3

EXTERNALITIES

AND PUBLIC GOODS

IT WAS ONCE a commonplace among economists that market institutions could not be relied on to generate Pareto optimal outcomes in the presence of externalities and public goods. Richard Musgrave referred to externalities as among those conditions "where the forces of the market cannot secure optimal results," and to public goods as a condition "where the market mechanism fails altogether." ¹

For most neoclassical theorists the conclusion was obvious: While markets are ideal institutions for allocating resources in the case of "normal" goods, some "political" mechanism must be sought for determining allocation of resources in "abnormal" cases of externalities and public goods. Musgrave put it succinctly:

Since the market mechanism fails to reveal consumer preferences in social wants, it may be asked what mechanism there is by which the government can determine the extent to which resources should be released for the satisfaction of such wants .... A political process must be substituted for the market mechanism. ²

The purpose of this chapter is to:

1. Summarize the state of public finance theory as of 1970

2. Call attention to the work of the most noteworthy previous critic of the neoclassical view and treatment of externalities and public goods, E. K. Hunt, which has been, regrettably, almost totally ignored

3. Situate recent work on "incentive compatible mechanisms" in the context of the traditional theory of public finance

4. Explain how, in our view, the traditional paradigm served to delay rather than facilitate progress in this area

3.1 Theory of Public Finance as of 1970

1. In the presence of externalities and public goods competitive market equilibria could not be expected to yield socially efficient resource allocations. This was due to "special" characteristics of externalities and public goods called "nonexcludability" and "market thinness," or what was more commonly called the "free rider problem."

2. This was not to say that in the presence of externalities and public goods Pareto optimal allocations did not exist. Indeed, as Samuelson showed, if one knew the preferences and technological capabilities of all agents in an economy, Pareto optimal allocations could be calculated even in the presence of externalities and public goods. ³

3. Though there was disagreement regarding the best course of action to be taken concerning externalities and public goods, there was a general consensus that no known mechanism that might be substituted for the market could be relied on to generate efficient allocations. Not only were markets combined with "voluntary associations" judged predictably inefficient, all proposed alternatives to the market mechanism and "voluntary" initiatives were also considered demonstrably imperfect. Sentiment was strong that a procedure that "solved" the "free rider problem" was theoretically impossible, and the best that could ever be hoped for was maximum amelioration.

4. The preceding conclusions were not regarded as particularly disheartening because most goods were viewed as private goods. Externalities and public goods, while troublesome, were exceptions to the rule.

We quote Richard Musgrave, a senior spokesperson in the field, to demonstrate that these are not our conclusions alone. Regarding the inefficiency of markets and "voluntary" initiatives, Musgrave said, "social wants are those wants satisfied by services that must be consumed in equal amounts by all. People who do not pay for the services cannot be excluded from the benefits that result; and since they cannot be excluded from the benefits, they will not engage in voluntary payments. Hence, the market cannot satisfy such wants.' ⁴

While sympathetic to "solutions" according to benefit received, Musgrave admitted, "the benefit approach has the advantage of tying the expenditure and tax sides of the budget together and relating both to individual preferences. Yet it fails in that it provides no way of ascertaining the true benefits." ⁵ But according to Musgrave, "the ability-to-pay approach to the theory of taxation proves even less satisfactory. It gives no clear-cut principles of tax distribution and fails altogether to answer the problem of expenditure determination. Later versions, categorizing the budget as a welfare plan, introduced the expenditure side of the budget and have the merit of emphasizing the planning aspects. However, these versions do not show how to determine the social preferences, without which determination the requirements of the welfare approach are an empty shell." ⁶

Musgrave spelled out the common wisdom regarding problems with various voting systems in detail. "In evaluating various voting systems, we must ask ourselves whether they will give a solution that comes close to the utility frontier, and what principles underlie the choice between various points on the frontier. In comparing the rationale of various systems, we can hardly avoid reasoning as if interpersonal utility comparisons can be made, the difference between the old and the new welfare economics of this type being less than is frequently supposed. Disregarding considerations of strategy, much is to be said in favor of the more sensitive ordering resulting from a system of point or plurality voting. Allowing for considerations of strategy, results of the cruder system of majority voting may be the better choice ... [But] a voting mechanism, together with a compulsory application of the budget plan thus decided upon ... [and the] necessity for compulsory application of a general tax formula means that the resulting solution will not be optimal." ⁷ And regarding the solution proposed by Tiebout ⁸ of having people "vote with their feet," Musgrave aptly observed, "the possibility of moving to other communities establishes something equivalent to a market mechanism in local finance, [but] the determination of social wants within any one community remains a problem in social wants." ⁹ In other words, Tiebout's scheme avoided rather than addressed the fundamental problems of public finance.

All of which led Musgrave, and many others to conclude that while "there is a wide array of situations where the market mechanism involves varying degrees of inefficiency in resource allocation-inefficiencies that arise collateral to the satisfaction of private wants-nevertheless, the satisfaction of such wants in most cases is best left to the market." ¹⁰

3.2 Previous Treatments

3.2.1 E. K. Hunt

No criticism of neoclassical welfare theory is more direct than E. K. Hunt's. Yet no criticism is so little known or has fewer echoing voices. ¹¹

The Achilles heel of welfare economics is its treatment of externalities....When reference is made to externalities, one usually takes as a typical example an upwind factory that emits large quantities of sulfur oxides and particulate matter inducing rising probabilities of emphysema, lung cancer, and other respiratory diseases to residents downwind, or a strip-mining operation that leaves an irreparable aesthetic scar on the countryside. The fact is, however, that most of the millions of acts of production and consumption in which we daily engage involve externalities. In a market economy any action of one individual or enterprise which induces pleasure or pain to any other individual or enterprise and is under priced by a market constitutes an externality. Since the vast majority of productive and consumptive acts are social, i.e., to some degree they involve more than one person, it follows that they will involve externalities. Our table manners in a restaurant, the general appearance of our house, our yard or our person, our personal hygiene, the route we pick for a joy ride, the time of day we mow our lawn, or nearly any one of the thousands of ordinary daily acts, all affect, to some degree, the pleasures or happiness of others. The fact is ... externalities are totally pervasive .... Only the most extreme bourgeois individualism could have resulted in an economic theory that assumed otherwise [emphasis added]. ¹²

From which Hunt concludes "the theory's absolute inability to handle pervasive externalities should more than suffice to convince any reasonable person of its utter irrelevance." ¹³

Hunt's argument is frighteningly direct. The implicit assumption underlying traditional welfare theory that "private goods" are the rule and "public goods" the exception is exactly backward. The presumption should be that a good has "social" aspects until proven otherwise. But the neoclassical assumption is seldom made explicit. And it rarely occurs to anyone to try to test it empirically, which only makes it more difficult to challenge. Hunt is brushed off as a Don Quixote tilting at windmills largely because the assumption is well hidden in the traditional paradigm. Accepting the "external effect exceptionality assumption" is part of the ideological leap of faith underlying acceptance of the whole traditional paradigm.

Hunt's first point concerning the relative prevalence of social aspects of economic actions is as simple as it is significant. His second observation is equally important but more subtle. Traditional welfare theory "simply takes the externalities, for which property rights and markets are to be established, as somehow metaphysically given and fixed. In ignoring the relational aspects of social life their theory ignores the fact that individuals can create externalities almost at will." ¹⁴ Hunt's explanation is as eloquent as it is penetrating and begs to be quoted in full:

If we assume the maximizing economic man of bourgeois economics, and if we assume the government establishes property rights and markets for these rights whenever an external diseconomy is discovered [the preferred "solution" of the conservative and increasingly dominant trend within the field of public finance], then each man will soon discover that through contrivance he can impose external diseconornies on other men, knowing that the bargaining within the new market that will be established will surely make him better off. The more significant the social cost imposed upon his neighbor, the greater will be his reward in the bargaining process. It follows from the orthodox assumption of maximizing man that each man will create a maximum of social costs which he can impose on others. D'Arge and I have labeled this process "the invisible foot" of the laissez faire ... market place. The "invisible foot" ensures us that in a free-market ... economy each person pursuing only his own good will automatically, and most efficiently, do his part in maximizing the general public misery.

To see why this principle has some validity, note that a self-oriented individual will maximize the value, to him, of participating in organized markets and creating nonmarket transactions. Taking this production possibility set for creating external diseconomies, he will select only those with a higher return than he could earn by engaging in market transactions. But by so doing, he will maximize the cost to others in that his gain is someone else's loss. All individuals acting independently to maximize the cost imposed on others will yield a maximum of these costs or payments to society, that is, by selecting only highly productive external effects. The recipient of contrived or inadvertent external diseconomies will undertake defensive expenditures or pay bribes until the usual marginal conditions of efficiency are fulfilled. Thus, the recipient's cost will be minimized for each external diseconomy, and an efficient pattern of external effects will emerge.

But if external diseconomies, in terms of value to the generator, are maximized in the society and if they are efficiently contended with by recipients, then we have a mirror image of consumption theory and Pareto efficiency. That is, instead of allocation of a good to its highest value use with its production costs minimized, we have allocation of a bad (external diseconomy) to its most costly impact, with the impact being minimized in terms of recipient cost as well as production costs. The economy, of course, is efficient but efficient only in providing misery.

To paraphrase a well-known precursor of this theory:

Every individual necessarily labors to render the annual external costs of the society as great as he can. He generally, indeed, neither intends to promote the public misery nor knows how much he is promoting it. He intends only his own gain, and he is in this, as in many other cases, led by an invisible foot to promote an end which was no part of his intention. Nor is it any better for society that it was no part of it. By pursuing his own interest he frequently promotes social misery more effectually than when he really intends to promote it. ¹⁵

We will pursue Hunt's lead regarding external effects in building our alternative paradigm in chapter 5 and analyzing markets in chapter 7. But in some respects we find his criticisms of traditional welfare theory overdrawn or wide of the mark. Since he is not so lonely in areas where, in our opinion, he errs, as in areas where he hits the mark, these points are worth mentioning.

We find no fault with the concept of Pareto optimality as a formalization of the intuitive notion of social efficiency. Therefore, we reject Hunt's charge that "the neoclassical notion of market efficiency encountered in every branch of applied economics, as well as the bourgeois notion of rational prices encountered in so many discussions of the role of the market in socialist society, have absolutely no meaning whatsoever other than the belief that a free competitive market will tend toward a Pareto optimal situation in which by definition, resources are said to be efficiently allocated and prices are said to be rational. There is no further criterion or justification for using the words efficient and rational than the assertion that the particular resource allocation and price structure obtaining in a free competitive market will have some connection with that envisioned in the analysis of Pareto optimality ." ¹⁶

In our view no further justification is needed for deeming an outcome efficient" than that it is Pareto optimal since the concept of Pareto optimality captures the useful aspects of the notion of social efficiency very well. ¹⁷ We have already stipulated that the concept of Pareto optimality has proved exceptionally helpful in formulating difficult propositions and conclusions concerning efficiency with greater precision and rigor. And if a set of prices accurately summarized the relative social costs and benefits of different goods, we would not hesitate to call those prices "rational."' ¹⁸

Hunt also voices another claim not uncommon among radical economists. "It is ... obvious that this theory is applicable only where individual preferences or tastes do not change over time." ¹⁹ We certainly must dissociate ourselves from this view since in chapter 6 we reformulate the three fundamental theorems of welfare economics in a formal model in which individuals' preferences are endogenous and change over time and thereby lay this intuitively appealing criticism to rest. We also find the derivative criticism that traditional welfare theory is inherently static overdrawn as well.

In the practical world of development policy all the problems Hunt catalogs above do exist, and it is critical to recognize these ambiguities. But in theory, and we emphasize "in theory," it is perfectly clear what is to be maximized for any individual and how "new" individuals must be treated. What is to be maximized for any individual is total utility over his or her entire life. This of course requires an individual well-being function that evaluates the relative contributions to well-being of utility in different time periods. But this is analytically no different from requiring individual preference orderings that evaluate the relative contributions to an individual's well-being of different consumption activities during the same time period. "New" individuals can appear on the scene at any time, as long as they are similarly equipped. Viewed in this way, the problem of "interpersonal" tradeoffs in well-being between present and future generations is analytically no different than the problem of "interpersonal comparisons" and tradeoffs between individuals living at the same time. In neither case does the concept of Pareto optimality have anything to tell us. But this is not a new, or different, problem, as Hunt implies.

Hunt is also wrong in thinking that the so-called "Cambridge Controversy" has implications for neoclassical welfare theory as applied to "Debreuvian" type models. Among conditions Hunt lists as necessary to guarantee Pareto optimality in private enterprise market economies are: "(c) homogeneous inputs and outputs each divisible into units of any desired size," and "(f) production functions being of smooth curvature, not having increasing returns to scale, and having diminishing marginal rates of substitution along any isoquant curve." ²¹ He then argues that "assumption (c) on homogeneity of inputs (particularly capital) and (f) about properly behaved production functions, have both been definitively shown to be untenable by the recent Cambridge capital controversy." ²²

This is a misinterpretation of the so-called "Cambridge Controversy." In fact, the Cambridge, England, side of the Cambridge debate successfully challenges the theoretical coherence of the neoclassical concept of "capital" as a homogeneous input that can be used as an argument along with homogeneous "labor" in an "aggregate" production function that is subsequently used to derive the rate of profit in the economy. But homogeneous "capital" is not the kind of homogeneous input relevant to welfare theory. In Debreuvian models the inputs that must be "homogeneous" are particular types of machinery. If two machines thought to be "homogeneous" turned out not to be, a Debreuvian model simply defines two inputs (and two prices) instead of one. Put differently, there are no aggregate variables in Debreuvian models, so any difficulties in defining the aggregate variable, "capital," are of no consequence.

Similarly, in Debreuvian models production functions are individual production functions, or more general yet, production possibility sets. The theoretical coherence and properties of an aggregate production function are irrelevant to a Debreuvian model because no such concept ever appears. In any case, the properties of production functions to which Hunt refers in (f) are relevant to the existence of general equilibria in private enterprise competitive market economies, not to whether or not any equilibria that do exist would be Pareto optimal in any case. The smoothness property of production functions (by which we assume Hunt means infinite input substitutability) is not even necessary for existence of equilibria once one substitutes convex production possibility sets for production functions.

Finally, Hunt and many other critics make much of what is called "the theory of the second best." We in no way dispute the theory as it was admirably demonstrated by Lipsey and Lancaster, ²³ but we believe it is often misinterpreted and misapplied.

In Hunt's interpretation "according to the theory of the second best, policies designed to remedy only some and not all of the defects (since simultaneously remedying all would obviously be impossible) will often result in effects diametrically opposed to those envisioned by the authors of these policies." ²⁴ In our view, a more careful interpretation would be: In a context of multidefects, there is no reason to suppose that implementing corrective policies in a random order will not move the economy farther away from optimality rather than closer at each step, up to the last. Before comparing these interpretations, two points need clarifying:

1. The theory of the second best cannot be a direct criticism of neoclassical welfare theory. A welfare theory is a system for ranking outcomes. The theory of the second best says nothing about the particular (incomplete) system of ranking outcomes that we call neoclassical welfare theory except by using it to implicitly accept it. Instead the theory of the second best warns that in a market economy a policy that would move us from a state neoclassical theory it ranks lower to a state it ranks higher if there were only one defect, might instead

2. Nor can the theory of the second best be seen as relevant to only market economies. In a centrally planned economy, if two or more goods were misvalued, correcting the evaluation of one and not the others might well move the economy farther away from Pareto optimality . We might say a "general 'general theory of the second best"' applies to economies that are highly integrated technologically (indecomposable) irrespective of the economic institutions chosen for administering them.

To summarize while Lipsey and Lancaster demonstrate that piecemeal policy correctives might be counterproductive according to the particular system of ranking we call neoclassical welfare theory in the particular kind of economy called competitive capitalism, their result is most often not sensitive to what system of ranking and what economic institutions one chooses. ²⁵ Instead, the "problem" the theory warns against is the result of technological and social possibilities that are highly integrated.

With all this said, what is the point of the theory of the second best, and how does our interpretation differ from that of Hunt and others? Hunt interprets the theory to mean that the numerous inefficiencies of market economies are hopelessly insensitive to policy correctives as calculated by a partial equilibrium application of neoclassical welfare theory. In short, Hunt feels that the neoclassical claim that market imperfections and failures are correctable via tax, subsidy, and antitrust policies is, in fact, a delusion in a world of multidefects and piecemeal policy making. Interestingly enough, conservatives interpret the theory of the second best as
proving there is no point in trying to implement policy correctives at all for what they see as the few demonstrable cases of market inefficiency.

The key to piecemeal policy correctives is selecting the order of implementation with a little common sense. Imagine an economy that produces two intermediate products and one final good. Suppose the final good is produced by a single firm and sold to hundreds of millions of consumers, one intermediate good is produced by millions of enterprises and the other is produced by fifty firms, and all intermediate goods are purchased as inputs by the firm making the final good. If the corrective policy is antitrust, two industries need to be busted to create competitive conditions everywhere. If we bust piecemeal and select where to begin by flipping a coin, we are obviously as likely to make matters worse as better after step one. This is what the theory of the second best tells us.

But the theory does not say we cannot break up the monopoly/monopsony first and take care of the mildly noncompetitive intermediate industry second. It does not say there is not a second best that falls short of the "best" option of implementing both antitrust policies, but improves upon doing nothing at all. It just cautions that there may be a fourth best that is worse than the third best of standing pat, and that moving without any caution whatsoever is as likely to move us from third to fourth as third to second.

Once this is understood, the practical question becomes: How obvious is the order in which piecemeal corrective policies should be implemented likely to be if we want to achieve the practical goal of improving matters at each step? Our answer is: Frequently, fairly obvious, although technical and social interrelatedness can complicate matters. But we repeat, in any case, these are primarily practical matters equally problematical for all reasonable welfare theories and all highly integrated economies. ²⁶

3.2.2 Theodore Groves and Company

As late as 1972, Milleron devoted only two of fifty-eight dense pages in his excellent article, "Theory of Value with Public Goods: A Survey Article," to proposed "solutions." After fifty-six pages reviewing work on the meaning of Pareto optimality , core, and equilibrium in an economy with public goods, Milleron needed only two pages to review mechanisms claiming optimality properties and conclude that none were satisfactory. He proved quite prophetic, however, in remarking, "One must remember that the investigations in this field are just beginning and that many new developments are to be expected in the next few years." ²⁷

Seven years later the Review of Economic Studies sponsored a symposium on "incentive compatibility" devoted to "solution" mechanisms to the "free rider problem." The ten new papers published as part of the symposium, together with the path-breaking papers of the previous few years, required an entire essay by Peter Hammond to introduce them. There was suddenly a torrent of sound where before there had been silence. As Hammond explained:

In other words, suddenly a number of "mechanisms" with claims of "solving" what had been considered the "irresolvable" problem of efficient provision of public goods appeared on the scene. In his contribution to the symposium one of those most responsible for the intellectual breakthrough, Theodore Groves, gave the following interpretation:

While Vickrey had described examples of demand revealing mechanisms in a paper published in 1961, ³⁰ the idea languished until Groves ³¹ and Clarke ³² independently picked up the theme in the early 1970s. We attribute the fact that nobody recognized the importance of Vickrey's discovery for so long partly to the influence of the traditional welfare paradigm we have been criticizing. That many continue in ignorance of the significance of incentive compatible mechanisms today is a further effect of a traditional paradigm that minimizes the importance of the problem and steers one away from solutions. It is of particular interest that the full flurry of creative activity that succeeded by the close of the decade in substantially elaborating the set of incentive compatible mechanisms only occurred after Hurwicz had taken a long step toward redefining critical parts of the traditional paradigm in papers published in 1972 and 1973. ³³

The search for "demand-revealing" or "incentive-compatible" mechanisms departed from the assumption that efficient provision of public goods would be possible if only consumers could be induced to reveal their true preferences for public goods; but the problem was that rational consumers would presumably consider how their "messages" about the desirability of public goods affected not only the quantity of public goods the government would end up supplying, but the taxes they would be assessed to pay for them. If reporting high benefits could be expected to raise one's assessment, false reporting and "free riding" would make sense. The key idea was to sever the link between reporting high preferences and receiving high assessments in order to eliminate the incentive to report untruthfully.

Dreze and de la Vallee Poussin ³⁴ drastically weakened the link by distinguishing between assessments to finance the cost of producing the total quantity of a public good and assessments to finance the cost of producing a marginal unit of a public good. If I am asked my willingness to pay for another unit of a public good, and my assessment for the entire amount of the public good that is provided is a positive function of my response, I have a strong incentive to disclaim interest. But if all that is at stake when I report my willingness to pay for another unit of a public good is how much I will be assessed for provision of an additional unit, the incentive to underreport is drastically reduced. Effectively, my answer can have a significant impact on my budget in the first case but not in the second. In the authors' own words:

Clark³⁶ incorporated this idea into what has come to be called the "pivot mechanism", which Groves and Loeb ³⁷ generalized. The "pivot mechanism" charges each agent for the sum reported willingness of all others to pay for substituting the public good supply that would have resulted had the agent reported complete indifference to all supplies of public goods for the public good supply resulting from the preferences the agent actually reported. In other words, if by expressing a preference for public goods an individual changes what others have to live with, the individual is assessed the sum inconvenience others claim.

The formulation that most directly revolutionized the field is due to Groves and Ledyard. In Groves and Ledyard's demand-revealing mechanism, "consumer i is simply assessed his proportional cost share of y [the package of public goods provided] minus the reported consumers' surplus of the others plus possibly some lump sum transfer that is independent of his message." ³⁸

Under the incentive system established by this mechanism, when we ask, How can I reduce my tax bill? the answer is, To all intents and purposes, not at all. ³⁹ And when we ask, How can I best affect the quantities of different public goods I (along with all others) will consume? the answer is, By reporting my willingness to pay truthfully. ⁴⁰

Once consumers report their preferences truthfully the government has only to produce the quantity of each public good at which the total marginal willingness to pay, or marginal social benefit, equals marginal cost.

Lest our justifiable enthusiasm be interpreted to imply all problems in the field have been solved, we should note that several sobering "impossibility" theorems have already been demonstrated. A "dominant strategy equilibrium" is said to exist in a game if each player has a strategy that is best regardless of the strategy chosen by any other player. Gibbard and Satterthwaith ⁴¹ have proven that no nondictatorial mechanisms exist to yield dominant strategy equilibria for all conceivable preference profiles. Green, Laffont, and Hurwicz ⁴² have demonstrated that no mechanism exists for all quasi-linear preference profiles (roughly preferences in which collective choices can be compared quantitatively in terms of a "money" equivalent that displays no "income effects") that yields an efficient dominant strategy equilibrium. Moreover, Groves and Ledyard in 1977 acknowledged that demand-revealing mechanisms fail to generate totally efficient allocations because they cannot guarantee a balanced government budget and compliance with Walras' Law.

3.3 Influence of the Traditional Paradigm

The key concept--ordinal preferences that are not interpersonally comparable over commodity space-is very useful for individual commodity spaces that are separate and exclusive. In the private good context no less vague (more restrictive) conception of evaluations proved necessary. But thinking with the concept of ordinal preference orderings that are not interpersonally comparable retarded thinking about efficient ways of allocating resources when all individuals have the same commodity space and nothing is, therefore, separate and exclusive. Instead, a less vague (more restrictive) conception of evaluation was necessary, namely, that individuals have preferences over "public goods commodity space" and for a moneylike good that could be used for "compensatory transfers." But over one hundred years of "progress" in refining conceptions of preference orderings over "individual commodity space" and as "interpersonally comparable," in other words, in the opposite direction, stood in the way.

It proved necessary to define concepts that permitted talking about communication of values (information sets and message spaces); concepts that permitted thinking about negotiation and back scratching (iterative mechanisms and assessments based on how others are affected by one's choices); concepts that saw rational individual behavior as system defined and therefore malleable; and concepts that recognized incentives are defined by the rules of the game people are forced to play. But none of these concepts were forthcoming from the traditional paradigm, and all this was necessary before the welfare theorists' "game" could be recognized for what it should be: "Find the rules that create the incentive to speak truly."

This is hardly an incentive-compatible mechanism for eliciting maximum creative efforts from economic theorists in the publish or perish world of modern academia! The welfare establishment had to do a drastic turnabout before journal-page points and tenure could be earned by dreaming up "two, three, many new creative economic 'games."'

In hindsight it is perfectly obvious: the solution set you would expect to get from the traditional paradigm is exactly what we got-policy prescriptions for resolving external effects by assigning additional property rights and creating new markets-which is to say, no solution at all. A paradigm was needed that would "analyze the social forces underlying the incentive structure" to encourage a mind set "to alter the incentive system itself ... in order to move toward a better efficiency of the economy." ⁴⁴