The Problem of Second Best:

Are Partial Equilibrium and Third-Best Analyses Solutions?









Joseph Rebello

Advanced Antitrust

Judge Posner

May 15, 2002


I. Introduction

            The problem of second best poses a nagging and interesting dilemma for antitrust analysis. It has been dismissed as a valid but unworkable criticism of traditional antitrust doctrine. More recently, however, differing approaches on how to incorporate second best analysis into the fashioning of antitrust remedies have been advanced.

            The problem of second best deals with the question of whether interventions directed at specific market imperfections can improve overall social welfare. According to the theory of second best, correcting specific market imperfections while leaving others untouched will not necessarily improve social welfare. This leads to the conundrum – if it is unknown whether welfare will be improved, should courts interfere at all? The problem is alleged not to be so hopeless, however, as there may be situations in which it is possible to estimate whether an intervention is at least likely to improve welfare.

            The theory of second best in general states that in a system where conditions are such that a Pareto optimum exists, if one condition is changed so that it is no longer at its optimum state, then to reach a second best optimum (because the first best optimum cannot be reached), all the other conditions must be changed from their original first best optimum states.

One exposition of the theory of second best, by Professors Lipsey and Lancaster[i], is an analysis that looks at a general equilibrium system in which all conditions are such that a Pareto optimum exists. Then one condition is changed from its optimum value so that the Pareto optimum no longer exists. The changes in all other conditions necessary to reach a second best optimum are then examined. A second approach, by Professor Meade[ii], looks at a system not at an initial Pareto optimum where there are several conditions that are not at their optimum values. The effects of changes in these conditions are then examined to see whether an increase or decrease in welfare follows.

The theory of second best has general implications for the economic analysis of law. When legal rules are evaluated from the standpoint of whether or not they promote social welfare, second best concerns are usually ignored. However, according to the theory, improvements to some but not all conditions in a non-optimum state will not necessarily increase welfare. Some extent of second best analysis would appear to be necessary with even basic prescriptions for legal rules.

As far as antitrust law is concerned, the theory of second best can color both the range of market imperfections examined as well as the remedies that attempt to correct them. If the goal of the antitrust statutes is to increase social welfare, then a second best analysis of antitrust remedies would appear to be required. Of course, a second best analysis can be undertaken only if it is practicable. There are different ideas about the role that second best analysis could play, ranging from incorporation into judicially crafted remedies to broad legislative mandates. If a second best analysis is impracticable, there is also the question of whether an antitrust remedy should even be specified, given the uncertainty about its effects on social welfare.

The problem with second best analysis is that it is very difficult to execute in a comprehensive fashion. It may be impossible to consider the effects on all conditions necessary for a Pareto optimum to hold. The important question is whether a more limited inquiry, such as a partial equilibrium analysis or a third-best-allocative-efficiency analysis, will still yield enough clues to estimate whether a market intervention will increase or decrease welfare.


II. The Theory of Second Best

            The first approach to conceptualize the problem of second best is a formal exposition by Professors Lipsey and Lancaster – in general terms, the theory of second best states that in a system at a Pareto equilibrium, if a constraint is introduced that prevents the attainment of one of the Pareto conditions, all the other conditions must be shifted from their Pareto equilibrium positions to reach another optimum. This optimum is a second best optimum because, by definition, the constraint that has been introduced prevents the attainment of the first best optimum.[iii]

            It is important to note that the system at a Pareto equilibrium need not be free of all constraints. That is, a set of constraints may be assumed to exist and it may be not be easy to remove them. A tax or some other government regulation are examples of such constraints.[iv] With the constraints that are assumed to be invariant, however, there exists a Pareto optimum such that all other conditions are at their optimum points. The one additional constraint described above is thus applied to the subset of conditions that are at their Pareto optimum points given the invariant constraints.[v] However, there is no real difference between the constraints that are assumed to exist in the system and the additional constraint that is introduced to disturb the optimum – it is just another constraint.[vi]

            The second best optimum then is just a new Pareto optimum given one additional constraint. For instance, take a system at a Pareto optimum with ten conditions. Suppose two of these conditions are subject to constraints such that they are no longer at their optimum points. If the remaining eight conditions are then optimized, the new state could be termed a second best optimum in relation to the state where nine or all ten of the conditions are optimized. Or, if the two constraints are assumed to be invariant, the state with eight conditions optimized might be called simply the Pareto optimum.

In the example above, when the second best optimum is reached, the values of the eight conditions must all be different from what they were when the first best optimum existed. This principle must work in reverse as well – if, instead of imposing an additional constraint, an existing constraint is lifted from a system at a Pareto optimum, then the values of all optimum conditions must change to reach a new optimum (this is of course not a second best optimum). Thus, if a tax is repealed in a system at a Pareto optimum, then all conditions must change, or else the system will not be at a Pareto optimum.

            In the problem posed as above with the introduction of one additional constraint into a system at Pareto equilibrium with given conditions, there will be one second best optimum. However, for the given system with its subset of conditions that are at their Pareto optimum points, there can be several second best optima. This is because any one or several of the given conditions could be subject to constraints, resulting in a different second best optimum for each combination of constraints.[vii]

            The second best optimum necessitates a departure from all the optimum conditions that were satisfied in the first best optimum. Unfortunately, in general, the direction and magnitude of the changes necessary are not known.[viii] Thus, in a system at Pareto equilibrium where a tax is levied on only one commodity but returned as a gift to the consumers (so that the only effect of the tax is to distort relative prices), the second best optimum would require a tax or subsidy on all other commodities.[ix] Nothing can be determined in general about which commodities are taxed or subsidized or to what extent – but no commodity can go untaxed or unsubsidized.

            Because of this indeterminacy, different states in which no condition is at its optimum point cannot be meaningfully compared with each other.[x] If all that can be generally known is that all conditions must be changed in some way, but not in what way or how much they should change, then one state cannot be said to be better than the other – this is the “important negative corollary that there is no a priori way to judge as between various situations in which some of the Paretian optimum conditions are fulfilled while others are not.”[xi] Thus, for example, a state where every industry is a monopoly to the same extent is not necessarily better than one in which there are differing degrees of monopoly power in all the industries.[xii] This indeterminacy creates a problem for structuring antitrust remedies – not only does the Pareto optimum seem unattainable, but so do any improvements, as long as it is impossible to tell whether there is in fact an overall improvement.


            The analysis above explored the theory of second best by imposing one additional constraint on a system where all other pertinent conditions were at their optimum values. A second approach to the theory of second best by Professor Meade is to examine a system where there are numerous departures from the optimum conditions.[xiii] This system is not at an initial Pareto optimum. The question then is how to evaluate the effect of a change in one or more of the conditions.

            For example, consider an economy with ten industries, none of which are operating at the optimum points at which they would be if the system were at a Pareto optimum. Thus, in every industry, marginal value (or proportionally, price) is not equal to marginal cost. This divergence could be the result of an absence of perfect competition, of taxes or subsidies, of other government intervention, or of a difference between the social and private net product of a factor.[xiv] One condition is then changed – perhaps an intervention that results in reducing the divergence between price and marginal cost in one industry. This change can affect the amounts of divergence in the other nine industries depending on the nature of the interrelationships between the industries in the economy. The amount of divergence in each industry will determine the extent to which each industry is allocatively inefficient – the extent to which the sum of consumer and producer surplus is not maximized as it would be under perfect competition. Of course, allocative efficiency is a measure of social welfare that ignores distributional issues. Even if allocative efficiency is decreased overall, then, the shifting of income between and among consumers and producers, depending on the distributional weights attached to the various groups, might not result in a decrease in social welfare.

Determining the effect on welfare of a change in one condition thus requires combining the change in overall allocative efficiency with the change in distribution of income between groups. Distributional issues, however, are trickier to resolve than questions of efficiency because they involve questioning the initial endowment of resources within groups.[xv] Nevertheless, if distributional issues can be put aside, it has been hypothesized that if the intervention is directed at the industry with the greatest divergence, then the policy will at least be likely to improve welfare.[xvi] This will only be the case, however, if improving the divergence in the one industry does not result in changes in other industries with negative effects that together outweigh the positive effects of the initial change. The greater the extent to which this hypothesis holds true, however, the easier it is to be confident in an aggressive antitrust policy.


The two approaches to the theory of second best discussed above do have a conflict. The Lipsey/Lancaster “negative corollary” which states that there is no a priori way to determine whether one non-optimum state is better than another non-optimum state[xvii] implies that “there are unlikely to be any simple sufficient conditions for an increase [in welfare] when a maximum cannot be obtained.”[xviii] Thus, a policy of “piecemeal” welfare economics that does not aim to correct all imperfections simultaneously is unhelpful.[xix] This reasoning seems sound, and yet Professor Meade’s approach appears to imply some amount of determinacy – that is, that the overall effect on welfare of a changed condition could be determined if the impacts on all other conditions could be calculated.[xx]

Professors Lipsey and Lancaster do not offer a proof of their “negative corollary” – only the general theorem, that every other condition must change to reach a second best optimum upon imposition of a constraint on a system at Pareto equilibrium, is proven.[xxi] They do, however, offer the following elaborations on the negative corollary: 1) a situation in which more of the optimum conditions are fulfilled is not necessarily better than one in which fewer are fulfilled and 2) a situation in which all departures from the optimum positions are of the same direction and magnitude is not necessarily better than one in which departures vary in direction and magnitude.[xxii] But does the negative corollary apply to the situation where the departures of the greatest magnitude are corrected? Is a situation, for example, where the four greatest departures from the optimum position are corrected better than a situation where the four least departures are corrected? Professors Lipsey and Lancaster do not address this question directly, and Professor Meade implies that the answer is likely to be yes.[xxiii] Furthermore, Professors Lipsey and Lancaster favorably acknowledge Professor Meade’s approach in their paper and state that “Meade’s [approach] is probably the appropriate one when considering problems of actual policy in a world where many imperfections exist and only a few can be removed at any one time.”[xxiv] The Lipsey/Lancaster approach is more fitting to study general principles of the theory of second best.[xxv]

            The resolution of this conflict is important because it determines the implications of second best theory for normative welfare economics. If the Lipsey/Lancaster negative corollary is taken in its strongest sense, then any prescriptions to improve market imperfections that fall short of attaining a first best or second best Pareto optimum are useless. This disheartening thought would appear to “call for the end of antitrust policy.”[xxvi] Does the negative corollary in its strongest sense, then, actually follow from the general theorem? An answer to this question would determine whether an antitrust policy directed at the most significant market imperfections can be said to reliably promote the increase of welfare. For now, this answer will be reserved so that the application of second best theory to antitrust policy may be introduced.


III. Second Best Theory and Antitrust Policy

            The integration of second best theory into antitrust policy has followed typically three different lines of thinking. First, there has been a call to basically ignore second best criticism of antitrust policy because of the unlikelihood of second best concerns, the unworkable nature of second best analysis, and the extraneous social costs of market imperfections such as monopoly. Second, the uncertainty of the effects on welfare brought up by the second best theory has motivated arguments to end an antitrust policy predicated on welfare-improving grounds. And third, a belief that the theory of second best does not preclude some determination of situations in which welfare can be increased has led to the proposal of methods such as a partial equilibrium analysis and third-best-allocative-efficiency analysis to address second best concerns for antitrust policy.


            The first line of thought uses several points to justify a dismissal of second best criticisms of antitrust policy. The first of these is that the theory of second best “does not address itself to the probability of the bad result, but states it merely as a possible outcome.”[xxvii] The “bad result” here is that correcting a market imperfection will have an overall negative effect on welfare because of ramifications elsewhere in the economy. It is true that the Lipsey/Lancaster approach does not deal with the probability of lowering welfare through a market intervention. But that is precisely the point of the theory – because, as a general matter, it is virtually impossible to predict with any specificity the changes required in the remaining conditions to reach a second best optimum, it is also not feasible to predict with any certainty the effect on welfare of an isolated change in a condition. To make such a prediction is possible only in very simple cases after making many assumptions about the conditions involved.[xxviii] The assumptions that are required preclude finding a second best optimum for most real cases of interest,[xxix] or determining the probability in most real cases that a “bad result” will occur.

            Another idea that is used to justify ignoring second best criticisms is that courts (or any other institution) are incapable of undertaking a second best analysis. For example, if second best analysis were employed in a price-fixing case, the amount of divergence (difference between price and marginal cost) would have to be calculated not only for the industry that is the focus of the case, but also for all industries that make substitutes or complements for the products involved and for all other industries that are linked by the flow of resources. The change in all divergences if the price-fixing is abolished must then be predicted, as well as if this leads to an increase or decrease in welfare.[xxx] This task does of course seem extremely daunting. However, it is simply the method used in Professor Meade’s approach to calculate the effect on welfare of a single market intervention.[xxxi] Whether it is impossible is unclear – many complicated cost-benefit analyses are undertaken that are arguably similar and the proponents of partial equilibrium analysis to address second best concerns would argue that the calculation is possible to a limited extent.

            Whether calculating the effect on welfare is a task that can be handled by the courts, however, is a different question. Simply because the calculation may be possible is not to say that courts are equipped to perform it, either in terms of resources or of the time frame required for a decision. After all, many antitrust cases deal with companies enjoying monopoly power due to innovation and since these companies often succumb to the next great innovation, antitrust remedies should be timely. It has been suggested instead that calculating effects on welfare is a task more suited to the legislature.[xxxii] Courts are limited to the arguments before them, create rules only incrementally, and generally are to consider only the local effects of a rule. Legislatures, on the other hand, can look at a broader context and have a mandate to be as far-reaching in scope as required. In addition, a far greater number of interests can be represented before a legislature.[xxxiii] Legislatures therefore have the characteristics that are required for the economy-wide inquiry called for in a second best analysis. This observation about institutional competence to perform a second best analysis is a good response to a rejection of second best theory on the basis that courts are unable to implement it. Nevertheless, the further criticism would undoubtedly be made that a second best analysis is too great an undertaking even for the legislature, due to its inherent complexity. Then the argument again would reduce to whether Professor Meade’s method of calculating the effect on welfare is feasible.

            Finally, second best criticism has been urged to be rejected on the grounds that there are other costs to leaving market imperfections uncorrected that could outweigh any negative second best effects. These costs are the social costs that accompany the maintenance and pursuit of monopoly power.[xxxiv] They include costs incurred by cartels in concealing their activity from enforcement agencies because cost-minimizing but more detectable cartel methods must be avoided, and costs incurred by reserving excess capacity because competition might break out quickly with an unstable cartel.[xxxv] Costs are also incurred in the acquisition of monopoly power because there is competition to be a monopolist.[xxxvi] This competition is for the potential monopoly profit that awaits – the transfer from consumer surplus to producer surplus, which is not a loss from an efficiency standpoint, that occurs upon the reduction of output from the competitive level. This is typical rent-seeking behavior and aspiring monopolists will incur costs to acquire monopoly power as long as the expected return from the monopoly profit is greater than the total costs each will incur in pursuit. An approximation of the cost of the competition for this transfer payment is the amount of the transfer payment itself[xxxvii]n firms competing for monopoly profit p, each with 1/n chance of winning, will each spend up to p/n, for a total of p. The amount of the transfer payment generally dwarfs the deadweight loss that results from the reduction to monopoly output. This provides a justification for antitrust regulation for which second best criticisms are irrelevant.[xxxviii]

             The social costs of monopolies are thus independent of any concerns about allocative efficiency, either in the monopolized market or in other affected markets. This argument in favor of antitrust regulation does not face a ready rebuttal purely from the perspective of second best theory. However, the notion that there are social costs incurred in acquiring or maintaining monopolies might be disputed based on the assertion that the concept of rent-seeking behavior is not entirely compatible with principles of wealth maximization.[xxxix] The point is that under the criterion of wealth maximizing efficiency, or Kaldor-Hicks efficiency, voluntary market transactions should be treated as wealth maximizing.[xl] Each party enters into a voluntary market transaction because each values what it receives more than the other party, creating welfare gains on both sides. Therefore, because the social costs of gaining and maintaining monopoly power are the result of such voluntary transactions – between, for example, companies and lobbyists and lawyers – it is incorrect to treat them as social costs.[xli]

There is a gut feeling, though, that the expenditures made in the name of monopoly might be better and more productively spent elsewhere by the company – for research and development, or just invested in the stock market. But if the company could earn a higher return on its outlays by investing in a project other than acquiring monopoly power, it would obviously do so. It is clear that on the other side of the transaction, from the lawyers’ and lobbyists’ point of view, there is no difference between working for a company seeking monopoly power and working for a company doing anything else. Basically, if any voluntary market transaction is deemed to be a social cost or wasteful instead of wealth maximizing, it is because the preferences behind those transactions are judged to be unworthy. And the claim is that such judgments should have no place within efficiency considerations.[xlii]

Nevertheless, the costs of acquiring and maintaining monopolies could also be said to be wasteful because they are incurred in furtherance of an activity that is prohibited because it is itself wasteful, due to the deadweight loss of monopoly. But if the deadweight loss is small,[xliii] a point that is contested[xliv] – then the costs of acquisition and maintenance have been deemed wasteful again only because a value judgment has been made. However, such value judgments are made. In the case of theft, for example, the amounts spent on burglar alarms are usually categorized as social costs.[xlv]

Whether monopoly is to be grouped with theft is a line-drawing question. If it is, then the expenditures made to acquire and maintain monopolies will be social costs that constitute an independent justification for antitrust regulation of monopolies. Ignoring second best implications of the regulation would then be appropriate because the social costs of monopoly would outweigh any negative second best effects.


             The second line of thought in the integration of second best theory into antitrust policy has been to accept the Lipsey/Lancaster negative corollary in the strong sense and focus on non-efficiency grounds for decision.[xlvi] Thus, because there is no indication that correcting deviations will increase welfare absent the simultaneous correction of all deviations, the usual antitrust efficiency analysis must be given up.[xlvii] Instead, a fairness standard or populist criteria such as the promotion of agrarian over industrial interests or of small entrepreneurs over big businesses could be utilized.[xlviii]

            Such a move appears somewhat regressive, but if welfare improvements short of maximization are indeed impossible, then using these alternate criteria are actually consistent both with efficiency and overall welfare concerns. They are consistent with efficiency concerns simply because there would be no way to incrementally specify regulations that definitely improve efficiency – the status quo would then be just as preferable. And the criteria are consistent with overall welfare concerns because, if efficiency cannot be improved, then welfare can be improved only by transfers made along distributional lines.

            In the present day, of course, the populist criteria mentioned above should probably not be used as distributional guidelines. The number of small farmers has declined dramatically as innovations in the long-run shift the industry from labor- to more capital-intensive methods of production.[xlix] Farming is now mainly a corporate activity,[l] and a distributional goal of keeping small family farmers employed will not be particularly sensible. A goal of protecting small entrepreneurs, too, is suspect. There are probably few advantages today in terms of the amount of innovation or quality of service – large corporations can provide both. Furthermore, the prevalence of pension fund investment in corporations offers wide and easy access to the gains from corporate enterprises. Nevertheless, if new criteria are called for, there will undoubtedly be many disadvantaged groups vying for a larger share of resources.


            Instead of dismissing second best criticism outright or taking it at its full face value, the third line of thinking in the integration of second best theory into antitrust policy takes a more moderate tack. Second best analysis would be utilized to check whether antitrust regulation would produce greater offsetting harms through second best effects – but the analysis would be limited in some manner, either in the range of the economy examined or in the types and amount of data to be collected.

With partial equilibrium analysis, the basis for a circumscribed use of the theory of second best is the idea, mentioned above, that the theory points only to the existence of negative second best effects, and not to their frequency.[li] And taking this a step farther, the frequency of second best effects is also claimed to be low, because it depends on interrelationships within the economy, which are generally weak and may be ignored. The economy can thus be partitioned and the only the strongest interactions need be subjected to second best analysis.[lii]

            This sets the stage for the partial equilibrium analysis, in which one market would be examined in isolation, holding conditions in all other markets constant. General equilibrium analysis, on the other hand, looks at the interaction of all markets in an economy.[liii] The market to be examined in a partial equilibrium analysis is not necessarily the “relevant market” as it is defined in antitrust law using cross-elasticity analysis. In fact, the market seems to be defined rather circularly as a partition that “accommodates strong economic interconnections, while ignoring remote second-best consequences …”[liv]

            Once the market is defined, the second best analysis would proceed by examining the effects of the proposed market regulation on total welfare. A second best defense to the antitrust regulation would be allowed generally if the anticompetitive conduct is in response to a market failure and serves to increase total welfare.[lv] A market failure is basically any situation where the competitive outcome results in a decrease in total welfare, whether it be due to negative externalities or welfare-reducing nonprice competition.[lvi] Of course, this is not to say that comparing the effects on welfare of allowing or disallowing the anticompetitive conduct is straightforward.[lvii]

             The other type of limited second best analysis is termed a third-best-allocative-efficiency analysis, which takes into account the difficulty of performing a second best analysis.[lviii] This method, unlike the partial equilibrium analysis, does not examine a unique market. It uses a general-equilibrium model that considers the entire economy but limits the data examined by applying a cost-benefit constraint to the gathering and analysis of economic data.[lix]

It is suggested that the “third-best” method is more appropriate than a “second-best-allocative-efficiency” analysis, which attempts, using a general equilibrium model, to determine the most allocatively efficient way to correct specific Pareto imperfections and assumes that the total number of imperfections in the economy is small.[lx] The “third-best” method is posited to be better because there are a multitude of Pareto imperfections in the economy[lxi], and so it is apparently best to be limited by the costs of the analysis rather than by methodology. Nevertheless, the general equilibrium “second-best-allocative-efficiency”  type of analysis can yield valuable insights – for example, that regulation may reduce welfare if only a fraction of the monopolistic industries in an economy are regulated.[lxii]

            Both the partial equilibrium analysis and the third-best-allocative-efficiency analysis have their shortcomings, which are somewhat complementary. The partial equilibrium analysis uses a market that is defined to include all significant second best effects. Costs of identifying and analyzing those effects are thus not explicitly taken into account. It may be that the economic analysis called for would be prohibitively expensive. The third-best-allocative-efficiency analysis, conversely, places a priority on the costs of inquiry since, at least ideally, global second best effects are examined. However, if there are gradations in second best effects, ignoring them in choosing between second best effects for the sake of cost-benefit concerns is unprincipled.


IV. The Scope of the Problem

            In order to better choose between the differing ways to incorporate second best analysis into antitrust policy, it will be helpful to return to some of the questions previously left unexamined. In considering these questions, only an intuitive approach will be used – clearly it would be superior to have a more formal development, but perhaps some insights may nevertheless be gleaned.

            First, there is the question of the feasibility of Professor Meade’s approach – is it possible to calculate the effect on overall welfare of a market intervention? The answer must be that theoretically this is possible. If the amount of the divergence, the difference between marginal cost and price, were known for the industry in question as well as for all industries throughout the economy, the extent of allocative efficiency, and by proxy, the extent of diminution in welfare could be calculated. But the ability to actually perform this calculation may be lacking, not only for the overall economy, but even for a particular industry. Is the theoretical ability to perform this calculation, however, at odds with the Lipsey/Lancaster negative corollary? That is, since the negative corollary states that it is generally impossible without specifying the conditions for attaining a maximum of allocative efficiency to specify sufficient conditions for an increase in efficiency, does it follow that even a theoretical ability to do this is precluded? Here, the answer must be that what the negative corollary indicates is an inability to specify conditions for an efficiency (welfare) increase ex ante, without performing Professor Meade’s calculation for the entire economy. The negative corollary thus does not preclude an ex post determination of the effect on welfare of a market intervention.

            A second important point is that the optimum point for an industry may not be a position of zero divergence between marginal cost and price. Consider an economy with no divergences – every industry would be allocatively efficient with marginal cost equal to price. But if even one constraint is introduced into the economy, the theory of second best requires all other conditions to change to reach a second best optimum. Thus, with even a solitary constraint imposed, the optimum point for all industries will be a position of some divergence. In addition, according to the theory, the extents of divergence are too complicated to predict ex ante. Methods such as the partial equilibrium analysis, however, simply take aim at industries with the greatest divergence. But what if those industries are at an optimum position? Again, there would be no way of knowing short of performing the welfare calculation for the entire economy. Even supposing the industries were not at an optimum position, though, it would be not be certain that reducing the amount of divergence is best to increase welfare. An optimum position means that changes in the condition in either direction will act to reduce welfare. So a policy that is directed only at reducing the amount of divergence might be misdirected. Again, the only solution appears to be to perform the welfare calculation for the entire economy, which is tantamount to specifying the conditions required for an efficiency maximum.

            In addition, since the theory of second best requires all other conditions to change upon imposition of a single constraint, the theory strongly indicates that every market in the economy is interrelated and counters the assertion that industries in the economy may be effectively partitioned for purposes of a partial equilibrium analysis. It is difficult to conceive of interrelationships in the economy as weak if a single change necessitates such a global response – of course, this is simply a consequence of maximizing a welfare function that is dependent on economy-wide conditions.

            Thus far, it appears that methods such as partial equilibrium analysis and third-best-allocative-efficiency analysis miss the point of the theory of second best. Nevertheless, there is a significant caveat to this problem. The theory of second best says nothing about the magnitude of changes required in order to reach a second best optimum. Perhaps the changes required are generally very tiny, so that a constraint imposed would cause only minor adjustments throughout most of the economy and perhaps major adjustments in only a few industries. In this case, marginal cost pricing would be a very close approximation to the second best optimum[lxiii] and antitrust policy would be correct to focus on industries with the greatest divergences between marginal cost and price. This would also explain Professor Meade’s contention that correcting the greatest divergences in the economy would be likely to improve welfare.[lxiv] The problem of second best would thus reduce to the problem of performing something like a partial equilibrium analysis so that the few major second best effects would be accounted for in specifying an antitrust remedy.


V. Conclusion

            The integration of the theory of second best into antitrust policy can proceed in a manner that may impose substantial costs of inquiry, such as the partial equilibrium analysis. It can also result in essentially zero costs of inquiry if second best analysis is deemed irrelevant  or, on the other hand, if antitrust policy is deemed irrelevant. Choosing between these approaches depends partially on whether the social costs of monopoly are taken as wasteful expenditures. If the social costs of monopoly are indeed costs to be avoided, then the best approach is to disregard second best criticisms since the allocative efficiency loss is small in comparison. If expenditures on monopoly are not social costs, the choice depends on the magnitude of the global adjustments necessary to reach a second best optimum. If these are great, then antitrust regulation should be foregone, but if they are generally minor, then methods such as partial equilibrium analysis can be used to take the few significant second best effects into account in fashioning antitrust remedies.


[i] R.G. Lipsey and R.K. Lancaster, The General Theory of Second Best, 63 Rev. Econ. Stud. 11 (1956).

[ii] J.E. Meade, Trade and Welfare 102 (1956).

[iii] Lipsey and Lancaster, supra note 1.

[iv] Id. at 12.

[v] Id.

[vi] Id.

[vii] Id. at 13.

[viii] Id. at 12.

[ix] Id.

[x] Id.

[xi] Id.

[xii] Id.

[xiii] Meade, supra note 2.

[xiv] Id. at 13.

[xv] Thomas S. Ulen, Symposium on Second-Best Theory and Law & Economics: Article: Courts, Legislatures, and the General Theory of Second Best in Law and Economics, 73 Chi.-Kent L. Rev. 189, 195 (1998).

[xvi] Meade, supra note 2, at 104.

[xvii] Lipsey and Lancaster, supra note 1, at 12.

[xviii] Id. at 17.

[xix] Id.

[xx] Meade, supra note 2, at 107.

[xxi] Lipsey and Lancaster, supra note 1, at 11.

[xxii] Id. at 12.

[xxiii] Meade, supra note 2, at 104.

[xxiv] Lipsey and Lancaster, supra note 1, at 13.

[xxv] Id.

[xxvi] Robert H. Bork, The Antitrust Paradox 114 (1978).

[xxvii] Id. at 113.

[xxviii] Lipsey and Lancaster, supra note 1, at 28-31.

[xxix] Id. at 31.

[xxx] Bork, supra note 26, at 114.

[xxxi] Meade, supra note 2, at 106.

[xxxii] Ulen, supra note 15, at 217.

[xxxiii] Id. at 217-218.

[xxxiv] Richard A. Posner, Antitrust Law 15-17 (2001).

[xxxv] Id. at 16.

[xxxvi] Id.

[xxxvii] Id.

[xxxviii] Id. at 17.

[xxxix] Mark Glick, Is Monopoly Rent Seeking Compatible with Wealth Maximization?, 1994 B.Y.U.L. Rev. 499 (1994).

[xl] Id. at 506.

[xli] Id. at 510-512.

[xlii] Id. at 510.

[xliii] Id. at 503.

[xliv] See Posner, supra note 34, at 17 for the point that studies that measured and found the social cost of monopoly to be very small did so given the existence of antitrust laws; thus, they only measure the extent of failure of those laws.

[xlv] Professor Glick, however, would also liken these to the monopoly transactions and label them a change in preferences, not a social cost. See Glick, supra note 39, at 511.

[xlvi] See Lawrence O. Sullivan, Book Review, 75 Colum. L. Rev. 1214 (1975).

[xlvii] Id. at 1220.

[xlviii] Oliver E. Williamson, Assessing Vertical Market Restrictions: Antitrust Ramifications of the Transaction Cost Approach, 127 U. Pa. L. Rev. 953, 986 (1979).

[xlix] Richard G. Lipsey, Peter O. Steiner, Douglas D. Purvis, and Paul N. Courant, Economics 106 (1990).

[l] Id.

[li] Bork, supra note 26, at 113.

[lii] Williamson, supra note 48, at 987 citing Baumol, Informed Judgment, Rigorous Theory and Public Policy, 32 S. Econ. J. 137, 144 (1965).

[liii] Peter J. Hammer, Antitrust Beyond Competition: Market Failures, Total Welfare, and the Challenge of IntraMarket Second-Best Tradeoffs, 98 Mich. L. Rev. 849, 855 (2000).

[liv] Id. at 856 n. 18.

[lv] Id. at 882.

[lvi] Id. at 859-867.

[lvii] See Posner, supra note 34, at 13 n. 5 stating that there would be difficulties in evaluating a defense based on negative externalities to a merger by cigarette companies.

[lviii] Richard S. Markovits, Symposium on Second-Best Theory and Law & Economics: An Introduction, 73 Chi.-Kent. L. Rev. 3, 7 (1998).

[lix] Id.

[lx] Id.

[lxi] Id.

[lxii] Claus Thustrup Hansen, Second-best antitrust in general equilibrium: a special case, 63 Econ. Lett. 193 (1999).

[lxiii] Richard Schmalensee, The Control of Natural Monopolies 30 (1979).

[lxiv] Meade, supra note 2, at 104.