The Legislator’s Dilemma - The Moral Economy: Why Good Incentives Are No Substitute for Good Citizens - Samuel Bowles

The Moral Economy: Why Good Incentives Are No Substitute for Good Citizens - Samuel Bowles (2016)

VI. The Legislator’s Dilemma

“The superiority of liberalism has been mathematically demonstrated” was Le Figaro’s headline, slightly embellishing what Gerard Debreu had actually said in an interview with the newspaper in 1984.1 The basis for Debreu’s remarks was the first fundamental theorem of welfare economics, to which he (along with Kenneth Arrow and others) had contributed a third of a century earlier. But as we will see, research in economic theory since their “invisible hand theorem” suggests a quite different conclusion about what the math demonstrated.

Many years after Debreu’s Figaro interview, renewed discussions between Aristotle’s Legislator and his colleagues in economics are friendly enough. Since Richard Titmuss’s The Gift Relationship (1971) was published, economists have been intrigued—but mostly not persuaded—by his claim that policies based on explicit economic incentives may be counterproductive because they induce people to adopt a “market mentality,” and that such policies thus compromise preexisting values that lead people to act in socially beneficial ways.2

At the time of the book’s publication there were two reasons to doubt Titmuss. First, little hard evidence showed that social preferences such as altruism, fairness, and civic duty were important influences on individual behavior or were in any way essential to the functioning of a market-based economy. Second, even had these social preferences been thought to be important influences on behavior, Titmuss’s book offered scant evidence that explicit economic incentives would undermine them.3

Another reason that most economists did not share Titmuss’s concern was the then-burgeoning field of mechanism design. This body of research, an extension of the earlier Marshall-Pigou tradition in what is called welfare economics, addressed market failures by means of green taxes, training subsidies, and other incentives. Recall that market failures are situations in which people’s uncoordinated actions result in Pareto-inefficient outcomes. This outcome, in turn, is defined as one for which there exists a technically feasible alternative outcome, given existing resources and technologies, in which at least one person is better off and nobody is worse off. Mechanism design held out the promise that even among entirely self-regarding citizens, well-designed incentives could, along with market prices, implement Pareto-efficient outcomes. As a result, at the time Titmuss’s book appeared, virtue was something economists thought they could safely ignore, just as J. S. Mill had advised them to do over a century earlier.

Economics Discovers Aristotle

Research in the intervening years has upset this complacency. As we have seen, experimental and other evidence has made it clear that ethical and other-regarding motivations are common, and moreover, as Titmuss claimed, that they can be crowded out by incentives. At the same time, economists were discovering that simple textbook metaphors like James Buchanan’s fruit stand, at which “both parties agree on the property rights,” were a poor guide to understanding a modern capitalist economy. Economic theory turned to the study of the exchange process when contracts are incomplete, that is, when it was not the case that everything that matters can be stipulated in an enforceable agreement. New models of the labor market, for example, recognized that work itself was not something that could be contracted for, so getting a job done well depends at least in part on the employee’s intrinsic desire to do the job well. Similarly, in the credit market even the best contract cannot guarantee the repayment of a loan if the borrower is bankrupt; lenders need to trust the borrower’s account of the project that the loan will finance.

The new microeconomics of labor, credit, and other markets enumerated the ways that, as Arrow said, social norms and moral codes could induce economic actors to internalize their actions’ costs and benefits to others when contracts failed to do this. The employee’s work ethic would induce her to internalize the cost that she would impose on her employer by keeping in touch with her Facebook friends while on the job. The borrower’s sense of honesty would deter his misrepresenting the riskiness of the project for which he was seeking a loan that would, in all likelihood, never be repaid if the project failed. Economists were coming to see why prices alone could not always do the work of morals.

Another advance in economic theory—in the seemingly unrelated field of macroeconomics—raised doubts about the “constitution for knaves” policy paradigm. A generation ago, Robert Lucas rocked economics with a simple observation: taxes and other governmental interventions in the private economy affect not only (as intended) the costs and benefits of citizens’ actions, but also their beliefs about the future actions of others (including the government). Thus, just as in the case of citizens’ experienced values, as illustrated in figure 3.1, incentives may have an indirect effect (in this case, via citizens’ beliefs) as well as direct ones. For example, announcing stiffer penalties for the nonpayment of taxes provides an incentive to pay up, but it also may convey the information that noncompliance is common, leading formerly honest citizens to cheat.

Lucas reasoned that the effect of a policy intervention can be predicted only if one takes into account these indirect effects on beliefs, and then studies the joint outcome in which citizens’ beliefs and the targeted economic actions are both mutually dependent. His point was that a new economic policy is not an intervention in a frozen model of how the economy works, but rather a change in the workings of the model itself: “Given that the structure of an econometric model consists of optimal decision rules of economic agents, and that optimal decision rules vary systematically with changes in the structure … relevant to the decision maker, it follows that any change in policy will systematically alter the structure of econometric models.”4 He concluded: “Policy makers, if they wish to forecast the response of citizens, must take the latter into their confidence.” This idea was taken to be so important that economists now capitalize it, bestowing an honor on the Lucas Critique withheld even from the invisible hand. Here I apply Lucas’s logic to cases in which incentives affect not only beliefs (as Lucas stressed) but also preferences.

Not surprisingly in light of these and other developments, economics began to change. Albert Hirschman chided his fellow economists who, he said, propose “to deal with unethical or antisocial behavior by raising the cost of that behavior rather than proclaiming standards and imposing prohibitions and sanctions,” probably because “they think of citizens as consumers with unchanging or arbitrarily changing tastes in matters civic as well as commodity-related behavior.” Economists, according to Hirschman, disregard the fact that “a principal purpose of publically proclaimed laws and regulations is to stigmatize antisocial behavior and thereby to influence citizens’ values and behavioral codes.”5 Michael Taylor, a political scientist, took up Hirschman’s idea that legal structures shape preferences and social norms. And he went further, turning Thomas Hobbes’s justification of state authority on its head and suggesting that Hobbesian man might be the result of, and not a reason for, the Hobbesian state.6 Reading Hirschman and then Taylor in the 1980s set me to work on the project that resulted in this book.7

I was far from alone. At a meeting of the American Economics Association in 1994, Henry Aaron pointed out “the failure of economists to take the formation of preferences seriously,” and went on to suggest that the Lucas Critique be extended accordingly.8 The economist Bruno Frey titled a paper “A Constitution for Knaves Crowds Out Civic Virtue.”9 The political scientist Elinor Ostrom, soon to be honored with a Nobel in economics, like Taylor worried about “crowding out citizenship.”10 The discipline of economics, which had spurned Titmuss a generation earlier, eventually rediscovered him. The subtitle of a paper in a top economics journal in 2008 asked “Was Titmuss Right?” (The authors concluded he was right about women but not men.)11

Economists, at least some of them, not only reread Titmuss but also rediscovered (or, more likely, discovered) Aristotle. What should his Legislator do, given this new knowledge about the possible anti-synergy between incentives and the social motives they now recognized as essential to a well-functioning economy and society? The new turn of research in economics, while recognizing the crowding-out problem, did not provide much of an answer. It had not yet systematically exploited the burgeoning experimental evidence for its policy implications, nor had it pursued the logic of recent developments in economic theory that would help clarify the Legislator’s dilemma and perhaps suggest a way out. These are the Legislator’s remaining tasks.

Mechanism Design: Can Prices Do the Work of Morals?

Not all economists were convinced that the importance of social preferences and the crowding-out problem warranted a new approach to policy. One could, they might insist, take on board the new experimental evidence and advances in theory without abandoning the self-interest-based policy paradigm.

Suppose that a policy maker, they reasoned, recognizes the need to address market failures and also understands that explicit incentives may crowd out the ethical and other-regarding motives. A holdout for the self-interest paradigm could go on to concede that these social preferences might otherwise induce actors to internalize the costs and benefits that their actions confer on others, thus mitigating the market failure that arises when contracts are incomplete. But the policy maker might nonetheless be confident that he could devise systems of subsidies, penalties, and constraints to implement socially desired outcomes, even if social preferences had been crowded out to the point that citizens were entirely amoral and self-regarding.

To think this possible, one would have to bar strong crowding out (so as to preclude cases in which incentives were counterproductive) and concede that the policy maker herself must not be entirely self-interested. But given these two caveats, a constitution for knaves just might do the job, as Hume recommended, even (as Hume never intended) in a world of not-at-all-hypothetical but instead very real, flesh-and-blood knaves.

This remains the canonical model of policy making in economics, and it is worth exploring whether it could possibly work. We will see that it could not.

To do this we need the theory of mechanism design.12 Founding works in this branch of economics appeared at about the same time as Titmuss’s book.13 Since then, the growing concern with environmental, public health, and other kinds of market failures has given new urgency to the mechanism designer’s attempt to give “the invisible hand a helping hand,” as the Economist put it in 2007, struggling to explain the challenging mathematics of the field when three of its leading scholars—Leo Hurwicz, Eric Maskin, and Roger Myerson—were awarded the Nobel Prize.

Helping the invisible hand inevitably requires that government play a role in the economy, but there is nothing Big Brother-ish about mechanism design. And surprisingly, this is where it runs into trouble.

The theory begins with recognition of the necessarily limited reach of the policy maker. The task of the mechanism designer is to provide a set of contracts, property rights, and other social rules—called a mechanism—that will attenuate or eliminate market failures. But both realism and a commitment to individual privacy require that the policies proposed can be implemented even when important bits of information about individuals are known only privately and hence cannot be used by the mechanism designer to implement the incentives, constraints, or other aspects of the proposed policy.

This privacy-of-information restriction excludes utopian solutions in which the designer could simply wish away the underlying reason for the market failure. If the proposed mechanism could use information about how hard a person worked, for example, or about a buyer’s or seller’s true valuation of some good or service, then the same information could have been used to write a complete private contract between the parties. And that contract would have eliminated the market failure that the mechanism designer had been called in to address.

To correct the problem of free riding in a work team, for example, a mechanism designer cannot simply propose an enforceable contract that all must work hard and well, or even that each member must truthfully report how hard she worked. If the mechanism designer could do this, then the team members could have done the same and simply paid team members according to how hard they worked. There would have been no need for the mechanism designer in the first place.

To its practitioners three conditions for an acceptable solution then define the challenge of mechanism design. The first is that the resulting allocation be Pareto-efficient.

Second, policies must rely on individuals’ voluntary participation in their economic activities. They must be free to choose their own courses of action, guided by their own preferences, including opting out of any possible exchange or other interaction. In economic language, the outcome must satisfy the individual’s “participation constraint”; each must regard participating in the mechanism as better than not doing so, meaning that participation is voluntary. The outcome must also be “incentive compatible,” meaning that whatever the individual does while participating must be motivated by the individual’s own preferences.

Third, there can be no restrictions on the kinds of preferences people may have. So the mechanism must work even if individuals are entirely self-interested and amoral.

Call these three conditions efficiency, voluntary participation, and preference neutrality. The first imposes a minimal collective-rationality condition (minimal because it ignores, for example, considerations of justice). The second rules out the confiscation of property or mandatory participation in an exchange. Along with the second, the third expresses the standard liberal commitments to individual liberty and the state’s neutrality on matters concerning individuals’ conceptions of the good.

The last has recently come to be termed liberal neutrality, expressed by Ronald Dworkin’s dictum that “political decisions must be … independent of any particular conception of the good life, or of what gives value to life.”14Peter Jones likewise writes: “It is not the function of the state to impose the pursuit of any particular set of ends upon its citizens.”15 Neutrality in this sense is not universally advocated by liberals. My broad interpretation of liberal neutrality—the permissibility of an unrestricted set of preferences—would be objected to by some, for example, who would point out that a preference for domination of others might be judged impermissible without violating liberal neutrality. But I use the requirement of an unrestricted set of preferences here simply to allow individuals the systematic pursuit of material self-interest, which could hardly be objected to under the liberal neutrality doctrine.

If mechanism design had succeeded in its quest for privacy-respecting rules that eliminated market failures while respecting the two quintessentially liberal requirements of voluntary participation and preference neutrality, Le Figaro could hardly have been faulted for its headline about what the math had demonstrated. Moreover, the holdout for the self-interest paradigm would have been vindicated. So if mechanism design had found its holy grail, we would have to conclude that a constitution for knaves would work, at least in this limited sense, however unpleasant it might otherwise be. But more than four decades of research in mechanism design has showed quite the opposite.

A (Liberal) Constitution for Knaves

A mechanism is a set of rules that the designer might impose on a population to influence the behaviors that determine how an economy’s resources are used. The fines and subsidies in the previous chapters’ experiments are found in the mechanism designer’s toolbox. So too are such familiar ways of determining resource use as majority rule, buying and selling in competitive markets, and, as we will see, more exotic rules as well. The setup defining the mechanism designer’s task is strikingly like the problem faced by Machiavelli’s benign ruler, “who would found a republic and order its laws,” and was charged with governing well through the imposition of enlightened laws upon a citizenry of “natural and ordinary humors.”16

This context of an enlightened ruler and self-interested subjects is exactly the setting for the “rotten kid” theorem devised by Gary Becker, another Nobel in economics. He identified conditions under which an altruistic household head (the mechanism designer) could impose a rule governing his transfers of income within the family that would induce all family members to act as if they cared no more for their own well-being than for each of the other family members: “Sufficient ‘love’ by one member guarantees that all members act as if they loved other members as much as themselves. As it were, the amount of ‘love’ required in a family is economized: sufficient ‘love’ by one member leads all other members by ‘an invisible hand’ to act as if they too loved everyone.”17

The theorem works, Becker explains, because the altruistic family head can arrange things so that every family member will “internalize fully all within-family ‘externalities’ regardless of how selfish … these members are.” Thus, under the right rules governing how income is distributed within the family, entirely self-interested family members will act as if they were entirely altruistic. This is a particularly dramatic case of prices doing the work of morals.

And it is exactly what the Economist had in mind: Becker provided an analogue for the invisible hand in family relationships, where Adam Smith’s variant does not work. Becker concludes, “Armed with this theorem, I do not need to dwell on the preferences of the nonheads”—even, apparently, the rotten one after whom the theorem is named. I studied Becker’s theorem at about the time I haplessly issued a price list for household chores that my not-at-all-rotten kids had previously done voluntarily. I did not see the connection at the time. You already know how that turned out.

Exported to the world of public policy, Becker’s rotten-kid theorem suggests that well-designed public incentives could dispense with virtue entirely as a foundation of good government (at least among “nonheads”). Just as importantly, it spares the policy maker or constitution writer the liberal embarrassment of having to foster in citizens some particular values—a concern for the environment, for future generations, or even for one’s own family, for example—over others.

But just as proof of the invisible hand theorem demonstrated how implausible were the axioms required to advocate a laissez-faire policy on efficiency grounds, modern mechanism design has clarified the limits of even the cleverest ways of giving the invisible hand a helping hand via public policy. Soon after the publication of Becker’s celebrated paper, for example, it was shown that the rotten-kid theorem relied on highly unrealistic mathematical assumptions, which limited its application to a set of rather special cases, quite remote from the general problem faced by family heads and mechanism designers.18 There would be no simple shortcuts for those who would “found a republic and order its laws.”

To see why a constitution for knaves could not work, let me set aside the objection that the constitution writer or mechanism designer had better not be a knave, and that a life surrounded by knaves would be unpleasant even for knaves. The basic idea in mechanism design is to align individual incentives with the objective of efficiency, defined over a large group of people. This may be done, as we saw in chapter II and as Becker explained, if we can find a way to induce each person to act as if he or she internalized all of the benefits and costs to others resulting from his or her actions. This is the condition of Robison Crusoe on his island; he alone “owns” the results of the work he has done, the risks he has taken, and the knowledge he has gained. Getting people to act as if they were like Robinson Crusoe alone on his island is the name of the game.

Just how difficult it is to do this for a large group of individuals with unrestricted preferences will be clear from the problem that arises in any process where people contribute to a joint output. Think about a group of software engineers working together to write code for new applications. To pose this as a problem of mechanism design, consistent with the preference neutrality condition, we suppose that each team member is interested only in his or her own material payoff, however determined. The mechanism designer (standing outside the team a bit like the legendary philosopher-king) can observe the team’s total output but cannot see how hard each member has worked (respecting the privacy-of-information restriction).

The problem for the designer is that if each team member receives as compensation simply an equal share of the output of the team, then a worker’s payoff for providing one unit of additional output will not be one unit of compensation, as it would be for Robinson Crusoe, but instead, because the additional unit will be shared among all of the team, 1/n units, where n is the number of team members. Under this shared-output compensation rule, working is analogous to contributing to a public good.

Each worker owns only a small fraction of the full benefit that his efforts have created, so if he can choose on his own (the voluntary-participation condition), he will provide less effort for the team than would be efficient, because he will not take into account the benefits that his effort confers on other team members. In this situation, every worker could be better off if each worked a little harder. Of course the designer knows that they would voluntarily work harder if they altruistically valued the contributions that their efforts made to the material payoffs of the other team members, but the preference-neutrality and voluntary-participation conditions prohibit him from imposing other-regarding preferences on the team or from compelling members to act counter to their preferences.

A mechanism that works for these individuals should result in each team member’s acting as if she were Robinson Crusoe, owning the results of her work. Each would then work up to the point where the discomfort of working a little harder was exactly compensated by the additional benefit of her work to the members of the entire team (more technically, the marginal disutility of work is equal to the marginal benefit that she gains from work). But if the designer cannot require the team members to provide any particular level of work (that is private information), meeting this objective seems to be impossible.

But it is not. The designer need only announce that he will pay each member the entire value of the output of the team, minus a constant sum to be determined by the designer. This bizarre mechanism ensures that any contribution by a member to the output of the team will be exactly compensated, giving each member the Robinson Crusoe incentives of an isolated individual who owns the entire fruits of his labor. The subtraction of a constant sum is necessary to allow the designer to balance the team’s budget; otherwise, he would be paying out n times the team’s total output.

Problem solved?

Not yet. To see why, we can introduce some real-world risk. Suppose the team’s realized production depends not only on the sum of the team members’ efforts but also on chance events affecting production but not controlled by team members. Call these positive or negative chance events “shocks,” and suppose, realistically, that like each member’s effort, they are not observable by the designer, so he cannot determine whether the team’s unexpectedly low output in some year comes from bad luck or from shirking workers. The contract offered by the designer would have to ensure that team members received an expected income (averaging over “good” shocks and “bad”) at least as great as their next-best alternative (the wage in some other job, unemployment insurance, or something similar).

But given the fluctuating level of total output as a result of a positive shock, for teams of any significant size each member’s realized income (total output minus some large constant sufficient to balance the budget) in any period could be many times larger than the workers’ next-best alternative. That is not the problem, however. The hitch is that shocks can also be negative, so realized income could also be much less than what the team member would have received in some alternative employment. It could, in fact, be a large negative number, meaning that the member would have to pay the mechanism designer rather than the other way around.

The problem arises because in order for this clever and odd mechanism to work, the pay of each member has to be keyed to the entire team’s realized output, and both negative and positive shocks to total output would realistically dwarf any individual’s average compensation in the long run. A contract under which, in some periods, a team member would not be paid and instead would be required to pay the team a substantial multiple of her expected salary is not likely to be attractive to many workers. No contract of this type would be voluntarily accepted by the team members, thus violating the condition of voluntary participation.

If the team members could borrow an unlimited amount of money in the bad years, they might be okay with such a contract. But this could never be the case, because a similar problem arises in the credit market: the contracted repayment of the funds is not enforceable if the team member is without funds. Thus, implementing the optimal contract for the team members just displaces the mechanism designer’s challenge to the analogous problem in the credit market: the incompleteness of contracts, where he would confront similar obstacles to a solution.

The Liberal Trilemma

This is not the end of the story, of course. Far more complicated mechanisms might address the problems that confront the designer’s simple “pay each the whole output” plan. But the verdict of the now-vast literature of mechanism design is that the designer’s problems are endemic, not specific to the team-production example. In 2007, when the Prize Committee of the Royal Swedish Academy of Sciences honored Maskin, Hurwicz and Myerson for their contributions to mechanism design, it explained what the field had discovered.

A paper by Hurwicz, the committee explained, had proved the following “negative result” concerning voluntary participation: in the presence of private information, “no incentive compatible mechanism which satisfies the participation constraint can produce Pareto-optimal outcomes.” Referring to Maskin’s joint work with Jean Jacques Laffont, and Myerson’s joint work with Mark Satterthwaite, the committee members wrote: “In a large class of models Pareto efficiency is incompatible with voluntary participation, even if there are no public goods.”19 Mechanism design, it appeared, had failed in its quest for mechanisms to address market failures that would have Pareto-efficient outcomes while respecting preference neutrality and voluntary participation.

A good example of the problems that account for these “negative results” is illustrated by one-shot buyer-seller interactions called “double auctions” because both buyers and sellers are looking for the counterpart who will give them the best price. In a double auction, prospective buyers and sellers are paired, each knows how much they value the good held by the prospective seller, and they simultaneously announce prices: for the seller, the least price at which she will part with the good, and for the buyer, the highest price he is willing to pay. If the seller’s announced price is less than the buyers, then a trade will occur, the actual price being either of the two announced prices or some price in between. If the seller’s announced price exceeds the buyer’s, then no transaction takes place.

The double-auction setting is important because it is generally thought that Pareto efficiency is not difficult to sustain in this type of interaction, even without the help of a mechanism designer, because it features none of the familiar impediments to efficient bargaining and market exchange, such as ill-defined property rights, incomplete contracts, or, as in the case of public goods, the nonexcludability of some aspect of the goods involved. Barring these impediments, a large number of buyers or sellers knowing only their own preferences and the prices offered or asked by their trading counterparts should be able to bargain their way to an efficient allocation, eventually exhausting all potential gains from trade and hence, by definition, implementing a Pareto-efficient outcome.

But surprisingly, this is not the case. The problem is that when traders meet, they have no incentive to reveal their true valuations of the goods that may be exchanged, since, as is plausible to assume, these stated valuations will influence the eventual price at which they transact if they make a deal.

Knowing this, the prospective seller of a good will overstate the lowest price at which he will sell, and the would-be buyer will understate the highest price at which she is willing to buy. As a result, the buyer’s announced price will sometimes fall short of the seller’s announced price, so no trade will take place, even when the buyer’s true valuation of the good is greater than the seller’s. This means that some mutually beneficial exchanges will not be consummated. Buyers and sellers will walk away with money left on the table; Pareto efficiency will elude the bargainers.

The key result of investigations of the double auction is that if traders care only about maximizing their expected gains, it is never in either party’s interest to truthfully reveal his or her evaluation. This is the case even if the other party is reporting truthfully. The fact that individuals may benefit by misrepresenting their preferences also prevents a benign social planner from eliciting the information he needs in order to provide incentives for the efficient provision of a public good.20

Kalyan Chatterjee invented an ingenious mechanism for which a trader’s best response (the action that maximizes her payoffs) is to truthfully report her valuations, with the result that all mutually beneficial trades occur.21 The mechanism implementing this Pareto-efficient outcome requires upfront payments between the traders that depend only on the announced values, irrespective of whether a trade ensues. The size of the payments depends on the losses that each trader’s misrepresentation of the true valuation would have imposed on the other, had the other responded truthfully. Chatterjee describes this mechanism as “the payment to each player of the expected externality generated by his action.”

The payment is effectively a tax on false representation of individual’s valuation of the good, a tax just sufficient to make truthfulness the best response. The clever part is that Chatterjee figured out how this could be done without violating the privacy of the traders’ true valuations of the good.

But as you have come to expect by now, there was a hitch. Given their true valuations, some traders would do better by refusing to make the upfront payment and instead withdrawing from the mechanism. But if they withdraw, the mechanism does not work. The mechanism therefore works only if either the parties (improbably) do not know in advance how much they value their own goods or, barring this odd situation, if participation is involuntary (violating the voluntary-participation constraint).22 This Big Brother aspect of the Chatterjee solution ruled it out.

Figure 6.1 summarizes what I call the liberal trilemma. What mechanism design has discovered is that the three conditions of liberal constitutional design and public policy—neutrality with respect to preferences, voluntary participation, and Pareto efficiency—are not generally compatible. Let’s consider each dyad—each pair of vertices of the triangle in the figure—and see why the excluded vertex is not feasible. Between each pair of liberal desiderata, the phrase in parentheses is the consequence of realizing the two desiderata connected by this side but excluding the third.

If voluntary participation is combined with preference neutrality—the base of the triangle—as we have just seen from the example of the double auction, some would-be mutually beneficial exchanges will not occur, so there will be unrealized gains from trade, violating Pareto efficiency. Pareto efficiency and preference neutrality are jointly feasible—the right-hand side of the triangle—if people can be compelled to participate in a mechanism (like that designed by Chatterjee) even if they would prefer not to, violating the voluntary-participation condition. Finally, voluntary participation and Pareto efficiency are jointly feasible if the traders were predisposed to report honestly their true valuations. Individuals would then never walk away from a potential mutually beneficial trade, but demanding that traders place a sufficiently high value on honesty requires a violation of preference neutrality.


Figure 6.1. The liberal trilemma: The impossibility of Pareto efficiency, preference neutrality, and voluntary participation

One way out of the trilemma is to abandon liberal neutrality. If the traders were altruistic, they would value the losses—not only their own but also those suffered by their trading partners because of forgone mutually beneficial trades. Because this would lead them to take account of the costs that their actions inflict on others, it would attenuate the resulting market failures. What degree of altruism would be needed to ensure that all mutually beneficial trades were implemented?

One’s intuition is that if both traders were perfectly altruistic, then they would not care at all about the price (they would treat others’ gains as equivalent to their own) and so would not misreport their true valuations. This is true. But Sung-Ha Hwang and I have shown, surprisingly, that it is sufficient for only one of the traders to be entirely altruistic; explaining why this is the case, however, would take us away from the lessons of the liberal trilemma.23

Recent research has not overturned the conclusion that preference neutrality, voluntary participation, and Pareto efficiency are not jointly possible, even in a very simple exchange of goods. Thus, even if we assume that those who design public policy are paragons of civic virtue, an assumption that the self-interest-based policy paradigm would like to dispense with for the rest of the population, mechanism design fails to produce a constitution for knaves fit for a liberal society. Among the reasons is the liberal aversion to requiring people to participate in economic transactions against their will, and to allowing states to favor some kinds of preferences over others.

Another source of the trilemma is the limited extent to which intrinsically private information may be placed in a mechanism designer’s hands. This is not only true as a practical matter but also, at least in a liberal society, a highly valued limit on the authority of states. Machiavelli stated the core message of the trilemma in his Discourses when he concluded that inducing the truly bad to act as if they were good “would be either a very cruel endeavor or altogether impossible,” and would require “extraordinary means, including violence and a resort to arms.”24

The literature on mechanism design reads like one piece of bad news after another. The negative results were important contributions, and as they accumulated, some in the field beat a strategic retreat, weakening the standard of efficiency in order to resolve the trilemma by the stroke of a pen. Again the Nobel committee: “Thus in settings where participants have private information, Pareto optimality in the classical sense is in general not attainable, and we need a new standard of efficiency.”25 Thus, mechanism design has (for the most part) abandoned the standard Pareto criterion in favor of “incentive efficiency,” which simply means “the best that can be done given the way people respond to incentives,” or in other words, given existing preferences. By this new standard, an outcome like the unrealized gains from trade in the double auction would be deemed efficient. But it is clear that the operation succeeded only because the patient disappeared.

This seemingly technical watering down of the field’s time-honored objective is a concession to the inconvenient fact that preferences indeed matter. We saw how they mattered in the case of the hypothetical team of software developers, in which a bit of altruism among the members would have attenuated their failure to reach an efficient outcome. Many of his fellow economists may now agree, perhaps ruefully, with Tim Besley: “Perhaps then the solution can only lie in creating better people.”26

The time has come, the Legislator muses, to take a second look at liberal neutrality.

A Second-Best World

The Legislator thus has some difficult tradeoffs to consider. The trilemma has affirmed that he was right to worry about the nature of people’s preferences. From earlier chapters, he knows that policies that would more effectively harness self-interest to public ends may compromise the ethical and other-regarding preferences on which the success of the Legislator’s constitution must also depend. And as we will see, the reverse is also true: policies that support the proliferation and expression of ethical and other-regarding motivations will sometimes reduce the effectiveness of explicit incentives in implementing efficient outcomes.

The resulting conundrum facing the Legislator is just a novel application of a venerable yet paradoxical idea in economics: policies that address market failures by moving the economy in the “right” direction may be misguided unless they go all the way and implement the ideal market and property rights on which the invisible hand theorem is based. This “all or nothing” advice to the policy maker is called the general theorem of the second best.27

Here is the idea in its original economic application. In a competitive economy of the type represented by the invisible hand theorem, recall that prices are signals to buyers that ideally measure the true scarcity of the good, as measured by its social marginal cost—that is, the cost of making an additional unit of the good available, taking account of not only the costs borne by the producer and seller of the good (the private marginal cost), but also any costs borne by all others.

The key assumption under which the theorem is true—that everything that matters in an exchange is subject to a contract enforceable at no cost to the parties of the exchange—ensures that competition among buyers and sellers will result in prices equal to the social marginal cost of each good. In other words, everything that matters will have a price, and the price will be right. Where the social marginal cost of a good exceeds its private marginal cost—for example, because of the costs of environmental damages for which the producer is not liable—the mechanism designer can tax the good by an amount equal to these excluded costs so that the price (now including the tax) will equal the social marginal cost.

But suppose there are two violations of the “price equals social marginal cost” rule. Imagine, for example, that a certain firm has a monopoly on some good; it restricts sales and profits by setting prices above the marginal cost of production. This is the first market failure.

The firm’s production of the good contributes to environmental degradation, so the private marginal cost of production to the firm’s owners is less than the social marginal cost. This is the second market failure.

Now consider what happens if we address one of these market failures—for example, by breaking up the monopoly into smaller competitive firms so that the excess of price over marginal cost will be reduced. The second-best theorem shows that this may take the economy further away from an efficient outcome. The reason is that the competitive firms that would then make up the industry would produce more than the erstwhile monopoly because, not being monopolies, they would not benefit by restricting sales. They would expand production until their marginal cost equaled market price. This would correct the standard problem of monopolies: firms sell too little in order to profit from high prices. But the increased production would worsen the environmental problems. Letting the company remain a monopoly might have been a better policy. If adopting optimal antitrust and environmental policies simultaneously (“going all the way”) is for some reason precluded, one cannot be sure that either policy alone would improve things and could well make things worse.

The intuition behind this result is that distortions caused by violation of one of the efficiency conditions may be attenuated by offsetting distortions induced by other violations. The remarkable result is that bringing the economy closer to fulfillment of the standard efficiency conditions may result in a net loss of efficiency.

A similar result arises from the nonseparability of incentives and social preferences. It follows from a now-familiar logic. Where market failures arise because contracts are incomplete, socially valuable norms like trust and reciprocity may be important in attenuating these market failures. In these cases, public policies and legal practices that more closely approximate the idealized incentives associated with complete contracting—fines imposed for insufficient back-transfers in the Trust game, for example, or for lateness at the Haifa day care centers—may exacerbate the underlying market failure by undermining these norms. The result is a less efficient allocation.

The theory of mechanism design has led us to doubt whether efficient outcomes can be implemented in an entirely self-interested population. So norms such as trust and reciprocity will remain socially valuable under any conceivable set of mechanisms, because there is no constitution for knaves that works. “Going all the way” is not an option. Aristotle’s Legislator thus lives in a second-best world in which interventions that would be appropriate in an ideal world may not only underperform. They may make things worse.

One of these trade-offs stems from the cultural downside of policies designed to perfect the workings of markets and to extend their role in determining how a society uses its human and material resources. To understand these concerns, we return to the idea in the previous chapter that markets and other social institutions are teachers—that is, they are environments that induce people to learn new motivations and to discard old ones. The concern now troubling the Legislator is that markets that approximate the ideal assumptions of the invisible hand theorem might provide an inauspicious environment for learning social norms, which, we have seen, are essential to market functioning.

Look at markets through this lens. To see how markets—at least those in economics textbooks—differ from other institutions, I will define two of the dimensions that characterize institutions as learning environments: first, interactions may be either durable or ephemeral, and second, they may be either personal or anonymous. A key characteristic of markets, first advanced by Max Weber and since stressed by Buchanan and other advocates of an expanded role for markets, is that they require neither personal affection nor long-term relationships among people engaged in exchange.28 Markets, in this view, work well when they are both ephemeral and anonymous.

Contrast this with what the mid-twentieth-century sociologist Talcott Parsons called the “two principal competitors” of the market: “requisitioning through the direct application of political power” and “non-political solidarities and communities.”29 Parsons was not good at naming things: we can call these two allocational systems, respectively, states and communities.

The former (“requisitioning through political power”) is in some respects as impersonal as a market—at least ideally. But there are differences too. Membership in a state is not typically something one chooses, but is instead an accident of birth. Entry and exit costs are high (often requiring a change in citizenship or at least residence). Moreover, the interpersonal contacts through which state allocations work are far from ephemeral.

Unlike states or markets, directly interacting communities with stable membership—often described by the terms “organic solidarity,” “clan,” “generalized reciprocity,” or “gemeinschaft”—are neither ephemeral nor impersonal.30Parsons’s “non-political solidarities and communities” are based on long-term face-to-face interactions among known partners.

Table 6.1. Idealized markets as learning environments


Note: The rows refer to the extent to which parties to an interaction may expect future interactions, while the columns refer to the irrelevance or importance, respectively, of the individual’s identity as an influence on the transaction.

Table 6.1 presents these three ideal types, along with a fourth: ephemeral and personal social interaction, in which one’s race or some other marker of identity is important. Racially segmented day-labor markets are an example, since they are personal (the participants’ ethnic identities matter), but the contact among participants is not ongoing.

Seen in this light, markets do not look like a school for generosity or concern for others. In their study of generosity toward the needy, George Loewenstein and Deborah Small found that emotional concern is greater for “victims who share our own affective state, who are geographically or socially proximate, who are similar to us or are presented in a vivid fashion.”31 It is rightly called a feature of markets, not a bug, that they allow for mutually beneficial interactions among individuals who share none of this closeness or vividness. But this aspect of market interactions may also attenuate concern for the good or harm to others arising from one’s actions.

Although many markets are impersonal or ephemeral—spot markets for day laborers, for example—labor markets also include the lifetime employment famously practiced by some Japanese firms, and the intimate relationships within a small family-owned firm.

To see why these differences matter, return to Adam Smith’s observation that merchants are trustworthier than ambassadors. His reasoning was that merchants are likely to interact repeatedly with many people, each of whom knows how the merchant treated others. Cheating one of them, Smith observed, would result in a costly loss in reputation for the merchant. Ambassadors, on the other hand, whose interactions are more ephemeral, “can gain more by a smart trick than they can lose by the injury which it does their character.”

Smith versus Smith

Smith’s reasoning about the probity of merchants and the untrustworthiness of ambassadors seems to be correct. I am not sure whether he is right about the results, but he could be: don’t forget that Costa Rican CEOs were more reciprocal than students in the Trust game, and many in the diplomatic corps at the United Nations are no paragons of parking probity. Also, we will see from a series of experiments and natural observations that the more that markets deviate from the spot market, with its idealized “flexibility” and arm’s-length impersonality, the greater is the tendency to form loyalties with exchange partners and to reciprocate their fair treatment and trust.

The sociologist Peter Kollock wanted to know how the extent of trust among people might depend on the kind of market in which they interacted. He investigated “the structural origins of trust in a system of exchange, rather than treating trust as an individual personality variable.”32 He wanted, that is, to see how trust might endogenously emerge in a market. He designed an experiment in which goods of variable quality were exchanged. In one treatment, the quality of the good was subject to contract (enforced by the experimenter), and in another treatment it was not, so quality became a matter of trust. He found that trust in and commitment to trading partners, as well as a concern for one’s own reputation and others’, emerged when product quality was not subject to contract, but did not emerge when quality was contractible.

Like Kollock, Martin Brown and his coauthors used a market experiment to explore the effects of contractual incompleteness on trust and loyalty among traders.33 The goods exchanged varied in quality, and higher-quality goods were more costly to provide. In the complete contracting condition, as in Kollock’s study, the experimenter enforced the level of quality promised by the supplier, while in the incomplete contracting condition, the supplier could provide any level of quality, irrespective of any promise or agreement with the buyer. Since buyers and sellers knew the identification numbers of those they interacted with, they could use information acquired in previous rounds to decide with whom they wished to interact, the prices and quality to offer, whether to switch partners, and the like. Buyers could make a private offer to the same seller in the next period (rather than broadcasting a public offer) if they wanted to initiate an ongoing relationship.

The complete and incomplete contracting conditions yielded very different patterns of trading. With complete contracting, 90 percent of the trading relationships lasted less than three periods, and most were one-shot. But only 40 percent of relationships under the incomplete contracting condition lasted less than three periods, and most traders formed trusting relationships with their partners. Buyers in the incomplete contracting condition offered prices considerably in excess of the supplier’s cost of providing a particular level of quality, thus deliberately sharing the gains from exchange.

The differences between complete and incomplete conditions were particularly pronounced in the game’s later rounds, suggesting that traders learned from their experiences and updated their behavior accordingly. As in the Kollock experiment, people learned to trust trading partners when contracts were incomplete and remained loyal to them (not switching when better deals were available elsewhere). This did not occur when contracts were complete.

The experiments document a synergistic relationship between incomplete contracts and social preferences. As we have seen, social preferences contribute to well-functioning markets when contracts are incomplete. In addition, the experiments show that exchanges conducted via incomplete contracts produce conditions—such as the durable and personal interactions that Smith found in the daily lives of merchants but not ambassadors—under which people tend to adopt social preferences.

Incomplete contracts cause market failures, as economists know, but they also encourage trust, which, as Arrow says, may be essential to attenuating market failures. This is part of the Legislator’s dilemma. But it could be the basis of a kind of virtuous circle: the trust that is essential to mutually beneficial exchange when contracts are incomplete appears to be learned in precisely the kinds of trading relationships that evolve when contracts are incomplete. The virtuous circle’s vicious cousin also exists, of course: lack of trust among trading partners will induce traders to make contracts as complete as possible, thereby making the evolution of trust unlikely.

We saw this in the control-aversion experiment of Falk and Kosfeld (in chapter IV): “employers” who did not trust their “employees” designed contracts that were as complete as the experiment allowed, thereby inducing “employees” to perform minimally, and fulfilling the employer’s expectations. If contracts could be made complete, then the disappearance of trust would not be an impediment. But from our exploration of the liberal trilemma, we know that this is not generally possible.

The complementarity of trust and contractual incompleteness illustrates the main message of this new application of the theory of the second best. Exchanges are durable and personal because market participants do not abandon their current exchange partner the moment a better deal appears. Economists call this “market inflexibility” and rightly (if one can set aside the problem of sustaining trust and other social norms) consider it an impediment to efficiency. But what are the consequences of making markets more “flexible,” as would occur in Brown’s experiment if traders were not allowed to know the identity of their partners, meaning that long-term loyalties could not develop? The population could well tip from the virtuous cycle to the vicious one.

To see how policies to implement a more impersonal and ephemeral ideal market may degrade the economy as an environment in which ethical or other-regarding preferences are learned, return to Smith’s claim that merchants are trustworthier than ambassadors. His reasoning can be understood as a repeated Prisoner’s Di lemma.34 We know that if social interactions are sufficiently long-lasting and people sufficiently patient, individuals with conditionally cooperative preferences (strictly: initially predisposed to cooperate and then to retaliate against those who did not cooperate in the previous round) will do better than noncooperators as long as most people are conditional cooperators. If, as seems plausible on empirical grounds, those who do well materially in a society are more likely to be copied, then long-lasting economic interactions with such individuals will support a society of conditional cooperators. The emergence and persistence of the probity of Smith’s merchants could well be explained by a similar process of cultural evolution.

But Smith’s reasoning leads in an unsuspected direction. Consider a group of villagers whose livelihood depends on a common-pool resource like a forest or a fishery that is subject to tragedy-of-the-commons-type overexploitation. These people could be the Colombian forest users whom we encountered in chapter III in the experiments of Juan Camilo Cardenas and his coauthors. Recall that these villagers cooperated especially well in sustaining their experimental “resource” when they could communicate with one another.

Suppose that in real life the resource has been sustained over centuries because most villagers are conditional cooperators, resisting the temptation to free ride on the restraint of their fellow villagers because a defection by one would provoke a defection by all. This is probably among the reasons why so many small communities of fishers, farmers, and forest users have averted the “tragedy of the commons.”35

It works because everyone in the village knows that they will interact with one another in the future, as will their grandchildren. Their common ownership of the natural asset increases the expected duration of these interactions, because those who leave the village surrender their claim on the asset. This provides the ideal conditions for effective disciplining of defectors in the management of the common-pool resource. Protected from exploitation by defectors, conditional cooperation in this population would be a sustainable social norm.

Privatizing the asset—giving each member a marketable share in the forest, for example—provides each with an incentive to sustain the resource and monitor those who would overexploit it. But by allowing the sale of one’s share, privatization also makes it easier to leave. This undermines the conditions that sustained cooperation. It reduces the expected duration of interactions, so the value of avoiding retaliation is reduced, possibly by enough to make overexploitation the more rewarding strategy. Those who remained conditional cooperators in this setting would experience frequent exploitation by their defecting fellow villagers. Privatization would thus favor the evolution of self-regarding preferences. One might not have anticipated this being the endpoint of Smith’s parable about the merchant.

Privatization versus Cooperation

The example is hypothetical, but similar processes are not difficult to come by in the field studies of historians, economists, and anthropologists. A curious twist in the history of land rights in highland Peru provided Ragnhild Haugli Braaten with a natural quasi experiment illustrating these effects.36

In an area long dominated and managed by the owners of large haciendas, a leftist military coup in 1968 initiated a series of land reforms, making farmers the de facto owners of the land they tilled. Because the government wanted to promote cooperative ownership of the land, it also initiated a time-consuming titling process in which communities were given formal joint ownership of the land. Eventually, all rural landholders’ rights were to be recognized in this manner. Though pushed in varying degrees by local land-reform officials, the recognition of joint land titles in the communidades campesinas reconocidas (as they were termed) drew little interest from the campesinos; the formal legal status of the land had no practical import for the farmers at the time. There was no market in land in any case, and land rights appear to have been understood as jointly held in all communities.

When civilian rule was reestablished ten years after the coup, the joint titling process was halted in midstream. In both the “recognized” villages and those that the military government had failed to reach before being thrown out, farmers continued their independent cultivation as de facto owners. In both sets of villages, too, a traditional community assembly of almost entirely male household heads deliberated questions of governance.

Among the assembly’s tasks was the organization of faenas, communal work parties that maintained the farmers’ complex irrigation systems, roads, public buildings, and other common resources. The assembly also specified the number of workdays each household was required to contribute and disciplined those who reneged on their community labor responsibilities. Their admonishments were backed up by the real threat that the land tilled by the free rider could be confiscated. In addition, men volunteered to help neighbors in a traditional custom of reciprocal sharing of farmwork called ayni. The joint titling appeared to change little about the farmers’ lives; most farmers did not know whether their community was “recognized.”

Things started changing dramatically in the late 1990s, however, with the introduction of formal legal titles to private individual landholdings, including the right to sell titles. For the first time, there was a market in land ownership, and farmers could use land as collateral for loans. By 2011, the Special Land Titling and Cadaster Project had issued a million and a half individual titles. But the new laws did not apply to the recognized communities, because they already held formal joint ownership. By the time Braaten arrived in the highlands to implement her Public Goods game experiments, the ownership status of each community was well known, and differences between the individually titled areas, called “private communities,” and the jointly titled areas had become apparent.

Braaten wanted to know whether the form of land ownership was associated with differences in the degree of cooperation among the campesinos. She interviewed 570 people and performed a Public Goods experiment with them; half were from seven jointly owned communities and half from eight “private” communities. Other than their different land-rights histories, the two sets of communities hardly differed in literacy, land area per household, degree of poverty, mean income, elevation above sea level, and even degree of trust (as measured by agreement with the standard survey statement “One can trust the majority of people”). But compared with those holding jointly titled land, people in the private communities volunteered less than half the number of faena days to the communal projects, and substantially and significantly fewer days to ayni, the system of reciprocal farming help, as well.

The campesinos immediately recognized the resemblance between Braaten’s experimental Public Goods game and the faenas. Among men, those from the joint-ownership villages contributed over a third more to the experimental public good than those from the private communities, controlling for individual and community characteristics. (Women from the joint-ownership communities were indistinguishable from those from the private communities, a finding that Braaten attributed to the fact that the communal governance institutions, as well as both faena and ayni work, were almost entirely male activities.) She concluded that the “recent formalization of individual land rights has … weakened the traditional forms of cooperation.”

As with individual land titling, the development of modern labor markets in the Peruvian highlands appears to have turned the traditional contributions of communal labor into an activity for chumps. Those exploiting the exit option afforded by a regional labor market increased their payoffs by simply ignoring what had once been a community norm.37 Other ethnographic and historical studies—from India, medieval traders in the Mediterranean (already mentioned), Mexican and Brazilian shoemakers—suggest that Braaten’s conclusion may have broad relevance.38

It would be a mistake to infer from Braaten’s study and the other evidence that the privatization and clarification of land and other property rights designed to allow for more complete contracts will fail to contribute to the economic development of the communities involved. But it is safe to conclude that efforts to perfect the workings of markets may have collateral cultural effects that make people less likely to learn or retain the exchange-supporting norms and other values essential to good governance.

The Legislator’s Dilemma

The Legislator’s visits to economics faculties have left him with five uncomfortable facts about incentives: incentives are essential to a well-governed society; incentives cannot singlehandedly implement a fully efficient use of economic resources if people are entirely self-interested and amoral; ethical and other social preferences are therefore essential; unless designed to at least “do no harm,” incentives may stand in the way of “creating better people”; and as a result, public policy must be concerned about the nature of individual preferences and the possibility that incentives may affect them adversely. The last point is not something to be celebrated by anyone committed to the idea that in a liberal society the government ought to stay out of the business of cultivating some values and discouraging others. But the Legislator does not see how it can be avoided.

He also knows that in the second-best world that he faces, devising good policies in light of these facts will be a challenge. Just like the incentives in our experiments, the kinds of policies advocated by economists to make markets work more efficiently in an economy of knaves—roughly, by putting a price on everything—may compromise exactly those ethical and other-regarding motives that are essential to a well-governed society. The functioning of markets themselves may, as a result, be degraded by these market-improving policies when, as is typically the case, there is no way to guarantee every aspect of an exchange by means of an enforceable contract. Some of the economist’s standard remedies to make markets work better may make them work worse.


Figure 6.2. The Legislator’s dilemma in improving market performance

This is the Legislator’s dilemma. He sketches figure 6.2. Implementing the conditions—such as clearly defined private property rights, competition, flexibility, and mobility—under which markets would work well if contracts were complete may compromise those social norms permitting mutually beneficial exchange when contracts are not. But the economic and social institutions that promote these social norms may impede market functioning because they widen the distance between the economy and the ideal world of the invisible hand.

To figure out how to address the challenge of this second-best world, Aristotle’s Legislator returns to his roots.