me clarify what this is not: This is not about my taking a pure libertarian position. That is, one could argue that public goods do exist as a unique, objectively determined category and still argue as a matter of morality that one cannot support the state. Nor am I arguing for an anarcho-capitalist state or a minarchy. (One of these days I will come back to minarchy, because Craig believes himself a classical liberal but I do not think he's a minarchist. I'd like him to explain the difference. Call this a marker for another debate.)

Stubblebine did have us read Buchanan's Demand and Supply of Public Goods, which makes the case I'm making. Buchanan has a response to the criteria which Craig, borrowing from Charles Murray (for whom I share Craig's respect and admiration), promulgates:

1. A public good must be non-exclusive, similar to second criteria above. A public good benefits everyone or anyone. Think national defense. Everyone benefits.
2. A public good can be consumed by one person without diminishing its availability to others.
3. A public good has a substantial neighborhood effect that is difficult to charge for. For example, roads benefit people that do not drive in that most goods delivered to retail outlets arrive by truck. Therefore, using general fund dollars to pay for roads can be a �public good.�

I remember reading this in Murray and thinking it wrong, and I want to set Craig on a path awy from Murray in this regard.

One major problem Craig has, and the point I was trying to make (too quickly) with David Strom was that exclusivity and divisiblility isn't an either/or proposition. Buchanan offers this graph to explain the spectra of public goods that are available. Along the vertical axis is divisibility, and along the horizontal axis is the size of the group negotiating for provision of the good. A very large group negotiating over a good that is indivisible -- the second of the Murray/Westover criteria -- would be characterized as a pure public good, and near the point 0' on the graph. A pure private good would be near 0 -- there's a great deal of divisibility and excludability (the latter a statement about technology more than anything else) and the interacting group is quite small.

I used a rather silly example on the radio of liking David Strom's shirt as an example of an externality. It certainly is (though think: Mini-skirt) and I certainly could negotiate with him to wear the shirt more often in return for some small payment. Does that externality make it a public good? But doesn't the shirt generate a "neighborhood effect" as Craig lists as criterion 3? What's the size of the neighborhood needed to make the case to collectively provide the good? Should there be a "Strom shirt district" consisting of those in close enough contact with David that David could tax to get enough money to get him to wear that particular shirt?

There are other, less silly examples though, as characterized by points 2, 3, and 4 on the graph. Buchanan describes them:

(2) Partially divisible goods and services, with interactions limited to groups of critically small size For a good or service that may be classified in this way, there must be some substitutability among consumption units, as among separate persons, but this is not one-for-one. If the total supply available to the group is fixed, the increase in consumption by one person will reduce the amount available to some other person, or persons, but not precisely by one unit, as in the purely private-good case, and not by zero, as in the purely public-good case. The "nonprivateness" extends, however, only over a relatively small number of persons. As the group size extends beyond these limits, all publicness elements vanish.

Examples of goods and services falling in this classification are those that involve small-number externalities. Fire extinguishers may be an illustration. A transfer of a fire extinguisher to my neighbor does not reduce my own fire protection from the extinguisher to zero, as would be the case with a purely private or divisible good or service. My neighbor's possession of the extinguisher continues to reduce somewhat the probability of fire damage to my property. However, this interaction is limited in range. The transfer of a fire extinguisher to someone who lives three miles from my house does reduce my own benefits from that extinguisher to zero, in which case the exchange becomes equivalent to that of a purely private good.

(3) Partially divisible goods and services, with interactions extending over groups of critically large size This category includes the large-number externalities, or Pigovian externalities. There are both publicness and privateness elements in a good or service, but the publicness or indivisibility elements extend over a group that is critically large in size. An example is inoculation against communicable disease. The securing of a shot provides me with some privately divisible benefits but, also, it provides some benefits to all other potentially exposed persons in a large group. By comparison with the small-number interaction, in this instance many persons are effectively my neighbors. As we shall demonstrate later, the organizational-institutional differences between goods and services falling in (2) and (3) may be significant.

(4) Fully indivisible goods and services, but with interaction limited to groups of critically small size This includes those goods and services that are characterized by the fact that there can be no increase or decrease in the quantity available for one person independently, so long as we are limited to groups of small size. Outside the common-sharing group, however, this pure publicness does not hold, and among separate small groups there may be no publicness elements at all.

Examples for this category are drawn from club-like arrangements, which provide the organizational norm for this set of goods and services. Swimming pools may be mentioned. The single pool may be equally available to all members of the swimming club, provided only that the size of the membership is limited.

So where does light rail go? It is certainly not a pure public good, and limiting the size of people on the train is critical to prevent overcrowding.

The purpose of broadening the classifications, Buchanan argues, is not to create a five-part rule or a ten-part rule, but that one simply can't use characteristics of goods to describe their publicness.

Any positive approach to this question must proceed on a case-by-case basis and provisional conclusions reached only after careful comparison of institutional alternatives in the broadest sense. The descriptive characteristics of a good or service, the technology of common-sharing and the range of such sharing, are important determinants of organizational efficiency. Care should be taken, however, not to presume that these characteristics, taken alone, allow a priori judgments to be made. The pound of ceteris paribus must be used with caution here, since other things are not at all likely to remain equal over the institutional variants that may be examined. The predicted working properties of the institutional structures, imposed as constraints on individual behavior, must be evaluated.

That is to say, the decision to provide some goods collectively because "market failure" would lead to non-optimal provision of the goods cannot be decided a priori. You must compare market failure to government failure. We don't decide to let David set the number of times he wears his colorful summer shirt because the externalities are de minimus. We do so because the optimal amount is closer met (at lower cost) by private market decisionmaking than collective decisionmaking. If a set of private market institutions would underprovide light rail by 50% (however you would measure this) but public provision leads to 100% oversupply, does it matter too much to the discussion whether light rail is category 2 or 3, or 5? Wouldn't that be the right way to handle the question of what is a "public health" problem, ratherthan the list of problem characteristics Craig offers?

As Stubblebine often said in his class, "a thing is neither good nor bad save the alternatives make it so." The argument over the alternatives isn't about the characteristics of goods, but the choice of institutions.