The Limits of Liberty: Between Anarchy and Leviathan

James M. Buchanan.
Buchanan, James M.
(1919- )
Display paragraphs in this book containing:
First Pub. Date
Indianapolis, IN: Liberty Fund, Inc.
Pub. Date
Foreword by Hartmut Kliemt.
11 of 14

Chapter 8The Punishment Dilemma


If man could but design a God who would punish for violations of man-determined rules, and would, at the same time, constrain his own impulse to power, stability and progress in social order might be insured. Only under some such scheme of things could the enforcer of basic constitutional contract be made genuinely external to the parties whose separate interests are to be protected without, at the same time, being granted power for potential exploitation on his own behalf. Only then could we think of social order as a game in which the umpire is neither himself among the players nor a potential seeker in the winnings. If all men should accept such a God on faith, on the "as if" assumption that such a God exists, and if all men behave accordingly, formal law embodied in the agencies of what we have called the protective state need not be observed. Abiding by the rules in existence, and secure in the prediction that others would follow the same rules, an individual could survive and prosper in an orderly regime of social intercourse provided that the rules themselves were tolerably efficient. But faith cannot follow design; the man who might imagine such a God could not himself faithfully abide by the precepts. Shivering man must rely on his own resources to pull himself from and stay out of the Hobbesian "warre."


As Hobbes perceptively noted three centuries ago, individuals at war with one another will join in contracting with an external peacemaker, and because they value peace so highly, they will surrender to him their own powers of resistance. In the Hobbesian model, the sovereign remains external to the parties who remain in potential conflict. For better or worse, Hobbes predicted the subjugation of individual men to a sovereign master, with the latter empowered to enforce "law" as he sees fit. But what if men acknowledge no sovereign to exist?


What if they recognize the necessity of selecting an enforcer from among their own ranks, one who has interests comparable to their own? How can the function of the protective state, the enforcer of the implicit contractual agreement embodied in law, be organized by those who are themselves to be protected? And, once organized, how can this state, this enforcer, be controlled? How is it possible to delegate enforcement power to an internal agent, and, once the power is delegated, to treat this agent as if it were external?


In this chapter, discussion will be confined to the issues raised by the necessity that the enforcement of law remain internal to the set of contracting parties. As the chapter's title indicates, these issues center around the dilemma of punishment, the treatment of those who violate law. In Chapter 9 the issues involved in controlling the enforcing agent will be examined.


Man cannot design a God, and man will not universally abide by the promises that he makes. The world is neither Christian nor Kantian, although Christians and Kantians inhabit it alongside their heathen and amoral brethren. The necessity for law enforcement must be squarely faced, regardless of our romantic yearnings for an imaginary paradise. Can man turn his scientific-technical talents toward the invention of some nonhuman agents, some robot-like entities, that will be internally constructed but which will function without moral impulse, without independent interest, without consciousness? Can an internal policing equivalent of the Doomsday Machine be invented and developed into a practicable institution? Are the constraints technical or are they based on Frankenstein uncertainties?


Questions such as these should not be dismissed out of hand. We need not reach into the extremities of science fiction to think of devices that could serve as automatically programmed enforcers. To an extent, such devices exist and are in operation, for example, electric fences and gun traps. But suppose that modern science, if applied with intent, could produce mechanical agents that would detect violations of law readily and, in the same process, impose severe penalties on violators in the form of physical pain. For example, suppose that it should prove possible inexpensively to install devices on TV sets that would cause them to explode violently if moved some specified distance from a preprogrammed location. The owner would, of course, have access to a code that would deactivate the device, allowing him to move the set about at will. A potential thief would not, however, have access to this code, and the device would be relatively invulnerable to deactivation in any other manner. This automatic enforcer would seem to meet formal requirements. It would protect individual ownership rights; those who were informed about the device would be effectively deterred from law violation. Those who were not informed would have nothing to fear unless they explicitly violated the law. There would be no danger that the device would do more than it was programmed for; it would not rise up and become the master.


Nonetheless, it is clear that modern value attitudes would probably not allow such a device or devices to be installed nor, indirectly, would they allow investment for research and development of such instruments. In fact, should a person install such a device and have his TV set stolen and should the thief be physically harmed, modern courts would probably force the owner to compensate the thief.*102 The basic punishment dilemma appears even in this quasi-fictional account of robot enforcement. Why are men apparently unwilling to punish those among themselves who violate the terms of the implicit existing contract, the law that defines individual rights?

The Cost of Punishment


From his professional base, the economist's primary contribution to discourse is his emphasis on the trade-off among conflicting objectives. His principle is that of opportunity cost; more of one thing can be secured only at the sacrifice of another. In Chapter 7 we discussed law, legal structure, as a public capital good, along with some of the implications that may be derived from this conception. As noted, there is for each person some efficient or optimal quantity of general law, defined in terms of a marginal equality between the costs of further restraints on liberty and the benefits of further order in society, with the increased stability that this embodies. Although it is less meaningful, and less operational, we might also define an efficient or optimal quality of law for the community by some aggregation over persons.


This derivation of the efficient or optimal quantity of law as discussed to this point is based on the implicit presumption that once some agreement is reached, once mutually acceptable limits are defined, each party abides by contract. As the analysis has indicated, however, each person retains a private incentive to default, and some behavior toward default can be predicted to occur in the absence of effective enforcement. Any enforcement is costly, however, and a more general model must allow for the simultaneous determination of the preferred or optimal quantities of law and the quantities of enforcement. Laws that might well be efficient in the presence of costless enforcement may be inefficient when the presence of potential violations and the costs of punishing violators are taken into account. The delineation of rights in constitutional contract, the definition of legal structure, is therefore related to the costs of enforcing rights, and exogenous changes in these costs will affect the system of rights that seem socially viable. When this is recognized, we must modify somewhat our earlier discussion of the protective state. Having once defined rights, individuals may be willing to turn over the enforcement task to an external agent, but with the proviso that this agent will carry out the enforcement task only to the limits preferred by the participants. The orders to the external agent cannot be, "Enforce these rights, regardless of cost," because the cost must be borne by those who assign the enforcement to the agent, and not by the agent itself. "Law-abiding," that which is achieved by enforcement, is a public good, but it is one that is not produced without cost.


Enforcement has two components. First, violations must be discovered and violators identified. Second, punishment must be imposed on violators. Both components involve costs. Resource outlays are required to search out violations and to identify those who are responsible; these outlays increase as the desired amount or quantity of law-abiding increases, given an unchanging set of individual preference functions. Once discovered and identified, however, violators must be punished or penalized for their failure to live up to the terms of the implicit contract under which law is established. Resource outlays are also required in punishment (prisons, guards, security systems, and so on). But the primary cost of punishment cannot be represented directly in a resource dimension. The basic costs of punishment are subjective, and these can best be conceived in a utility dimension. The imposition of penalties on living beings, whether or not these beings have violated law, causes pain, utility loss, to the normal person who must, directly or indirectly, choose these penalties. "Punishing others" is a "bad" in economic terms, an activity that is, in itself, undesirable, an activity that normal persons will escape if possible or, failing this, will pay to reduce.


This "badness" aspect of punishment is accentuated when the individual recognizes that he may himself fall victim to the very standards that he lays down. He may himself be among the "others" for whom the punishment is chosen. For some laws or behavioral rules, the individual's self-interest may override adherence, at least in certain circumstances. Traffic violations offer a good example here. Recognizing that he may himself violate traffic regulations on occasion, the individual may be reluctant to accept institutions that impose severe penalties, despite his preferences that all "others" than himself should be led to obey the general rules by sufficiently severe sanctions. Just as the individual prefers that all others abide voluntarily by law while he remains free to violate it, so, too, he prefers that differentially severe punishment for law violation be meted out to others than himself. Similar effects on a choice among institutions arise from the recognition that errors may occur in the enforcement-punishment process. Even if he knows in advance that he will violate no laws or rules generally laid down for the community, the individual must consider the prospects of becoming an innocent victim. In fact, "protection for the innocent" has been a central if not dominant objective in English and American jurisprudence.


Care must be taken to distinguish between the pain costs, the utility losses, that are involved in making a choice among institutions that impose pain or suffering on individuals for law violation, and the pain costs, or utility losses, involved in the actual imposition of the prechosen punishment. The former is under discussion and analysis; the latter may, in most cases, be delegated to an agent or agency. The "punishment dilemma" arises from the elementary fact that to secure the public good of law-abiding the public bad of punishment must be accepted.*103 We do not live in a world where men abide by law without threat of punishment; if we did, there would be no need for law, as such. We could achieve ordered anarchy.

The Time Dimension of Punishment


The benefit-cost calculus involved in determining a preferred level of punishment is more complex than a simple public-good, public-bad trade-off might suggest. Punishment, as carried out, is necessarily ex post; a person who violates law is punished after the fact. The objective of punishment, on the other hand, is ex ante; punishment institutions are chosen only for the purpose of preventing or deterring violations. Consider a person who is conceptually participating in a decision on the establishment of collectively determined rules for punishment of a particular crime. The benefits from punishment are measured in the anticipated deterrent effects; the costs of punishment are measured in the pain, or utility loss, suffered by the knowledge that overt damage must be imposed on those who are not deterred. This seems straightforward enough, until we recognize that, once a punishment rule is established and violation of law then occurs, the individual must suffer a quite different "cost of punishment" as he observed the rules in operation.*104 He observes persons being coerced and harmed by the rules, after violations have occurred. The damage that the violation itself represents has been done; no punishment will restore the status quo ante. At this stage, the individual member of the inclusive community, which includes the person who has violated law as well as those who are damaged by the law violation, may be strongly tempted to modify or to change the rules that he may have indicated to be preferred in his planning or constitutional frame of reference. The presence of uncertainty concerning the actual identification of the violator will serve to accentuate this mood. The normal person may suffer utility losses on punishing known criminals; he suffers even more from the prospect that some persons apprehended will, in fact, be falsely charged. In this context, the planning frame of reference may be jettisoned altogether; and persons may begin to juxtapose the observed costs of punishment measured in the suffering of offenders against the future benefits of having punishment carried out; the rationally chosen rules may well be abandoned or relaxed.*105


Insofar as currently observed infliction of punishment is allowed to inform decisions or choices about preferred levels of punishment generally, the subjective trade-off or discount rate between present and future will influence the outcome. As this rate rises, the preferred level of punishment will fall, providing that observed punishment inflicts utility loss on the observer. There are unstable elements introduced into the behavioral system once the effects of the subjective discount rate are recognized. Those persons who violate law also make their decisions in a time dimension. As noted in an earlier chapter, law violation amounts to a destruction of social capital. Those who violate law seek immediate gain; they try to compare the expected value of this gain against the expected and future costs that punishment involves. A decision to break law represents a trade-off between anticipated future loss and present gain. From this calculus, an increase in the subjective discount rate will lead to increased law violation. But, as noted, for those who are influenced by current infliction of punishment in their choice of optimal levels, an increase in the subjective discount rate will, ceteris paribus, lead to a decrease in punishment imposed. The analysis suggests that an exogenous upward shift in the subjective discount rate will simultaneously increase law violation and decrease punishment. This may, in turn, promote further violation on straightforward benefit-cost adjustments. As the observed breakdown in standards of order proceeds, the subjective discount rate may increase further. The system of order may be degenerative until and unless exogenous forces intervene to restore stability.


Exogenous changes in "tastes" may also set off unstable reactions under the conditions of the model implied here. A modern example is provided in marijuana consumption. The law, as it existed before the 1960s, nominally imposed severe penalties on those who consumed and traded marijuana. Presumably, during the 1960s, an exogenous change in taste occurred, increasing dramatically the number of marijuana users, and, predictably, the number of those who are subjected to punishment as established in law. The nonusing members of the public observed this punishment and, finding it unpleasant, began to clamor for a reduction or relaxation of legal standards. The effect is, of course, to increase demands for marijuana usage, with subsequent further demands for reduction in punishment, or "decriminalization." The sequence may be stopped only after the legal restrictions on behavior are abandoned and marijuana usage shifted to that set of social interactions that is organized anarchistically. The implication that may be drawn from the analysis, and the example, is that the basic social decision on whether such restrictions should be imposed should be made in an ex ante planning or constitutional stage and that this decision should not be influenced unduly by observed impositions of penalties.

The Strategic or Constitutional Dimension of Punishment


There is no escape from a punishment dilemma until and unless the relevant choices of institutions imposing punishment are made strategically, at the stage of constitutional choice-making, rather than expediently. The community will remain frustrated by its own complex institutional processes if it does nothing other than respond to violations of law, once committed. A response strategy can be, and will be, exploited profitably by potential law violators, and the community will be forced to remain in undesired positions, imposing less than efficient levels of punishment and accomplishing less than preferred levels of compliance, while at the same time suffering the pain of having to punish at all. It is precisely the response mentality in thinking about and in discussing issues of law enforcement that lies at the source of modern confusion about deterrence. Punishment that is imposed ex post cannot be a deterrent ex ante for the same offense. Hence, treated solely in a response setting, and neglecting the retributive aspects, the explicit introduction of punishment may seem arbitrary infliction of pain, action that involves disutility to persons who observe and, ultimately, select the punishment. The utility-maximizing response of the potential enforcer, as an agent for the community, may therefore be that of nonpunishment or of unduly light punishment, by comparison with that punishment which might be chosen strategically at the constitutional stage by the same community.


By saying that preferred punishment institutions should be strategically chosen at the time of constitutional decision-making and not at the postconstitutional response stage, I mean only that the basic law enforcement policy and structure should be selected before explicit violation of law occurs, and independently of observed violation, even in the recognition of the serious utility losses that may, in certain cases, be borne by those who delegate the authority to punish. In this respect, as in others already discussed, the instructions should ideally be given to an external agent, instructions which should be irrevocable once made.*106


The argument may be more carefully developed with simplified numerical illustration. Figure 8.1.  Click to enlarge in new window.Consider figure 8.1. Individual B has, we assume, already committed a crime; he has violated a law. In a pure response setting, the choice remaining for individual A (who is, we assume, acting as a participant in the community's decision but is not, personally, damaged by B's crime) is to decide whether or not B should be punished. (I shall use the simple either/or alternatives here; more complex analysis could allow for continuous variability.) In the pure reaction context depicted in the figure, A may decide to leave punishment undone. He will do so if he suffers utility loss in imposing pain or penalty on another person, at least sufficient to offset the benefits stemming from carrying out retribution.


In figure 8.1, the left-hand number in each cell represents a utility indicator for A. As presented, if A confronts his punishment decision ex post, he will refrain from taking action. To carry out punishment, to impose pain on B, generates a higher utility loss to A than his failure to respond. Remaining inactive in the knowledge that B is allowed to escape scot-free after having violated law may also involve utility loss to A, but this may well be less than that which punishment embodies. The right-hand numbers are utility indicators for B. If he is punished, his utility loss may be relatively large. If he is not, he may secure utility gains over and above those already achieved in the act of law violation itself; he may experience joy in the knowledge that he has successfully exploited the "softness" of A.


As I have emphasized and as the analysis should have made clear, it is inappropriate to conceive punishment in the strict response setting depicted in figure 8.1. The individual A, here conceptualized as a participant in the community's decisions involving punishment policy, should look at the choice ex ante, that is, before law is violated. His decision should be based on predictions concerning the influence of his choice of punishment institutions on the decisions of persons, all the Bs (who will, in this instance, include A) acting under the legal system in postconstitutional stages, to abide by or to violate law. All persons are potential violators, including those who select the preferred punishment, and it is the general effect of punishment on behavior that should be relevant for the constitutional-stage strategic decision in which policy is laid down.*107


Figure 8.2.  Click to enlarge in new window.The situation may be depicted in figure 8.2, where B's behavior is not predetermined. If A chooses to commit an agency of the community to impose punishment, after violation, and if this choice is known in advance and is believed by a potential B, this fact, in itself, may influence behavior. To the extent that it does, B may not violate law, in which case A need not suffer the utility loss involved in observing the infliction of penalty.


As before, the numbers in the cells are utility indicators, the left-hand numbers applying to A, the right-hand numbers to B. Note that, in a nonstrategic or pure response setting, there is row dominance for A. Regardless of the course of action chosen by B, the response adopted by A would be that of nonpunishment. If, however, A predicts that the punishment strategy selected in advance will modify B's behavior, he may maximize his own utility by selecting the punishment alternative. Let us suppose that punishment, which we might assume to be a six-months' jail sentence, may be attached to a single violation of law, which we assume to be petty theft. Individual A, the participant in the policy decision, may choose to impose this punishment for this crime, or he may allow thieves to go free. Suppose that A predicts that an ex ante commitment to punishment will reduce by two-thirds the probability that B will commit petty theft. In two cases out of three, therefore, A's ex ante decision to impose punishment will produce a column 1 solution, one that is clearly preferable to either solution in column 2. On the other hand, if A does not select the punishment, he predicts that B will violate law in three cases out of three. If these probabilities (shown in the bracketed terms) inform A's choices, he will select punishment over the nonpunishment alternative. The expected utility for each alternative is indicated by the numbers to the right of the matrix. Individual A can expect two units of utility from choosing the punishment strategy, even if he knows that, in one out of three cases, he will suffer a two-unit utility loss.


In more general terms, A's choice will depend critically on, first, the relationship between the anticipated benefits from law-abiding behavior, from order in the community, and the anticipated utility losses consequent on the actual necessity to inflict punishment. Second, the choice will depend on the predicted influence on the behavior pattern of those subjected to the decision, the Bs. To the extent that A places a high value on order and/or is relatively indifferent to the infliction of punishment (a "hard man," who believes in retribution), he will, ceteris paribus, be more likely to select a relatively strict punishment schedule in his constitutional choice. To the extent that A considers the externalities implicit in disorder minimal, is relatively sensitive to the imposition of pain (a "soft man," who thinks all people are basically noble), and who considers himself as a potential recipient, whether as a result of law violation or error, he will tend, ceteris paribus, to choose relatively mild institutions of punishment. The predicted effects on B's behavior, impounded in "ceteris paribus" in the above statements, may, however, overwhelm the other determinants. If A predicts that the behavior of B (all the potential Bs) is highly sensitive to the penalties imposed on violations of law, the constitutional choice may rationally include a severe punishment strategy even if, otherwise, he fits the "soft man" characterization. Conversely, if the Bs are predicted to be relatively insensitive to variations in anticipated punishment, even the "hard man" may refrain from choosing a severe punishment strategy. If A predicts absolute invariance in B's behavior, there need be no strategic consideration in his choice; rules could simply not be enforced; the group would remain in Hobbesian anarchy.


The analysis may easily be generalized to allow for variation in punishment levels, allowing us to define the conditions for attaining optimally preferred or efficient institutions. As noted, an increase in the level or severity of punishment, chosen ex ante, will reduce the probability of law violation. This is the benefit side of the constitutional choice. But such an increase will also involve a higher utility loss because of the implied imposition of stiffer penalties on those who do break the law, the cost side of the ledger. For any person, A, considered as a participant in the community's choice of legal structure, the preferred or efficient level of punishment is attained when the margins are equated; that is, when an incremental increase (or decrease) in punishment generates marginal gains in behavioral order that just match the marginal losses which the more severe treatment of offenders involves, both measured in value standards of the person who makes the choice.


An additional simplifying assumption of the analysis must be modified at this point. I have implicitly discussed punishment levels, which may be conceived as involving a severity component (length of jail terms, amount of fines, physical pain, and so on) on the implicit assumption that the certainty component is exogenously determined. Characteristic of modern discussions of law enforcement, however, has been the emphasis on the effects of both of these components, certainty and severity, and on the trade-offs that are possible between them. To the extent that potential law violators predict with certainty that they will be subjected to punishment, the severity of the punishment itself may be reduced, and vice versa. The trade-off between certainty and severity introduces an interdependence between pecuniary and nonpecuniary aspects of law enforcement. Increasing the certainty of punishment for law violation may require substantial outlays on the discovery, investigation, and identification of offenders, that is, on improved policing, without substantial change in the aggregate disutility that punishment embodies. Consider an example. Suppose that improved police methods insure that five rather than two of each ten offenders are apprehended, convicted, and sentenced to jail terms of three months each. This policy may produce the same effect on the behavior of potential law violators as one that increases the jail terms from three to six months with no change in police input. The first policy alternative requires an investment of funds, and the costs in utility terms arise only because these funds must be drawn from otherwise desirable uses. The second policy alternative may be accomplished with relatively little additional resource commitments, but it generates nonpecuniary reductions in utility.*108

The generality of punishment rules


We have discussed the punishment choice for A, treated as a participant in the selection of the basic legal rules, on the presumption that all potential violators or offenders are identical in their predicted behavioral responses to alternative punishment strategies. A major difficulty arises, however, when the number of potential violators is large and/or when predicted behavioral responses vary widely. In this case, which is descriptive of any real-world setting, the ideally preferred and efficient punishment strategy will differ over different potential offenders.


Consider the earlier numerical illustration. Suppose that the predicted response behavior of one group of potential law violators, the B1's, is that shown in figure 8.2 above. The advance commitment to a six-months' term for petty theft will reduce by two-thirds the number of violations. As indicated, under these probabilities, the preferred strategy for A (or the group of A's participating in a collective decision) is that of punishment. If, however, there should exist a second set of potential violators, the B2's, which are characterized by the pattern of response behavior depicted in Figure 8.3.  Click to enlarge in new window.figure 8.3, the preferred strategy will be different. The discrete punishment alternative shown to be utility maximizing when applied to the enforcement of law of the B1's will, when applied to the B2 group, reduce the violation of law by only one-tenth. Faced with this array of choice alternatives, A will rationally refrain from adopting the punishment strategy. In this instance, the utility loss which punishment itself embodies more than counters the relatively minor utility gains secured by the additional adherence to law that the imposition of punishment produces.


Figure 8.4.  Click to enlarge in new window.We may discuss the problem somewhat more fully with the aid of figure 8.4, I and II. Here we allow for three rather than two punishment alternatives. The model could, of course, be expanded to allow for continuous variation in the punishment set, but this extension is unnecessary for current purposes. If the individual participant in the choice, A, considers only the potential offenders in the B2 group, and if his predictions about their response behavior are those summarized by the probability coefficients in figure 8.4(II), he will prefer the severe punishment strategy. However, if A considers only the potential offenders in the B1 group, whose response behavior is summarized by the probability coefficients in figure 8.4(I), he will prefer the moderate punishment. The illustration clearly shows that preferred punishment varies with differences in predictions about response patterns. Note that, in the illustration, there is no other differentiation as between the B1 and the B2 group. The utility indicators in the matrices are identical for the two cases.


The constraints within which A must choose may, however, require that the same punishment be imposed on all persons who break the same law. It might, of course, be relatively straightforward for an external Hobbesian sovereign to impose differential punishments among different groups and, by so doing, achieve a higher level of efficiency. Our question is the different one: How can a community of persons, having agreed on some initial definition of rights, agree on a set of enforcement institutions that will be tolerably efficient when behavioral response patterns are predicted to be widely different? Even if we leave aside all of the difficulties of establishing identification of violence-prone individuals and groups, why should members of these groups, at the time of the conceptual agreement, accept the imposition of differentially higher penalties? The benefits from the imposition of such penalties may accrue largely to other members of the community rather than to members of those groups that should, in some "social rationality" sense, be subjected to discriminatory punishment.


Any contractually selected discrimination in punishment seems out of the realm of possibility, regardless of the strict efficiency basis for such discrimination. As noted earlier, much of the public discussion and attitude toward punishment fails to embody the conceptual constitutional approach and, instead, tends to reflect simple ex post reaction to violations of law. In this context, there is no basis for differential treatment. In our example B1 and B2 violate the same law, commit the same crime. In its reaction, the community may or may not impose punishment, but there is surely no apparent basis for treating one law violator more favorably than the other. Any plausibly truthful scenario must, therefore, be constructed on the requirement that a single set of punishment institutions be applied generally to all persons in the community, even in the face of acknowledged differentials in predicted response patterns. This suggests that, for any set of institutions chosen, the sanctions will be unduly and unnecessarily repressive for some potential violators and unduly and unnecessarily permissive for others. Formal requirements for selecting the most efficient set of institutions and rules could, of course, be defined by assigning utility weights for the opportunity losses in the two directions.

Public Choice of Punishment


The formal analysis of efficient law enforcement has been developed in some detail by several modern economists, and this analysis need not be elaborated here.*109 In the elementary discussion of the preceding sections, punishment strategy was examined in a two-person construction, with both A and B taking on "everyman" characteristics as appropriate. Individual A, the participant in a presumed collective choice at the constitutional level, one that involves the selection of a set of institutions for punishment, can conceptually pick out a preferred option, given his own utility function, his own endowments and capacities, his predictions about the behavior of potential offenders in response to alternative punishment strategies, his predictions about the functioning of the institutions chosen, and some knowledge of the resource outlays required to implement the alternatives. His preferred solution will embody some mix between the certainty and severity components. If this were the end of it, perhaps the most difficult aspect of the punishment dilemma need never arise. In such a simple interaction, A might accept the necessity of making his choice strategically at the constitutional level and accept the implication that this choice, once made, could not be tampered with in response to expediency considerations that arise subsequently. But there are many A's in the community, including also all of the potential B's, and the selection of a set of enforcement-punishment institutions must be collective rather than individualistic. Each member of the group may reach a personal decision on his most preferred institution, but the separate individual choices must somehow be combined to produce a unique social, community, or collective outcome. All of the problems of aggregating individual orderings seem to emerge.


The set of alternative enforcement-punishment institutions that satisfy the separate personal preferences may be large indeed. In our earlier discussion of the conceptual constitutional contract, the problem of reaching general agreement was examined, especially with reference to general agreement on the quantity of law, in Chapter 7. Insofar as the first leap from anarchy takes the form of a disarmament pact, with persons agreeing to honor the rights of others, specific terms may be more readily agreed on. Tizio and Caio may accept a mutual disarmament compact in which they agree to refrain from invading each other's domain. This is not, of course, to suggest that the definition of separate individual rights emerges in some natural sense. It does suggest that, once definition is attained, the contract is more or less complete. The definition of the appropriate dividing line between those interactions subjected to formal law and those that are not was shown to be a much more difficult aspect of social contract. As we have conceptualized it, the basic constitutional contract must also include the terms under which the community may undertake collective or joint action in postconstitutional stages; that is, the basic constitutional framework must lay down the rules for making collective decision concerning the provision and financing of public goods and services. Individuals may differ among themselves over the working properties of alternative rules, and, because of this, they may optimally prefer separate structures. I have not discussed this problem as such; I have simply presumed that some general agreement on such decision rules is achieved.*110


I have, in this chapter, argued that punishment institutions and rules should also be included as a part of the conceptual constitutional contract under which a society operates. The attainment of general agreement on a set of preferred punishment rules may, however, create more difficulties than almost any other aspect of the basic constitutional settlement. In this respect it is akin to the problem of reaching agreement on the range of law, discussed in Chapter 7. Even within the constitution-making process, some initial agreement on a decision rule may be required. Each member of the group will, presumably, have an interest in laying down some rules for punishing those who violate the basic terms. But since different members will disagree over the severity of punishment, agreement may first have to be reached on how a unique set of punishment rules can be selected. If a majority rule is chosen as the instrument, the familiar Condorcet paradox may be present. It is quite possible that separate individual orderings may not be single-peaked or single-troughed, in which case cyclical patterns may be generated.*111 Even if we disregard this prospect, the dissatisfaction of participants whose preferences are not median for the group must be acknowledged. A majoritarian decision amounts to satisfying the preferences of the median man. The selection of a set of enforcement-punishment institutions which makes the median man happy must leave others on both sides of the choice spectrum unhappy. There will be some persons who consider the median choice to be overly restrictive and others who consider the median choice to be unduly permissive in its operation and effects.


As with the conceptual contractual negotiations over the range of behavioral restrictions, there must exist some means of securing general agreement on levels of punishment, provided that appropriate side payments or compensations separate from the punishment choice itself can be made. Those who intensely prefer severe punishment may, in some instances, purchase the agreement of those who prefer less severe penalties for law violation, or vice versa. Conceptually, agreement may be attained, but, practically, the choice of a set of punishment institutions presents more difficulties in attaining acceptable compromises among differing preferences than almost any other aspect of the imagined constitutional contract.


With the collective choice of an enforcement-punishment institutional structure, difficulties present themselves which do not appear in other aspects of constitutional agreement. As the analysis suggests, it is essential that the choice of punishment rules be made at the constitutional stage, where strategic effects of alternatives can be assessed and predictions made. That is to say, punishment rules must be chosen before punishment becomes necessary. To the individual participant in the constitutional choice, however, the strategic implications may not be evident. He does not feel individually responsible for the outcome that emerges from the group deliberations; the costs and benefits are diffused generally among all members of the community, and among many time-periods. The individual participant behaves as if he is purchasing a genuinely public good. He will not be motivated to invest in information about the choice alternatives.*112 To the extent, therefore, that a strategically rational approach to the selection of a set of enforcement-punishment institutions requires a more sophisticated analysis than a simple response, the collectivization of decision at the constitutional level introduces major complications. The rules for punishment that might emerge from a deliberative process may not reflect careful weighing of alternatives. The outcome may seem almost arbitrary, which, in turn, offers the temptation for tampering with the rules in a postconstitutional response setting. It is one thing for the analyst to suggest that community decisions about punishment should be made at the constitutional level, and that these decisions should be sophisticated in the strategic sense. It is quite another thing to suggest that the community decisions on punishment will be made in this fashion, either in terms of the levels of decision or in terms of the informational-analytical content.

Public Choice of General Rules


When both of the difficulties discussed in the two preceding sections are joined, the existence of extreme departures from individually preferred punishment strategies is not surprising. Consider an individual whose preferences and predictions lead him to hope for a much more severe set of punishment institutions than those that he observes operative in the community. This person will be dissatisfied with the punishment meted out generally to all potential law violators. But he will be even more acutely unhappy with the application of general rules to those potential violators who stand at the violation-prone end of the response spectrum. Such a person endures a double opportunity loss; the general institutions are nonpreferred, and the necessary generality in application accentuates the intensity of his dissatisfaction. The individual whose preferences and predictions are the obverse is equally frustrated by the observed operation of enforcement-punishment institutions. The set is nonpreferred in the direction of being overly repressive; liberty is sacrificed for order beyond his own personal limits of preference. And the application of the rules is also painful when those who are highly sensitive in response are subjected to the general treatment offered to everyone.


There is no escape from the conclusion that the punishment dilemma is genuine for any community that seeks to ground its legal structure on individual values. As earlier chapters have discussed, a conceptual contractual origin for the delineation of individual rights, the initial constitutional contract, can be derived. Furthermore, we can at least conceptualize a contract with some external enforcing agent (sometimes called a contract of government), with this agent being charged with the strict policing function. Such an agent must, however, be constrained in its application of punishment for law violation, and, presumably, rules or institutions embodying punishment norms must be selected by the members of the community in some quasi-constitutional setting. It is here that the dilemma emerges in its sharpest form. The genuinely democratic regime will tend to be reactive rather than strategic in its decision processes, and it will tend to renege on the punishment choices that are made in some prior stage of deliberation. The punishment institutions, as observed, will tend to reflect individuals' current motivations of retribution, justice, and compassion, rather than their rationally chosen long-term interests as embodied in quasi-permanent rules. The result can only be some structure that generates widespread dissatisfaction among members of the community, dissatisfaction which, in itself, tends to undermine the respect for rights and the enforcement of rights, respect that is essential to maintain the social capital that law, in its entirety, represents.


This much may be granted. But what is the alternative? We have discussed the possible delegation of authority for enforcement to an external agent, even if this agent is created and manned by persons who are simultaneously internal to the community. This authority may be allowed relatively free reign in locating and in identifying the violations of rights laid down in constitutional contract. But can this external enforcing authority also be empowered to make its own choices among alternative punishment strategies, independent of the preferences of citizens? There are historical examples where specialized professional classes of jurists have been allowed relatively complete powers of punishment. This avenue of escape from the dilemma, however, immediately presents another: How is the external agent to be controlled?

Notes for this chapter

As witness the infamous gun-trap judgment in Iowa. For a general discussion of state laws, see Richard A. Posner, "Killing or Wounding to Protect a Property Interest," Journal of Law and Economics 14 (April 1971): 201-32.
For a discussion of the punishment dilemma in a different setting, see my "The Samaritan's Dilemma," in Altruism, Morality, and Economic Theory.
For a generalized discussion of the different conceptions of "cost" that has relevance here, see my Cost and Choice: An Inquiry in Economic Theory (Chicago: Markham Publishing Co., 1969).
My discussion suggests that the observed punishment of offenders, ceteris paribus, imposes a utility loss on representative members of society. This may not be true, especially with respect to major crimes. In observing retribution or "justice done," individuals may actually secure benefits or utility gains. This will not modify the analysis, as such, although it will make the problem under discussion less serious than otherwise. In the extreme, if the average or representative person in the community should actually enjoy seeing others punished, the direction of the bias introduced by taking this into account would be the opposite from that which I have suggested. My own view is that modern social attitudes are much more accurately described by the model which my discussion implies than by its opposite.
Cf. Montesquieu, The Spirit of Laws, vol. 38, Great Books of the Western World (Chicago: Encyclopaedia Britannica, 1952), p. 35.
The argument developed here is, of course, essentially a contractarian theory of punishment, as distinct from a utilitarian theory, although there are related aspects. In a genuine contractarian theory, there is no problem raised concerning the "right" of some persons to punish others, since, in effect, individuals who find themselves in the implicit social contract that any legal order represents have presumably chosen to be punished as the law directs when they violate law. My argument seems in these respects to be close to that presented by Kant and, especially, by Hegel. On the Hegelian theory, and as related to the use made of this by Marx, see Jeffrie G. Murphy, "Marxism and Retribution," Philosophy and Public Affairs 2 (Spring 1973): 217-43.
On this trade-off, along with a discussion of some of the policy implications, see Thomas Ireland, "Public Order as a Public Good."
See Gary Becker, "Crime and Punishment: An Economic Approach," Journal of Political Economy 76 (March/April 1968): 169-217; Gordon Tullock, The Logic of Law (New York: Basic Books, 1971); George Stigler, "The Optimum Enforcement of Laws," Journal of Political Economy 78 (May/June 1970): 526-36; Gary Becker and George Stigler, "Law Enforcement, Corruption, and the Compensation of Enforcers," mimeographed (Paper presented at Conference on Capitalism and Freedom, Charlottesville, Virginia, October 1972).
In part, this neglect of the discussion of preferred rules for making ordinary collective decisions, and of the efficiency properties of alternative rules, stems from the earlier treatment of these questions. See James M. Buchanan and Gordon Tullock, The Calculus of Consent.
The modern seminal works on the voting paradox are Kenneth Arrow, Social Choice and Individual Values (New York: Wiley, 1951); and Duncan Black, Theory of Committees and Elections (Cambridge: Cambridge University Press, 1958).
For a more extended discussion of these points, see my "Individual Choice in Voting and the Market," Journal of Political Economy 62 (August 1954): 334-43, reprinted in Fiscal Theory and Political Economy (Chapel Hill: University of North Carolina Press, 1960); and Gordon Tullock, "Public Decisions as Public Goods," Journal of Political Economy 79 (July/August 1971): 913-18.

End of Notes

11 of 14

Return to top