1. INTRODUCTION
To adjudicate among competing political philosophies, or competing policy prescriptions, we must delineate the feasible set. In other words, we must decide how utopian we are willing to be. To give a simple example, pure communism may sound good as an abstract ideal. But we reject the idea of a world without scarcity as excessively utopian. We instead opt for some more practicable vision of how the world ought to be.
Similarly, we all dismiss Charles Fourier's belief that socialism would bring oceans of lemonade and ship-pulling dolphins. Nonetheless, consider Milton Friedman's proposal to eliminate milk price supports. We can imagine some voice, call it the dairy lobby, defending the status quo. What if the dairy lobby were to invoke excess utopianism against Milton Friedman? After all, eliminating milk price supports does appear to be more utopian than maintaining the status quo. Can we dismiss laissez-faire in milk, as we dismissed Fourier, on the grounds that it is outside the feasible set? After all, it has not happened to date.
Friedman (Reference Friedman and Boaz2002), in his essay for the Cato Institute, laid out the general dilemma starkly, albeit unintentionally. His essay title referred to “Markets – the Ultimate Free Lunch.” He argued that market economies capture gains from trade and can make everyone better off, if only we would rely on them more. But if markets are such a free lunch, why are they not more popular? What hidden cost of instituting markets has been ignored? Has Friedman not already told us elsewhere that “There Is No Such Thing as a Free Lunch”? Is not asking for more markets simply another kind of utopianism, no more relevant than a plea for a free lunch?
The basic dilemma is this: Many reform proposals wish to have it both ways. They require that some degree of utopianism is acceptable. An underlying premise is that we should advocate good outcomes for their own sake, without necessarily predicting their adoption. Without this willingness to be at least somewhat utopian we cannot elevate a good reform proposal above the status quo. Nonetheless there will exist other better, yet more utopian, proposals. We reject (many of these) proposals by arguing that they are excessively utopian. Why do some reform proposals stand within the feasible set and others not?Footnote 1
At the textbook level, economists use the idea of a budget constraint to delineate the utopian from the feasible. In this view “moving along the budget constraint” (reshuffling resources) is feasible, whereas “wanting the budget constraint to shift out” (i.e., more resources for nothing) is excessively utopian. But this distinction begs the question. A society cannot move from one point along a budget constraint to another point without cost. The resources measured by the budget constraint are all owned and controlled by various agents, and in the absence of interference agents will allocate these resources one way rather than another. To ask for one allocation rather than another is to stipulate that some of these constraints and incentives be changed. We are simply asking for more resources, or for different resources, albeit in disguised fashion. (In essence, we are missing at least one dimension from our axes; often this dimension involves transactions costs.) The real question is what the social budget constraint looks like in the first place. Once viewed in these terms, we can no longer invoke the budget constraint as an a priori solution to the problem.
Economists believe that feasibility matters when it comes to policy or institutional design. In fact economists usually regard feasibility as a trump card of sorts. Yet this concept passes without much critical scrutiny. In practice the economist typically makes a methodological decision to treat some variables as “given” and others as “free,” depending on what feature of the real world he is trying to shed light on. However useful this approach may be for positive science, it leaves important normative questions unresolved.Footnote 2
2. WHY FEASIBILITY MATTERS
Following Pareto and Lionel Robbins, economists typically divide political disagreements into disagreements about matters of fact and disagreements about values. But disagreements about feasibility also can generate very different final normative stances.Footnote 3
To construct a concrete example, consider classical liberalism and social democracy, two commonly understood political categories. This example will prove useful, but as we shall see shortly, the argument is more general than any particular comparison might indicate. We require only that individual rankings of feasibility and desirability do not strictly coincide across agents; rather the two notions may conflict.
The overall desirability rankings might look like this, with the most desirable option listed on the top:
“Most desirable” rankings:
1. Oceans with lemonade and no scarcity
2. Social democracy with relatively wise and benevolent rulers
3. Classical liberal polity as it operates when found
Both the classical liberal and the modern liberal could well agree on this ordering. That is, the classical liberal might agree that a sufficiently competent and benevolent government could indeed outperform an alternative with smaller government and greater reliance on the market. Nonetheless the classical liberal will not endorse no. 2, typically out of the belief that it is too utopian. The disagreement boils down to questions concerning feasibility. To see this more clearly, consider some feasibility rankings, with the least feasible option listed on top:
“Most to least utopian” rankings:
1. Oceans with lemonade and no scarcity
2. Social democracy with relatively wise and benevolent rulers
3. Classical liberal polity as it operates when found
Again, both the classical and modern liberal might agree on these rankings in broad terms. Nonetheless we can observe two sources of disagreement to justify why the classical liberal prefers no. 3 and the modern liberal prefers no. 2. First, the classical liberal might place greater weight on feasibility than does the modern liberal. That is, the classical liberal might discriminate more heavily against no. 2, on the grounds that it requires an improvement in governance capabilities. (Of course the example is illustrative, and the positions could be reversed, with the classical liberal favoring some radical change to increase the scope of markets; I do not wish to argue that modern liberalism is intrinsically a more utopian position than is classical liberalism.)
Second, the classical liberal might believe that no. 2 is very utopian, largely out of cynicism about democratic politics. In contrast, the modern social democrat will see “relatively wise and benevolent rulers” as within our grasp to a greater degree, even if current rulers are not perceived as wise or benevolent. The classical liberal therefore will favor option 3 and the social democrat will favor option 2, even though they agree fully about both facts and values. Instead the disagreements are rooted in differing notions of feasibility and its import.Footnote 4
The claim is not that the feasibility dimension constitutes the entirety of all political differences. Real world disagreements are complex and multi-dimensional and involve many issues of contention. Nonetheless differing views on feasibility suffice to generate political disagreements, even when the parties agree on the other relevant facts and values. In this regard the “feasibility dimension” of normative reasoning is critical.
Differing judgments about feasibility pop up in a wide variety of political debates. In a modern American political context, conservatives charge that human self-interest will turn benevolent-sounding social programs into corrupt destroyers of social values. In other words they claim that the left-liberal vision of elevating the poor is not feasible. Similarly, liberals claim that the realities of politics and the instability of markets will prevent conservatives from disassembling big government without chaos. Critics from both left and right charge that libertarianism would create an unstable power vacuum and could not persist. In this view the libertarian program of wishing the state would go away is no more meaningful than wishing there were no hurricanes.Footnote 5
We also might find that two economists agree on each and every particular predictive claim. If we ask “will repealing this tax increase wealth?,” and numerous other questions of this kind, we might get the same answer from each economist. The two economists nonetheless may hold very different political views, disagreeing on the overall scope of policy and political philosophy. Implicitly they may have different mental models of what is feasible and how much feasibility should count in our overall evaluations.
A skeptic might deny that we can rank outcomes, even very roughly, in terms of their degree of feasibility. In this view we have no good metric, either empirically or conceptually, for such a ranking. But such nay-saying would not solve the problems we will face. If such rankings are meaningless, we cannot define the feasible set at all. This view will turn out to have radical consequences, namely that we must either be extreme utopians or extreme conservatives. Since I will return to these options in due course, let us put the skeptical attitude to the side.
Alternatively, common sense may suggest that some points lie clearly in the realm of the non-feasible. Perhaps Fourier's oceans of lemonade simply are absurd as a policy proposal.Footnote 6 But while this common sense approach may rule out some extreme options, it does not solve the more general problem. Many practical problems involve comparisons that cannot be treated or dismissed so easily. In many cases we start with two (or more) groups of intelligent people with different notions of the feasible set, or different notions of what is excessively utopian. Perhaps one group is right and the other wrong, but neither is postulating a policy as ridiculous as oceans full of lemonade. We can put the extreme cases aside by focusing on the large common-sense range where absurdities do not arise. We remain stuck with the basic dilemma.
Finally, I do not look to formal modal logic to help resolve policy questions about what is feasible. Formal modal logic is a well-developed philosophic literature which looks at what it means to analyze or speak of “possible worlds.” Nonetheless, without intending any criticism of the genre, modal logic is not well suited to the more applied task at hand.Footnote 7
It is beyond the scope of this article to survey modal logic in its complexity, but modal logic differs from this endeavor. First, the major approaches to modal logic deal with a very broadly circumscribed notion of what is feasible or possible. For instance, it is frequently accepted that “talking donkeys,” however strange the concept may be in common sense terms, belong to the set of possible worlds. Modal logic usually operates within a broader notion of the feasible than would resolve extant debates over the feasible set in policy or political philosophy. In contrast, we are looking for a concrete method of judging feasibility, with some epistemic applicability. Second, modal logic itself presents many unsolved dilemmas, not the least of which is what the concept of “possible worlds” means. For these reasons, we do not find ready-made answers to the above dilemma in the writings of modal logicians.
3. WHERE DO WE STAND?
We might look to (at least) three approaches to adjudicate disputes over feasibility. I refer to extreme positions, “practical advocacy,” and feasibility as a matter of degree. Of these three, I see the latter as the most promising. Let us survey each in turn.
3.1 Extreme positions
The “feasibilist” and extreme utopian views each use only a single dimension to evaluate world states. They are polar reactions to the dilemma at hand.
The extreme utopian view favors what is best, without worrying about feasibility at all. We already have mentioned Fourier and his oceans of lemonade. Along different lines, John Stuart Mill defended the perfectibility of mankind as a central political vision. He believed that the quality of human understanding could rise to extremely high levels across a broad cross-section of humanity. Turgot, Marquis de Condorcet, and Herbert Spencer all believed in extreme progress and human perfectibility, albeit with varying scenarios (see Manuel and Manuel Reference Manuel and Manuel1979).
The feasibilist view does not admit utopian speculation at all and thus favors the observed status quo. Along these lines, some economists make (or verge on making) the extreme claim that everything we observe is efficient. Nobel Laureate George Stigler in particular has been associated with this view, though he never made it in print as far as I can tell.Footnote 8
Virtually everyone rejects this view and many people scorn it. Nonetheless it would resolve the above dilemmas. Since we are already doing as well as we possibly can, we do not have to worry about normative rankings. Any beneficial improvement (that we do not already have) is too utopian and thus should be dismissed as an impossible “free lunch.” In essence the social budget constraint is now a single point.
When any claim of inefficiency comes up, the feasibilist has a rejoinder: “The current state of affairs would be inefficient, if the relevant parties could bargain or trade to bring about a better outcome. But apparently they cannot. Correcting the so-called problem is too costly. The existence of the problem is efficient, once we take all constraints and all costs, including the costs of bargaining, into account. To claim otherwise is simply to wish that things would be better, a kind of utopian dreaming. Such an argument can be invoked whether the market or government is cited as the source of the supposed inefficiency. Of course the feasibilist view need not be thought of as especially optimistic. We are in the “worst of all possible worlds” as well as in the “best of all possible worlds.”
Both the extreme utopian and feasibilist views are internally consistent, so it is difficult to refute them. Nonetheless few people find them plausible or useful. So let us look for alternative options.
3.2 Practical advocacy
Another response conceives of the problem in purely practical terms. Imagine having the option of advocating a more utopian policy option, or a policy option closer to the status quo. Perhaps we should choose the advocacy that will do the most for our notion of good consequences. Of course in making such a calculation we must consider the respective benefits from each potential change, our chance of making a difference, the chance that our advice turns out to be wrong, and so on. I will refer to this answer as “practical advocacy.” Practical advocacy tells us what would be most useful or practical for us to say.
Note, however, that practical advocacy does not provide coherent rankings across policy alternatives. Instead it ranks our personal pronouncements. We can derive the claim “John should advocate X,” but this offers no demonstration that X is good, that X is better than Y, that X is feasible, that X is the best feasible outcome, and so on. In essence, practical advocacy skirts the problem by looking toward a realm – personal speech – where all statements appear feasible (“just say it!”) and only desirability matters. It is not hard to rank statements in terms of their desirability, given that all statements appear to fall within the realm of feasibility.Footnote 9
Practical advocacy runs a number of dangers. For instance practical advocacy does not restrict us to making true claims and indeed it is likely to suggest many patently false claims. Arguably all societies are based on myths and legends in religious, political, and ethnic realms. To cite an example, it may be desirable preach that wrongdoers will be sent to the fires of hell. It does not follow that we should create such tortures for wrongdoers, were we able to. Nor does it follow that such tortures exist. Yet talking about such tortures, and endorsing them, may help inculcate morality. Similarly, political order may require that many people have false beliefs about the sanctity of their nation-state. Nations would find it much harder to defend themselves in wartime under fully accurate and realistic beliefs about the morality of war and the historical arbitrariness of many national borders. The point is not to debate the empirical relevance of these examples, but rather to show that advocacy and goodness are conceptually distinct. There is in principle a conceptual gap between an argument for advocating a policy and an argument for the policy itself. We can find many cases where false advocacy will lead to good results; yet we do not, in our final account of things, wish to endorse what is being advocated.
Practical advocacy also leaves us with the possibility of conflicting recommendations. It will sometimes suggest that different individuals should advocate conflicting policies and philosophies. For instance, perhaps Robert Nozick had beneficial influence on one area of economic policy, and John Rawls had beneficial influence on another area. Practical advocacy must then be comfortable with the scenario of Nozick pushing a libertarian line, Rawls pushing a more redistributionist attitude, and no overarching perspective to adjudicate one position as being correct and the other false.
Practical advocacy is surely a relevant consideration. We should not waste our time promoting policies or world states that cannot happen. We should concentrate our efforts where we can do some good. But practical advocacy does not resolve questions about feasibility or the basic dilemmas discussed above.
3.3 Feasibility as a matter of degree
Feasibility is most likely a spectrum rather than an all-or-nothing category. Some specified world states are more utopian than others, but as a matter of degree rather than of kind.
In most cases it is easy to see how feasibility differences might be of degree. For instance it might be argued that it is feasible for citizens to give one tenth of 1% more of their income to charity. At the same time it is too utopian to expect everyone to give 70% of his or her income to charity. Human motivations would have to be “too different” for so much charity to be forthcoming. But the judgment of non-feasibility does not appear to kick in at any particular quantitative amount of giving, as could be defined in terms of an exact percentage, penny, or dollar cut-off. Rather the more giving is specified, the lower the degree of feasibility that appears to obtain.
We can see degrees of feasibility even in the Fourier vision. Forget about oceans, what if we were told that socialism would bring us a mid-sized lake full of lemonade? A small pond? A swimming pool of lemonade? And so on. Again, it is difficult to find a distinct cut-off at which the specified world states clearly and definitely cross from the realm of the feasible into the realm of the infeasible. In epistemic terms, it is unlikely that we could identify or verify such a cut-off point, even if it were to exist.
A spectrum of feasibility might offer several dimensions. A more feasible vision, compared to a less feasible vision, might be “more like” the world we know in terms of fact, more like the world we know in terms of adherence to laws of science, or more likely to come about in the future. David Lewis has suggested some standards for ranking worlds in terms of their similarity to each other. We could think of the more “similar” worlds to our own as somehow “more possible” or “less utopian.” Lewis's (1979: 472) measure of similarity involves a lexical ranking of the following four qualities:
(1) It is of the first importance to avoid big, widespread diverse violations of [physical] law. (2) It is of the second importance to maximize the spatio-temporal region throughout which perfect match of particular fact prevails. (3) It is of the third importance to avoid even small, localized, simple violations of law. (4) It is of little or no importance to secure approximate similarity of particular fact, even in matters that concern us greatly.
I do not mean to defend Lewis's particular proposal; rather it is one particular vision of feasibility as a general albeit multi-dimensional continuum.Footnote 10 More generally, we also might incorporate considerations of probability and cost into our multi-dimensional concept of feasibility.
4. HOW TO DEAL WITH FEASIBILITY AS A MATTER OF DEGREE
Some individuals might wish to construct functions with both desirability and feasibility as relevant arguments. (It is a moot point whether we call them “social welfare functions,” or whether we moving to a category above and beyond welfare by invoking the concept of feasibility as well.) Imagine a positive valued function that reflects the importance of both notions:
(1) X = f (desirability, feasibility)
Note that such a function need not be separable, but we would expect both desirability and feasibility to enter positively into the function along all relevant margins. If we had such a function, we could combine assessments of desirability and of feasibility to produce what might be called a “feasibility-weighted best outcome.”
We might be tempted to convert this function into a kind of expected value calculation, where probability stands in for feasibility. We could then multiply value outcomes by their probability to arrive at a best course of action, all things considered. But such a procedure would fail to grasp the nature of the feasibility dilemma. Degrees of feasibility do not (currently?) exist on an operational metric that could be multiplied numerically with a value dimension. We have at best broad brushstrokes of degree. Furthermore the category of feasibility – while poorly understood – is distinct from that of probability. For instance, an option can be highly feasible but extremely improbable. Working within the standard economic paradigm, consider any point along the budget constraint but not close to the point of utility maximization. The feasibility of this point will be absolute but its probability will be low; arguably its probability is zero if individuals are rational utility maximizers. Alternatively, it could be argued that the outcomes on either a two-sided or six-sided die are all “outright feasible,” even if their probabilities of appearing differ substantially (one in six vs. one in two). So we should not treat feasibility as synonymous with or somehow concordant to probability. Probability may be one relevant notion for the feasibility concept, but the link will not be a very tight one.
In any case, the prospects of drawing up a globally defined function for both feasibility and desirability are daunting. Since Kenneth Arrow, it has been hard enough to develop a commonly accepted social welfare function for desirability alone. By trying to solve desirability and feasibility together, in functional terms, we would be taking on Arrow's problem squared.
We should instead proceed in qualitative terms to develop a method for improving the quality of political debate. Individuals should specify not only their values and beliefs about facts, but also their beliefs about feasibility. In particular, they should outline two considerations. First, they should explain how much importance they assign to nominating a feasible outcome, and what degree of feasibility they assign to their own proposals. Second, individuals should outline their more particular views on which outcomes are especially feasible or not. If some political disagreements arise from feasibility differences, as outlined above, this should help resolve or at least identify those differences. The individuals involved can either proceed to debate questions of feasibility, or they can “agree to disagree.” We can think of such dialogues as a kind of iterative programming procedure, designed to identify why individuals do not always converge on the same answers to normative problems.
The particulars of these feasibility debates will involve a broad range of considerations. If we return to the comparisons of classical liberalism and social democracy from section 2 above, we can see relevant questions for investigation. First, just how infeasible are benevolent public rulers in a democracy? We could address this question by looking at a broader cross-section of democratic rulers in history. We would then study whether the benevolent ones occurred under particular preconditions, and whether those preconditions might plausibly obtain under the policy environment in question. This would help determine the size of the “infeasibility penalty” to be placed on visions with benevolent democratic rulers. That is, empirical analysis will help us determine where a proposed alternative stands along a feasibility spectrum.
Second, and with greater difficulty, we must confront the question of how utopian we should be in our political visions. This is a more subjective question, and is not so easily resolved by empirical or historical investigation.Footnote 11 We might, however, proceed by some empirical investigation. Many individuals, even if they are not literally feasibilists in the sense discussed above, see relatively little scope for contingency in human affairs. Such people might believe that even if Hitler had never been born, modern Europe would somehow have taken roughly its current form. Europe would have had the same geography, many of the same technologies, faced similar geopolitical constraints, and so on. Alternatively, other individuals believe that small changes in initial conditions could have made a huge difference to the modern world. Perhaps with no Hitler, Europe would be twice as wealthy and Russia would have developed more rapidly than did Japan, and so on.
Individuals who stand relatively close to the feasibilist position, and who see little contingency in human affairs, tend to have a lower tolerance for utopian thinking. In their view, making the world fundamentally different would require a significant change in initial conditions. Alternatively, believers in contingency should be more predisposed toward utopian thinking. Only a small change in initial conditions would be required to possibly usher in a grand improvement.
Of course these are modal claims, and while they are related to matters of fact, they are not susceptible to direct empirical testing. We cannot run historical experiments and observe more than one possible outcome; we cannot redo European history without Hitler. Nonetheless, issues of contingency are not entirely removed from social science investigation. It is well known that differing social science models imply varying degrees of contingency for future events. For instance, models of non-linear dynamics and increasing returns imply a relatively strong influence for contingent features of the environment. Other approaches, such as the Solow growth model, imply that small changes are more likely to fade into significance. The implication of the Solow model is that the initial change must affect the rate at which new ideas are produced; otherwise convergence will set in eventually. So the relative stature of these models (and others) influences the extent to which we should be willing to think in utopian terms. This is hardly a formal algorithm for solving a policy disagreement. Nonetheless it allows us to have a meaningful debate, it shifts the terms of discourse, and it can help us make some progress in the future.
5. CONCLUDING REMARKS
My goal in this paper has been to bring about greater political agreement at some meta-level. If we do not deal with feasibility dilemmas, political agreement can be impossible to generate, as demonstrated above. We might still disagree about what is feasible, or how much feasibility matters, but at least we can figure out what we are really debating.
I think of this enterprise as starting with the fact of persistent disagreement, and working to eliminate that disagreement by examining how greater consensus might come about. Recognizing the relevance of thinking clearly about feasibility is one brick – albeit only one brick – in that broader problem. So while I do not offer a generalized solution to the problem of feasibility, we can find some clues for how to think about feasibility. We have taken one step in the direction of knowledge and greater agreement on normative issues.
To sum up, we can see several implications for normative economics:
1. Facts and values, as those concepts are used traditionally, do not exhaust all possible sources of political disagreement.
2. Recognizing this would improve the quality of policy debate. Different views on feasibility suffice to generate substantial disagreements, although of course they do not exhaust such disagreements.
3. Whether an option stands in the relevant domain is often a question of degree rather than of kind, at least under current knowledge.
4. Our best current hope of progress is a kind of iterative procedure designed to identify and resolve different assumptions about feasibility.
5. While this inquiry suggests some new normative challenges, and greater agnosticism on some questions, we also are left with paths toward knowledge.
