Sandy Levinson argues that our Constitution is undemocratic, and that this is a bad thing.1 In his essay for this symposium and in his new book,2 Levinson’s principal targets are the hopeless malapportionment of the Senate, the strong status quo bias produced by our quadricameral lawmaking structure (House, Senate, presidential veto, and judicial review), the episodically harmful influence of the electoral college, life tenure for Supreme Court justices, the excessive aggrandizement of presidential power, and, related to the last, the Constitution’s failure to set up any real regulation of emergency powers or provision for government continuity in case of a catastrophic attack. The solution is a citizen-initiated convention that will give us a new Constitution, informed by our democratic commitments and by the last two centuries of experience with constitution-making around the world.
Levinson’s iconoclasm and his bent for present-oriented or forward-looking law, as opposed to ancestor-worshipping law, are refreshing. However, his indictment of the Constitution may not succeed even if his theoretical premises are right. Levinson offers a piecemeal critique, walking through the Constitution’s central features and condemning them one by one. This procedure overlooks that our constitutional order may be more democratic than the sum of its parts. Interactions among institutions that are undemocratic, taken one by one, can produce a kind of emergent democracy at the aggregate level. This possibility arises because second-best effects cause our Constitution’s democratic failures to offset each other, at least in part.
Put conversely, adopting all of Levinson’s proposals would produce a first-best democracy, by Levinson’s lights. But if political constraints rule out even some of his many proposals, then it need not be best, even on Levinson’s criteria, if the remainder are adopted. Where political constraints rule out full attainment of the first-best regime, the general theory of second best entails that there is no general reason to prefer moving as close as possible to the first-best.3Indeed, further departures from the first-best will sometimes be necessary to compensate for the initial departures. Unless all of Levinson’s proposals succeed, Jeffersonian democrats like Levinson might have good reason to prefer our current, democratically-flawed constitutional order to a regime that tried to approach the first best as closely as possible. For Jeffersonians, although our current constitutional democracy is imperfect, it may well be a tolerable second-best. At least that is the next conversation Jeffersonians need to have, even if they are convinced by everything Levinson says.
I. Offsetting Failures of Democracy?
Levinson argues that each of our national institutions is undemocratic, taken one by one, because those institutions fail to implement the preferences of current national majorities. Even if this is right, not much follows, not necessarily anyway. It is wrong to suppose either that (1) if the constitutional order is democratic, each or even some of its component institutions must themselves be democratic; or that (2) if some or all of the component institutions are undemocratic, the product of their interaction cannot be democratic. These are fallacies of aggregation: respectively the fallacy of division in case (1), of composition in case (2).4 An array of institutions might produce a system that is more democratic than any member of the array.
Two mechanisms might produce emergent democracy at the aggregate level, even if the underlying institutions are democratically objectionable. One possibility involves second-best offsets or compensating adjustments: an institution that departs from the democratic ideal (whatever it may be) in one respect might serve to cancel out the undemocratic effects of another institution in the array. Another possibility is that by distributing undemocratic power widely, to a range of different groups, the Constitution might produce political interactions that are tolerably democratic overall. These possibilities are speculative, but they spotlight some tensions within Levinson’s thesis.
To begin with, some of the undemocratic arrangements that Levinson condemns are themselves the cures for other arrangements that he also condemns. Levinson, for example, does not like the status quo bias of our national lawmaking system.5 The main cure for this, which culminated during the New Deal, is lawmaking by the executive branch through administrative agencies that combine legislative, executive, and judicial powers-the very fusion that the unwieldy quadricameral system was designed to prevent. The benefit of executive-branch lawmaking is that within the bounds of broad and vague statutory delegations, it bypasses the vetogates and can thus be adjusted to changing circumstances as new administrations, with new views, succeed each other over time.6 The cost is inflated executive power; critics of executive lawmaking routinely invoke the so-called “nondelegation doctrine” to argue that the administrative state dilutes “democratic accountability,” here equating democracy with congressional policymaking.7
Levinson dislikes the New Deal’s pumped-up executive, but that is just the byproduct of the New Deal’s attempt to cure the status quo bias he also dislikes. Of course the first-best, from Levinson’s perspective, would be to restrict executive power and also to abolish bicameralism or in some other way reduce status quo bias. But a regime with both status quo bias and a strong executive is better, not worse, than a regime with only one of those two features, according to Levinson’s own criteria. The piecemeal critique misses this interaction effect and thus ranks the possible regimes incorrectly.
Similar things can be said about political parties, about which Levinson seems appropriately ambivalent.8 At least under unified government, political parties lubricate the wheels of the quadricameral system and thus partially offset the excessive status quo bias, or bias against implementing the preferences of current majorities, that Levinson identifies. That effect may or may not obtain under divided government as well; but under divided government political parties at least exercise some oversight of the executive branch, which Levinson thinks especially desirable in an era of emergencies.
In related writings Levinson praises proportional representation (PR) and condemns the Anglo-American “first past the post” system of plurality voting in single-member electoral districts.9However, one of the clearer findings of comparative politics is that the combination of a presidential system with PR is dangerous one, because PR tends to produce multiple small parties and thus removes a counterweight to presidential power that a single major opposition party provides.10 This evidence comes mostly from Latin American democracies that are often shaky in other ways, but conditional on having an independently-elected executive, there is real tension between praise for PR and fear of presidential autocracy. First-past-the-post systems reinforce checks and balances in a presidential system.
Under the second mechanism I mentioned, the broad distribution of undemocratic power may produce a tolerable approximation of democracy, by allowing many groups a privileged forum in which to express their grievances, or by forcing all holders of democratically unjust entitlements to come to the bargaining table. Levinson writes that “the dreadful fact is that none of the great institutions of American politics can plausibly claim to speak for the majority of Americans, even though all assert such claims.”11 But even if the various constitutional arrangements that Levinson condemns all depart from the democratic ideal, specified in majoritarian terms, they do not all depart in the same way, or in the same direction. The varying departures may cancel each other out or even push the overall system towards, not away from, the democratic ideal.
Consider that the Senate favors small states; the Electoral College favors groups with influence in battleground states (which may or may not be small states); the administrative state favors groups who can organize to influence agencies and congressional committees; the prestige and power of the Supreme Court benefit the legal elites who feed in the Court’s wake. However, there is no one group or interest or social class that is uniformly favored by each of these undemocratic institutions. Just as the only real restraint on predation by feudal lords was competition from other feudal lords, so too the undemocratic power of the favored groups, within their domains, is checked by the fact that other groups have undemocratic power in other domains. Getting rid of all feudal lords was best, but failing that, many feudal lords would be better than one or a few.
Here the quadricameral lawmaking system does useful service by forcing many of these groups, each with power that is by hypothesis undemocratic in some domain, to argue or bargain together in order to jointly agree on national policy. Economic models of the separation of powers suggest that where several players, each with dictatorial power in its domain, must agree to a common policy, the results may improve social welfare as compared to a system that relies on elections alone.12 The domain-specific dictators, even if elected, do not act democratically in Levinson’s sense within their individual domains; because of information asymmetries and the imperfections of elections as a disciplining technology, they do not always do what current majorities would prefer. But so long as all must agree on a common policy, the interaction between or among them pushes the whole system closer to satisfying popular preferences, producing a kind of second-best democracy.
The conceptual point that failures of constitutional democracy can be mutually offsetting is not offered to explain the genesis of the relevant institutions or structures. I do not claim that one failure produced the other, although in some cases that may be so. I do suggest, in justificatory or normative terms, that the constitutional order is quite plausibly more democratic than the sum of its parts.
II. The Constitutional Status Quo as a Second-Best
There is a corollary of all this: even given everything else Levinson says, it is not obvious that Jeffersonian democrats should hope for a constitutional convention. I put aside the Burkean claim that a popular constitutional convention might produce disastrous results, such as a constitutionally mandated theocracy.13 Although that is sometimes cast as a claim about the unintended consequences of mobilizing citizens to effect constitutional change, Burkeans would not approve of the product of such a convention even if it worked precisely as intended by producing a genuinely Jeffersonian democratic order, which is Levinson’s hope.
Instead I offer a narrower point: unless all of Levinson’s prescriptions are adopted,14 even a committed Jeffersonian might prefer the current constitutional order on second-best grounds. In the examples above, the worst scenario for a Jeffersonian might arise if the convention adopts some of Levinson’s prescriptions but not all. If, for example, the convention reins in executive lawmaking but leaves quadricameralism unchanged, status quo bias will strangle current majorities; if it retains a powerful presidency but abolishes the Senate and adopts proportional representation in the new truncated Congress, a real worry about executive dictatorship would arise.
This caution, rooted in the theory of second-best, does not show that bad consequences will always result from piecemeal reform; all it shows is that piecemeal reform is not always best, because approximating the good as closely as possible can sometimes produce the worst possible outcomes.15 It then becomes necessary to go beyond Levinson by offering a detailed institutional analysis of the conditions of American constitutional democracy to see whether particular reforms will improve matters. The caution remains operative at this stage, however. Suppose that because of the causal forces that operate in politics, constitutional changes must be effected through “interdependent packages.”16 If so, then simply urging as many piecemeal Jeffersonian changes as possible may produce far worse results, even on Jeffersonian grounds, than the constitutional status quo. The first-best must not be made the enemy of the tolerable, even if we agree with Levinson that the status quo is woefully undemocratic.
* Adrian Vermeule is Professor of Law, Harvard Law School. Thanks to Frank Michelman, Mark Tushnet and Abby Wood for helpful comments, and to the Harvard Law Roundtable for helpful discussion.
 Sanford Levinson, The Democratic Deficit in America, 1 HARV. L. & POL’Y REV. (Online) (Dec. 4, 2006), http://www.hlpronline.com/2006/06/levinson_01.html.
 SANFORD LEVINSON, OUR UNDEMOCRATIC CONSTITUTION: WHERE THE CONSTITUTION GOES WRONG (AND HOW WE THE PEOPLE CAN CORRECT IT) (2006). For a similar argument, see ROBERT DAHL, HOW DEMOCRATIC IS THE AMERICAN CONSTITUTION? (2002).
 R.G. Lipsey & Kelvin Lancaster, General Theory of Second Best, 24 REV. OF ECON. STUD. 11, 11-12 (1956).
 For an introduction to these aggregation fallacies, see generally Adrian Vermeule, The Judiciary is a They, Not an It, 14 J. CONTEMP. LEGAL ISSUES 549 (2005).
 LEVINSON, supra note 2, at 29-49.
 Cf. Chevron U.S.A v. Natural Res. Def. Council et. al., 467 U.S. 837 (1984); Motor Vehicle Mfrs. Assn. v. State Farm Mut. Automobile Ins. Co., 463 U.S. 29 (1983).
 David Schoenbrod, Politics and the Principle that Elected Legislators Should Make the Laws, 26 HARV J. L. & PUB. POL’Y. 239, 279-80 (2003).
 See LEVINSON, supra note 2, at 62-66. For an excellent account of the second-best role of parties in structural constitutionalism, see Daryl J. Levinson & Richard H. Pildes, Separation of Parties, Not Powers, 119 HARV. L. REV. 2311 (2006).
 Sanford Levinson, “Imposed Constitutionalism”: Some Reflections, 37 CONN. L. REV. 921, 929-30 (2005).
 Bruce Ackerman, The New Separation of Powers, 113 HARV. L. REV. 633, 656-57 (2000). Levinson notes this evidence. See Levinson, supra note 9, at 930.
 LEVINSON, supra note 2, at 49.
 Torsten Persson et al., Separation of Powers and Political Accountability,112 Q. J. ECON, 1175-92 (1997).
 Cf. Cass Sunstein, It Could be Worse, NEW REPUBLIC Oct. 16, 2006, at 32.
 At least the major structural proposals. The point about second-best effects does not apply to all of Levinson’s recommendations – for example, his suggestion for abolishing the bar on foreign-born citizens serving as President – but I have argued that it covers the most important ones.
 See Avishai Margalit, Ideals and Second Bests, in Philosophy for Education 77 (Seymour Fox ed. 1983).
 Vicki Jackson, Narratives of Federalism: Of Continuities and Comparative Constitutional Experience, 51 DUKE L.J. 223, 272-79 (2001). See also Mikhail Filipov, Peter C. Ordeshook, & Olga Shvetsova, DESIGNING FEDERALISM: A THEORY OF SELF-SUSTAINING INSTITUTIONS 300-01 (2004) (emphasizing that because institutions are interdependent, the constitutional designer cannot evaluate them piecemeal). Thanks to Mark Tushnet for these references.