If democracy is to be justified, it will have to be in consequentialist (or, if we prefer, “pragmatist”) terms; and as it seems prima facie implausible to think that all political and social institutions could or should be democratic in a first-order sense, only a second-order version of the consequentialist case for democracy can succeed. Democracy will have to be justified, if at all, as the best second-order decision procedure for allocating competences among first-order institutions, including institutions that are arguably or partially undemocratic when viewed in isolation, such as the administrative state, markets, and constitutional courts. Such is the core of Johnson and Knight’s argument in The Priority of Democracy, and I won’t stop to explain why it seems to me both correct and important.[^important]
Instead let me press on to sketch two political risks that second-order democracy[^democracy] will have to face. The materialization of either one would deeply compromise Johnson and Knight’s project. Now these are just contingent risks, which will be more or less grave under varying political and social conditions; I make no pretense of offering conceptual critiques or analytically necessary arguments. But in a pragmatic spirit, a central task of second-order democracy will have to be risk management. Institutional experiments and ongoing adjustments imply that managing the risks of democracy will be inescapable.
The two risks are that second-order democracy will devolve into non-democracy, and that it will devolve into simple first-order democracy. By simple first-order democracy, I mean a state of affairs in which first-order democrats prove unable to sustain or unwilling to tolerate institutional forms that embody non-democratic, countermajoritarian or technocratic decisionmaking. In either case, non-democracy or simple first-order democracy, the risk is that Johnson and Knight’s complex and carefully balanced two-level structure may topple to one side or the other, as second-order democracy proves unstable.
The first risk, that of a collapse into non-democracy, is straightforward. An old problem in democratic theory is the question whether democratic institutions might end up abolishing themselves, perhaps by a procedurally impeccable (super)majority vote among free and equal citizens who happen to believe, after suitable deliberation, that first-order authoritarianism of some sort is best in the circumstances. Perhaps the second-order democracy is embedded in a brutally competitive military and political environment, or faces an internal threat from antidemocratic forces attempting to subvert the regime, and therefore decides that a temporary suspension of first-order democracy is in order. The hope is presumably that some form of commissarial dictatorship will itself be the best way to preserve democracy in the long run.
The second-order democracy may understand full well that there is a countervailing long-run risk that the dictator will refuse to return to his farm when his tasks are done, but the second-order democracy may see that as a risk worth running, in light of the alternatives. We may assume that in the long run, over time, the epistemically best regime will be a second-order democracy, for all the pragmatic reasons Johnson and Knight give. Even so, at some particular time that very regime – precisely because it is epistemically best – may make a reasonable decision, given the information it currently has, that the long-run risk is worth incurring in order to survive the short run. Of course justifiable decisions may turn out to be erroneous, so the second-order democracy needn’t be right about any of this; indeed its decision may well prove to be a catastrophic blunder. Johnson and Knight are the first to admit that the second-order democracy is fallible, and indeed that fallibility is the precondition for pragmatist virtues such as institutional experimentalism.
This risk of democratic self-abolition is familiar, but I believe that Knight and Johnson face a particularly virulent form of the problem. It is not clear even in principle what they could say to an assembly of second-order democrats who decide that, given the short-run risks, a suspension of first-order democracy looks desirable on balance. The priority of second-order democracy, on Johnson and Knight’s account, is fundamentally epistemic. The second-order democracy is thought to do better than experts — including experts on political science and theory — at assessing the pragmatically best allocation of institutional competences over time. That entails, doesn’t it, that the normative analyst herself should recognize the epistemic superiority of the second-order democracy’s decision to incur the long-run risks of a suspension of first-order democracy — at least so long as standard procedural preconditions (free deliberation, equality of opportunity to influence politics, and so forth) are respected in the making of that second-order decision. The analyst who subscribes to Johnson and Knight’s account has cut out from under herself any epistemic standpoint from which she might object to the second-order democracy’s act of (impeccably democratic) self-immolation.
Here I think Johnson and Knight slide into an inconsistency, when they say both that (1) “democratic governments can legitimately resort to repression” so long as they do so “intelligently” and that (2) “the official reaction to antidemocratic forces … must not be based on emergency powers” (p. 284-85). The second clause seems to be an arbitrary add-on, unjustified by the underlying theory that generates the first clause. If the second-order democracy decides, intelligently and quite pragmatically, that an emergency suspension of first-order democracy is the best course of action all things considered, despite the long-run risk to epistemic democracy, Johnson and Knight aren’t entitled to have any objection. But they are clearly reluctant to let first-order democracy be an entirely contingent property of the regime, one that is entirely at the mercy of whatever decision the second-order democracy makes; so they flinch from the full implications of their approach.
Let me now turn to the second risk, a partial or total collapse into simple first-order democracy. Here the risk is that the demos will be unable to tolerate the existence or authority of non-democratic institutions, even if those institutions have been chosen at the second-order level by wise democrats who seek to bind their own hands against myopia, weakness of the will, blinding passion, or other forms of collective improvidence over time.[^elster] The classical example is the episode of the Arginusae generals, hurried to their deaths in 406 BCE by an ad hoc order of the inflamed Athenian assembly. Although extant second-order rules required a jury trial and thus constrained the first-order decisions of the democracy, a critical mass of the assembly shouted that “it is shocking not to let the people do whatever they wish;”[^xenophon] and what they wished to do, right then anyway, was to order the execution en masse of a set of specialized military commanders whose innovative tactics had in fact won the naval battle at issue, although a storm had prevented them from rescuing a number of stranded Athenian sailors.
What exactly is wrong with the slogan the assembly shouted? I think there are two ways to understand the problem, one conceptual and one motivational. In the conceptual version, it is a fallacy of division to assume that if a given polity is democratic overall, then every institution and lawmaking procedure within the democratic order must itself be run on democratic lines. On the contrary, some set of non-democratic procedures or countermajoritarian institutions, such as constitutional judicial review, an appointed civil service or a specialized military corps, may be a necessary balance-wheel against unconstrained first-order democracy, precisely in order to sustain second-order democracy over time.[^constitution] Yet this fallacy of division is commonly observed in constitutional theory, where people go on endlessly about the “countermajoritarian difficulty”[^counter] posed by particular non-democratic institutions at the first order, even when the relevant institution itself has an impeccable second-order democratic pedigree. It may not be too highfaluting to attribute a similar mistake to passionate first-order democrats like the members of the Athenian assembly, who slip too quickly from the premise that the regime should be democratic to the conclusion that every decision within the regime should be made by the assembled demos.
On the motivational interpretation, second-order democracy suffers from a problem of commitment or time-inconsistency: precisely when and because the demos contains too many passionate first-order democrats, second-order democracy will be unable to commit, over time, to a fully intelligent and pragmatic choice among first-order institutional forms and procedures. In particular, pragmatically desirable countermajoritarian institutions, and technologies of collective self-binding, will be incentive-incompatible and thus out of reach.
Of course all this just identifies a risk, not an iron law, but it is a risk that has materialized over and over again in observed democracies; Johnson and Knight seem sanguine about the stability and incentive-compatibility of whatever institutional decisions the second-order democracy happens to make. The Arginusae episode is dramatic, but the same issue constantly arises on a smaller scale in modern polities. Sensible technocratic decisions by risk-regulation experts, for example, are constantly vulnerable to override by inflamed first-order democratic publics,[^cascades] even if second-order democrats would prefer that risk-regulation experts enjoy substantial autonomy from irrational first-order public beliefs and emotions.
The two risks I have identified have a common implication. At bottom, the public psychology of a second-order democracy threatens to be intrinsically unstable. It is not clear whether it can remain both second-order and a democracy; one or the other feature may have to go. The Athenians who (may have) shouted that “it is shocking not to let the people do whatever they wish,” in a first-order sense, are the predictable result of a political theory that endorses democracy in a second-order sense. The endorsement tells the people that they are competent to manage complex questions of institutional allocation, and it is not obvious — even when it is true — why that high level of competence does not also entail competence to make first-order decisions. Anticipating this, officials within the regime, especially elected officials, will shy away from expounding upon the limited first-order competence of the people. Let me give Walter Bagehot the last word on the subject:
A people never hears censure of itself. No one will tell it that the educated minority whom it dethroned governed better or more wisely than it governs. A democracy will never, save after an awful catastrophe, return what has once been conceded to it, for to do so would be to admit an inferiority in itself, of which, except by some almost unbearable misfortune, it could never be convinced.[^bagehot]
[^important]: For the notion of a “second-order assembly of democrats” designing first-order institutions and allocating powers among them, see Adrian Vermeule, MECHANISMS OF DEMOCRACY: INSTITUTIONAL DESIGN WRIT SMALL 246 (2007).
[^democracy]:“Second-order democracy” is my term, whereas Johnson and Knight refer more elaborately to the “second-order priority of democracy.” I use the former phrase merely as shorthand, for whatever institutional arrangements they envision as instantiating and enforcing the latter.
[^elster]: For an intrapersonal debate over the feasibility and desirability of collective self-binding, compare Jon Elster, ULYSSES AND THE SIRENS (1984) with Jon Elster, ULYSSES UNBOUND (2000).
[^xenophon]: At least in Xenophon’s telling. See XENOPHON, HELLENICA I-IV (Trans. Charleton L. Brownson) (Loeb Classical Library 1968) pp. 71-85. For an overview of the episode and the failure of the graphe paronomon procedure that was supposed to enforce the second-order commitment, see Adriaan Lanni & Adrian Vermeule, Precautionary Constitutionalism in Ancient Athens 19-20 (available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2115570 ). For my purposes it isn’t important whether (1) the episode actually occurred as Xenophon claims, or (2) whether the episode was militarily disastrous for Athens, as the conventional wisdom has it. (For the view that the Athenian democracy partially recovered and managed to create new and stable mechanisms of democratic self-constraint in the 4th century BCE, see Josiah Ober, DEMOCRACY AND KNOWLEDGE: INNOVATION AND LEARNING IN CLASSICAL ATHENS (2008)). Whatever the historical truth about those matters, Xenophon’s version provides a usefully vivid thought-experiment to illustrate one type of risk that afflicts second-order democracy.
[^constitution]: See Adrian Vermeule, THE SYSTEM OF THE CONSTITUTION 50-56 (2011).
[^counter]: See Alexander Bickel, THE LEAST DANGEROUS BRANCH 16 (1962).
[^cascades]: See, e.g, Timur Kuran & Cass R. Sunstein, Availability Cascades and Risk Regulation, 51 STAN. L. REV. 683 (1999).
[^bagehot]: Walter Bagehot, THE ENGLISH CONSTITUTION 36-37 (Oxford World’s Classics ed. 2001).