How should the administrative state be organized from the epistemic point of view? The institutions of the administrative state must aggregate both preferences and judgements. The preference-aggregation side of the ledger is the subject of extensive literatures in law and economics, public choice, political economy and positive political theory. Theorists of preference aggregation in regulatory and administrative institutions study the industrial organization of Congress, the White House and the agencies; lobbying, litigation and other forms of individual and collective action; and a range of other activities undertaken by rational agents whose preferences differ from those of other agents, whether or not their beliefs differ as well.
By contrast, the epistemic organization of the administrative state is relatively terra incognita, with useful exceptions in the literatures on political economy and industrial organization.Footnote 1 On the epistemic side, the main issue is not the aggregation of conflicting preferences but the aggregation of diverse judgements. Information must be generated, aggregated and deployed by executive officials, administrative agencies and other actors who share important fundamental preferences but who have differing beliefs and thus have different derived preferences over policies. The breadth of the topic is daunting; I will examine only one of the major fault lines that structures debates about the epistemic capacities of administrative agencies and the administrative state more generally.
At issue is the conflict, tension or trade-off between local and global knowledge.Footnote 2 Across a variety of controversies, I will suggest, the same issue can be discerned in varying forms, whether explicitly or implicitly. Proponents of local knowledge, often taking their cue from Friedrich Hayek's work, argue that the scope of the administrative state or its internal organization within that scope should be arranged so as to privilege context-specific knowledge about particular economic or regulatory problems, especially tacit or practical knowledge. Proponents of global knowledge, among whom the best known may be Justice Stephen Breyer of the United States Supreme Court, argue for a kind of synoptic technocratic rationality that implies an expansive scope for the administrative state and, within that scope, attempts to maximize the epistemic coordination of regulatory approaches across different agencies and different subject matters.
I will examine the trade-off between local and global knowledge at two levels. The first is the scope of the administrative state's regulatory jurisdiction; this is the large-scale question of government versus markets which is central to the Hayekian program. The second level is the internal organization of the regulatory bureaucracy within the area committed to the administrative state's regulatory jurisdiction. Here, the industrial organization literature, as we will see, has adapted Hayekian questions to new settings. These two levels – the scope of the administrative state and its internal organization – parallel a conventional distinction in the industrial organization literature between the scope of firms and the internal structure of firms.
The framing and analysis of these two questions is meant to have value independent of the answers I will offer. However, once the trade-offs are stated and evaluated, I go on to suggest that the Hayekian arguments for local knowledge fare poorly at both levels. At the level of the scope of the administrative state, the Hayekian position emphasizes the benefits of local knowledge and adaptation to the contingencies of time and place, but it overlooks or downplays a major trade-off: centralized synoptic regulation is indispensable for epistemic coordination. Spill-overs, externalities and lost opportunities for economic synergy may arise not only because of conflicts of interest and problems of collective action but also for epistemic reasons: actors with thick localized information may be myopic about what other actors are doing. Consequently, a major role for synoptic national regulation is not (only) command and control but epistemic co-ordination and the creation of common knowledge – measures for generating and sharing information that dispel the local myopia of market actors.
Furthermore, even setting this trade-off aside, Hayek and many of his successors overlook that markets are just one institutional arrangement for aggregating local information. The administrative state itself can and actually does embody a range of institutions for aggregating thick local knowledge, including the tacit practical knowledge whose importance Hayek underscored. Congress itself is a summation of representatives with local knowledge of dispersed constituencies, while the administrative agencies often incorporate actors with industry- or area-specific skills and information. The administrative state deploys far more than abstract or statistical technocratic expertise; rather, it has developed a representative bureaucracy devoted to the gathering and exploitation of local knowledge.
At the second level, the internal organization of the bureaucracy, the argument for local knowledge underscores that front-line agencies may have more issue-specific expertise and tacit practical knowledge than centralized reviewing institutions, such as the Office of Information and Regulatory Affairs (OIRA), can ever possess; centralized reviewers must always be boundedly informed. Yet, line agencies are also boundedly informed on other dimensions, as the very context specificity of their expertise makes them prone to overlook synergies and spill-overs across agencies and regulatory areas. The result is a problem of epistemic coordination among the regulators. This problem can be solved in principle either through centralized top-down coordination by a body such as OIRA or through decentralized bottom-up coordination by agencies interacting horizontally. Under a realistic assessment of the conditions of the American administrative state, however, top-down epistemic coordination will prove the superior approach. OIRA aggregates and coordinates dispersed information – information that is dispersed around the bureaucracy rather than society – and does so in a manner that cannot be replicated by horizontal coordination among agencies given background features of the federal government.
When both the scope and organization of the administrative firm are considered simultaneously in light of the knowledge problem, it becomes clear that certain combinations of views are inconsistent. We will see that Hayekians attempt to justify, on free-market grounds, the creation of a central planner for regulation – a synoptic overseer of the bureaucracy. I will suggest that such a justification rests on inconsistent premises about local knowledge and thus fails. Yet, there is no inconsistency in believing, on epistemic grounds, both that the administrative state should have a robust scope and that an OIRA-style centralized synoptic coordinator is necessary to oversee the bureaucracy; I will press this combination of views instead.
The first part of this chapter examines the scope of the administrative state and the trade-off between local knowledge and centralized epistemic coordination; a principal claim is that Hayekian arguments for a constrained administrative state overlook the ability of non-market institutions to aggregate local and tacit knowledge. The second part examines the internal organization of the bureaucracy and claims that under realistic conditions, at least in the American administrative state, the trade-off between global or synoptic coordination and local knowledge will have to be resolved by a centralized coordinating body such as OIRA. The third part brings the two halves of the discussion together by asking whether a theoretically consistent Hayekian may justify OIRA-style centralized oversight of the bureaucracy as a measure for protecting the dispersed local knowledge generated by free markets. In my view, this approach is logically untenable, although a robust role for OIRA is justifiable on other grounds. A brief conclusion follows.
The scope of the administrative state
Assumptions
The first question involves the scope of the administrative ‘firm’ – the area of regulatory jurisdiction that the administrative state will oversee. From an epistemic standpoint, the scope of the administrative state should be chosen so as to aggregate information and judgements in whatever way optimally promotes commonly agreed-upon ends. To get traction on these questions, we will need to make some simplifying assumptions.
First and foremost is the assumption that there are commonly agreed-upon ends somewhere in the picture. In some domains, there are not, of course; it is value conflict all the way down. In other domains, however, the questions about the performance of the administrative state are questions about how well agencies and the executive branch generally pursue social goals admitted by nearly everyone to be valid and whose ranking vis-à-vis other goals may be widely shared as well. Even where there is disagreement about goals, there is typically unanimous agreement that goals should be pursued in the most cost-effective manner possible, so the choice of efficient means is a matter of common interest even if goals are contested.
Here two points are important. Firstly, Hayekians, in contrast to more radical libertarians and anarchists, do not deny the existence of commonly agreed-upon ends,Footnote 3 although they believe that the administrative state does a poor job of promoting those ends. Secondly, observed disagreement among political or economic actors or even bitter conflict among such actors does not at all entail that there is any conflict of fundamental or bedrock preferences. Strictly epistemic differences arising from differing information or beliefs – beliefs about causal processes, means-ends judgements and the like – may produce deep disagreements about optimal policies. Derived preferences – preferences for particular policies – may diverge because of differences in beliefs, even if relevant actors share deep or fundamental preferences in a given area.
Some legal assumptions are also necessary to frame the issue. I assume that, as a first approximation, Congress may (1) regulate any aspect of the national economy and (2) delegate to the executive branch jurisdiction over anything which it could regulate directly. Both assumptions abstract from an elaborate body of constitutional law, which respectively establishes the bounds of Congress’ regulatory powers and the limits of its ability to delegate power to the executive. However, it is conventional wisdom among constitutional lawyers that both bodies of doctrine are impressively capacious; barring unusual circumstances,Footnote 4 Congress may regulate any and all economic matters either directlyFootnote 5 or indirectly through administrative delegation.Footnote 6
Finally, I assume that the alternative to national regulation is ‘the market’. This assumption is made strictly to engage Hayekian arguments against the administrative state that rest on an opposition between ‘government’ and ‘markets’. In federal systems such as that of the United States, of course, the opposition is somewhat misleading. A third alternative is federalism – regulation by states and localities, which holds out the promise of being decentralized but not market based. Absent some constitutional restriction, state governments can regulate even if the national government cannot or will not, so the market can be constrained by state regulation as well as federal; indeed, in a number of domains, federal regulation attempts to clear away state regulation or to impose national uniformity in order to reduce legal uncertainty or to unshackle markets.
However, absent national regulation, there is some class of activities that states cannot effectively regulate because economic actors with mobile assets enjoy low costs of exit; in the extreme cases, no state can regulate more than does the least-regulatory jurisdiction (the ‘race to the bottom’). And activities by market actors in one state may affect both market actors and regulators in other states, as in the case of pollution. These inter-state spill-overs imply some irreducible role for national coordination and regulation, so it is not entirely misleading to contrast national regulation with markets. In any event, I mean to bracket questions of federalism; the simplifying assumption allows me to more cleanly identify two problems with Hayekian arguments from local information. Those problems would also arise in many cases in which state rather than federal regulation is contrasted with markets, so bracketing questions of federalism does not seriously distort the substantive issues.Footnote 7
Trade-offs: local and global knowledge
Hayek's 1945 paper, ‘The Use of Knowledge in Society’,Footnote 8 argues quite brilliantly that central planning suffers from two problems, not one. Most obvious is the problem of incentives, created by the self-interest of the planner. Thus, undergraduates are prone to say that a genuinely benevolent dictator would be the best of all possible worlds in politics, but Hayek argues that even a benevolent dictatorial planner would fail due to inadequate information.
Hayek's main argument runs this way: the central planner may call upon experts who possess specialized technical information, usually embodied in statistical form, but economic actors possess dispersed, disaggregated information about the ‘particular circumstances of time and place’.Footnote 9 This sort of information about a myriad of rapidly changing variables is indispensable for on-going adjustments in a relentlessly dynamic economy; this is the Austrian side of Hayek. Crucially, the information takes the form of tacit practical knowledge, knowing how rather than knowing that. The key feature of this sort of information is that it cannot be transmitted to the planner, at least not at sufficiently low cost or in a sufficiently timely manner; according to Hayek, it is ‘knowledge of the kind which by its nature cannot enter into statistics and therefore cannot be conveyed to any central authority in statistical form’.Footnote 10 Thus, the planner, no matter how many experts he or she has on tap, will be unable to supply plans or even regulations that are sufficiently well adapted to the ever-changing localized problems of the economy. As a shorthand, I will call this highly contextualized and non-transmissible type of knowledge ‘local knowledge’; James Scott uses the Greek term metis.Footnote 11
Hayek is aware that market actors need to coordinate their behaviours with one another and that the flip side of local knowledge is a kind of informational parochialism or myopia. ‘[T]he “man on the spot” cannot decide solely on the basis of his limited but intimate knowledge of the facts of his immediate surroundings. There still remains the problem of communicating to him such further information as he needs to fit his decisions into the whole pattern of changes of the larger economic system.’Footnote 12 Thus, there is a need for coordination across local contexts and local knowers, but as I will discuss more extensively in the second part of this chapter, coordination need not entail centralization; there can be decentralized coordination. Hayek puts his faith in horizontal rather than vertical coordination – in coordination achieved through a myriad of decentralized interactions rather than through top-down commands. Market competition is a ‘discovery procedure’ that gives each actor the information necessary to adjust his or her plans to those of others.Footnote 13
For Hayek, the critical mechanism that produces informational coordination in markets is the price system. On this view, prices are a kind of summary statistic which impounds a bewildering array of variables, drawn from many localized decisions, and conveys them in useful form to other localized decision makers. The effects of any given actor's decisions on other actors, and vice versa, are communicated through prices, and this suffices to dispel the myopia of locally adaptive agents who have thick knowledge of their own circumstances but limited informational horizons.
The problems with this view are manifold. Firstly, under a range of conditions, prices will supply highly misleading information about the real cost of resources; a stock example is monopoly, in which the monopolist sets prices above marginal cost and thus, in effect, sends a distorted signal about the social cost of the goods the monopolist supplies. Hayek therefore allows that regulation to curb monopoly is valid,Footnote 14 but this opens the door to other arguments that real costs diverge from perceived costs as embodied in prices. Secondly, even where there are no problems of externality or monopoly, genuine coordination problems may arise, an example being the choice of ‘compatibility standards’ for products and technology.Footnote 15 In such problems, there exists more than one mutually beneficial equilibrium for market actors; prices will not tell the actors which equilibrium to adopt. Thirdly, as emphasized by Karl Popper, administrative experimentation and intervention – Popper's ‘piecemeal social engineering’ – is precisely what generates new information and allows agencies to learn by doing. On this view, even if Hayek's diagnosis is correct, his prescription is backwards; profound ignorance about complex social and economic processes itself counsels in favour of (constrained) intervention and experimentation by regulators rather than in favour of free-market liberalism.Footnote 16 At a minimum, there is no epistemic warrant for a global presumption either for or against regulatory intervention and experimentation; intervention has informational benefits as well as risks of unintended consequences and unexpected costs, so interventions should be judged locally on their particular merits.
Finally – and this is the most general problem of all – it has been shown that as long as information is costly, no market can be fully informationally efficient, even in principle.Footnote 17 In simplified terms, the reason is that market actors will have incentives to acquire costly information only if they can profit from it. But if they can, then – precisely on Hayekian grounds – their information will be transmitted via the price system to competitors who may profit as well, even though they have not made costly investments in information. Anticipating this free-rider problem, no one may invest in information, but if this occurs, then everyone has incentives to do so – and so forth. In other words, the costliness of information creates a free-rider problem that has no general equilibrium solution, implying that markets for information are intrinsically unstable. The more efficient the price system becomes at conveying information, the worse the free-rider problem becomes, the less efficient the price system will be and so on in a circle. Hayek's appeal to the price system to give an informational rationale for free markets is in this sense self-defeating.
The upshot is that decentralized coordination through the price system cannot, even in principle, fully substitute for centralized coordination through governmental institutions. There is a real trade-off between local knowledge and adaptation, on the one hand, and epistemic coordination which attempts to take a broader synoptic overview of firm behaviour and of the economy generally, discerning spill-overs, externalities and opportunities for synergies across enterprises which go unnoticed by locally myopic actors. The administrative state is a very large institutional mechanism for pursuing epistemic coordination and managing the resulting trade-offs.
To be clear, nothing in this account limits the administrative state to operating through command-and-control regulation. Coordination often may be pursued through essentially informational measures.Footnote 18 The problem with Hayekian agents with thick local knowledge, on this account, is not that they are selfish (although they may be) but that they are locally myopic; they may overlook opportunities for cooperation with other agents and fail to understand how coordinating their behaviours with those of other agents could make all better off. Given these problems, agencies may pursue measures that collect globally relevant information and provide it to local agents, better enabling them to coordinate their behaviours with those of others. Indeed, in Hayekian vein, agencies often may be able to invent and deploy summary statistics that impound a great deal of global information about the behaviours of whole economic sectors or industries.
A key example of this technique is the promulgation, by agencies or quasi-public bodies, of various types of standards: voluntary industry standards which convey information to firms about the behaviours of other firms or technical standards where uniformity lowers the costs of cooperation and exchange among firms (the modern equivalent of uniform weights and measures). It has been shown that under plausible conditions, committees that promulgate standards – either regulatory commissions or quasi-public bodies such as the American National Standards Institute – will be more likely to produce widespread coordination than will decentralized action by ‘market leaders’ who attempt to start a bandwagon of subsequent imitation by other firms.Footnote 19 Centralized standard-setting measures amount to central planning without coercion, at least without the sort of ad hoc and discretionary administrative coercion that Hayek feared would undermine the rule of law.Footnote 20
Thus, there are real trade-offs between the benefits of local adaptation and the benefits of global synoptic coordination, especially epistemic coordination. Although Hayek was right that coordination does not logically entail centralization, the mechanism of decentralized coordination on which he relied – the price system – will not do the trick; absent any other candidate for a mechanism of decentralized coordination, the administrative state has important coordinating functions that require agencies to collect and provide synoptic information. I return to these points in the second part of this chapter, where, as we will see, the same trade-off between global and local knowledge is critical to the internal organization of the administrative state.
Nonmarket aggregation of local knowledge
The foregoing provides an external critique of the Hayekian argument for local knowledge, but there is an internal critique as well. Let us put aside the trade-off between coordination and local adaptation and suppose that local adaptation is a good that should be strictly maximized. Even on this assumption, there is an important lacuna in the Hayekian argument. The market is just one possible institutional mechanism for generating and then aggregating local knowledge. One cannot simply posit the importance of local knowledge and then conclude to the superiority of market mechanisms; rather, one must carry out an even-handed institutional comparison between, or among, the institutional possibilities. In particular, I will suggest, the institutions of the regulatory state themselves can be, and have been, justified in part as mechanisms for aggregating local knowledge.Footnote 21
Although it is not essential to my argument here, I believe that Hayek himself overlooked this point. As of 1945, Hayek repeatedly juxtaposed decentralized coordination through the price system, on the one hand, to ‘the single mind’ of ‘the planner’, on the other. But, of course, a single mind cannot encompass dispersed local knowledge. Hayek has here built his thesis on a particular kind of nirvana fallacy, comparing the worst-possible version of one institutional arrangement to the best-possible version of another. We might even dub this the Hayek fallacy: a comparison between an informationally rich decentralized market and an informationally impoverished regulatory apparatus.Footnote 22 Regulatory institutions are far from perfect aggregators of information, and no one would claim otherwise; I shall expand upon their imperfections in what follows. But the right comparison is between informationally imperfect regulatory institutions and an informationally imperfect market. As we have seen, markets are necessarily imperfect aggregators of information as long as information is costly; the problem is structural, not contingent or remediable.Footnote 23
Whatever the logical and causal problems with Hayek's argument, however, I want to focus on the affirmative case that the institutions of the administrative state themselves aggregate dispersed information and local knowledge through a variety of mechanisms. I will begin with a set of classical justifications for legislative representation and then move to the administrative state proper, in which delegation of power from legislatures to bureaucracies is the principal means of policymaking.
Legislatures and representation. Let me begin with one of the classical justifications for legislative representation. On this view, urged both by Publius and the Antifederalists in the founding era, a major function of representation is precisely to aggregate ‘local information’Footnote 24 – Hayek's ‘knowledge of people, of local conditions, and special circumstances’. Responding to the argument that federal representatives will make poorly informed decisions about taxation due to ‘want of a sufficient knowledge of local circumstances’, Publius asks, ‘[c]annot the like knowledge be obtained in the national legislature from the representatives of each State? And is it not to be presumed that the men who will generally be sent there will be possessed of the necessary degree of intelligence to be able to communicate that information?’Footnote 25
The problem of obtaining local knowledge was central to several questions of constitutional design in the founding era. One of the major debates during the ratification process centred on the House of Representatives and its representation ratio, capped by the constitutional text at no more than one representative for every 30,000 inhabitants.Footnote 26 Antifederalists held, as Publius put it, that the representation ratio of the House was too small, so ‘[the representatives] will not possess a proper knowledge of the local circumstances of their numerous constituents.’Footnote 27 Far from rejecting the normative premise of the argument, Publius embraces it and devotes a whole paper to arguing that the federal representatives actually will bring to their tasks an adequate degree of local knowledge. ‘The representatives of each state will not only bring with them a considerable knowledge of their respective districts; but will probably in all cases have been members … of the state legislature, where all the local information and interests of the state are assembled, and from whence they may easily be conveyed by a very few hands into the legislature of the United States.’Footnote 28
There is no logical inconsistency between this argument and Publius's better-known defence of representation and elections as a selection mechanism that ‘refine[s] and enlarge[s] the public views’Footnote 29 – yielding a set of representatives who possess a kind of synoptic understanding of the public good of the whole polity. The latter argument suggests that federal representatives will be well positioned to promote synoptic coordination, but Publius also thinks they will be well positioned to aggregate local information. The argument is cast in the alternative, and the alternatives are not logically exclusive. It is fair, however, to discern a pragmatic trade-off between the two aims of synoptic understanding of the federal law-making system as a whole, on the one hand, and local knowledge, on the other, in so far as time is a scarce resource and the time spent by federal legislators on acquiring one type of knowledge is time not spent on acquiring the other.
So far, we have seen that a leading justification for representation in The Federalist emphasizes the local character of the knowledge held by representatives, although Publius is candid that the benefits of local information trade off against other goods. A closely related idea, one not prominent in Publius's argument, emphasizes the dispersed or decentralized character of the knowledge held by representatives. One of the most piercing critics of representative democracy, Carl Schmitt, distilled the classical liberal account of representation by arguing explicitly that representation aggregates dispersed information. As Schmitt put it, ‘Parliament is … the place in which particles of reason that are strewn unequally among human beings gather themselves and bring public power under their control.’Footnote 30 It might be backwards to say that Schmitt's terms are explicitly Hayekian; although we have no information on the transmission of this particular idea, the channel of influence between the two thinkers seems to have run generally from Schmitt to Hayek.Footnote 31 Hayek's conception of dispersed information in competitive markets ultimately may derive from Schmitt's conception of dispersed information in representative government.
Representative bureaucracy. A great and neglected theorist of the American administrative state, John A. Rohr, agreed with the Antifederalists that the framers plausibly erred by choosing an excessively low representation ratio in Congress, particularly in the lower and more populist legislative chamber.Footnote 32 The House has an exceedingly low ratio in historical and comparative perspective; capped by the Constitution at one representative per 30,000 citizens,Footnote 33 it currently stands much lower, at around one per 700,000.Footnote 34 It is an open question whether representatives of that sort may plausibly be viewed as possessing thick local knowledge; let us suppose that such a claim would be implausible. The general problem is that in large modern representative democracies, with high representation ratios, the local-knowledge justification for legislative institutions becomes ever more attenuated. Sheer scale dilutes the ‘local information’ that both Madison and the Antifederalists valued. Likewise, assuming that economic and social conditions change more rapidly today than in the founding era, the balance struck by Publius between local knowledge and global knowledge may be askew; the legislative terms chosen by the framers may be too long in current conditions, even if they were optimal when chosen.
Nonetheless, Rohr suggested, the administrative state has generated – whether by happenstance, evolutionary adaptation or intentional institutional design – a second-best set of institutions that at least partially compensate for the high representation ratios and slow informational updating of modern legislatures. The main mechanism is delegation from legislatures to ‘representative bureaucracy’,Footnote 35 an idea drawn from the literature on public administration. Although representative bureaucracy comes in a bewildering variety of shapes and sizes, one version of the idea is that administrative institutions may themselves build right into their structure and procedures a kind of representation that brings local knowledge into the regulatory enterprise.
Under the Administrative Procedure Act of 1946, the main mechanism for incorporating the information held by the public and by affected groups into administrative policy making is ‘notice-and-comment rule making’, under which agencies issue a notice of proposed regulation, receive and consider comments by any interested person and then issue a final decision with a published rationale that is supposed to take the comments into account.Footnote 36 Notice-and-comment rule making, however, has both detractors and defenders. The main issues in debate are whether agencies take comments into account at all or instead merely pretend to do so, and if they do, whether the process of commenting is dominated by well-organized and well-funded groups with high stakes who swamp the efforts of public-interest groups and a diffuse citizenry.Footnote 37 In the limit, the dominance of well-funded interests may result in a form of ‘epistemic capture’, in which agencies act with a marked bias in favour of industry and other regulated parties not because of corrupt motivations but because the information agencies receive is itself skewed.
Given these problems, scholars such as Jody Freeman have in effect taken up Rohr's theme by studying alternative administrative mechanisms which in some way promise to create representative bureaucracy. In various forms of ‘collaborative governance’, agencies develop rules and policies through ‘negotiated rule making’ among multiple stakeholders.Footnote 38 The list of stakeholders may include representatives of trade and industry, public-interest groups which represent the interests of a diffuse citizenry, state and local regulators and other affected parties. Although part of the justification for collaborative governance is straightforward interest representation through bargaining, another major justification is epistemic and points to the deep understanding of the particulars of the regulatory problems – the local knowledge – that stakeholders possess. ‘The collaborative claim that problem solving [through collaborative governance] tends to produce higher-quality rules rests upon the belief that unanticipated or novel solutions are likely to emerge from face-to-face deliberative engagement among knowledgeable parties … [Accordingly, i]n addition to the federal agency, likely members of a negotiated rule-making committee include representatives from the regulated industry, trade associations, labor organizations, public interest groups, and state and local governments.’Footnote 39
Beyond negotiated rule making, Congress sometimes enacts mandates that create structural forms of representative bureaucracy, built right into agencies’ composition or procedure. The Advisory Commission on Childhood Vaccines within the Department of Health and Human Services (HHS), created by statute to advise the HHS secretary on vaccine-related problems, comprises ‘health experts, members of the general public (two of whom have children who have suffered vaccine-related injury or death), lawyers, and officials from relevant agencies’.Footnote 40 The Dodd-Frank Act that reformed financial regulation created an ‘Investor Advisory Committee, which is tasked with advising the [Financial Stability Oversight Council] on regulatory reforms to protect investors. The Committee is comprised of a mix of representatives of various stakeholder interests, such as state governments, senior citizens, and pension funds, in addition to relevant experts.’Footnote 41 Likewise, ‘the financial reform act includes a provision to establish a Municipal Securities Rulemaking Board, comprised of experts and representatives of brokers, investors, and the general public, to set standards for municipal securities advisors.’Footnote 42 In all these examples, technocratic experts, whose knowledge Hayek impeached as excessively general and abstract, are accompanied and complemented by actors with thicker industry-specific practical and local knowledge.
Crisis and delegation. The recurrent major crises of the era since 1914 underscore that the administrative state may itself be capable of aggregating and organizing local knowledge even more rapidly than the market. For these purposes, crisis may be defined as a condition that increases the optimal rate of policy adjustment. In a crisis, policy must be modified and updated more rapidly than in normal times. The increase in the optimal rate of policy adjustment militates in favour of delegation to hierarchical bureaucracy and in favour of enhanced executive power; counter-intuitively, markets adjust too slowly to changing circumstances, and legislatures are even worse in this regard. I will support these claims with a point about the rate of adjustment in markets, a point about legislatures and the speed of policy change and a point about delegation to bureaucracies as a response to rapidly changing circumstances.Footnote 43
Hayek claimed that the price system adjusts more rapidly than a central planner and thus enables an endless dance of mutual adjustment by economic actors; he suggests, in other words, that decentralized coordination works more quickly under rapidly changing conditions than does centralized coordination. This is in principle a testable hypothesis, and it is hardly obvious that Hayek is correct. Decentralized coordination plausibly requires more time to reach an equilibrium than does centralized coordination, which is why greater urgency implies greater benefits from centralization.Footnote 44 In the Second World War, across political regimes of different types, ‘[e]verywhere the price mechanism came to be regarded as a method of allocating resources which was too slow and too risky.’Footnote 45 The solution, across regimes, was bureaucratic coordination of the generation and distribution of resources.
What about legislatures? Publius addresses the speed of adjustment to changing circumstances when discussing the optimal length of the term for federal representatives in the House. In constitutional design, the speed with which legislatures can acquire updated information about changes in local circumstances is determined, and constrained, by the length of the legislative term and the frequency of elections. As to these issues, Publius affirms the value of relatively frequent elections as a means of supplying representatives with updated local information, yet he also praises synoptic knowledge of the operation of the whole federal law-making system. The former implies a shorter legislative term and more frequent elections; the latter, a longer legislative term and less frequent elections. The resulting optimization problem has no pinpoint solution, but the trade-off curve plausibly has an internal maximum; Publius argues that the Constitution's two-year term for the House strikes a sensible balance.Footnote 46
This is fine as far as it goes, but it does not get at the main question, which is whether legislatures are capable of keeping up with the necessary pace of policy change in crises. Publius, through Hamilton, famously argued that speed of policy adjustment requires an energetic executive.Footnote 47 But even this argument was offered in an environment in which legislative agendas were far less constrained than they are today and in which the capacities of the administrative bureaucracy – as opposed to the presidency per se – were childish compared to those of the modern administrative state. The severely impacted agenda of the modern legislature implies that delegation to a hierarchically organized bureaucracy is the main way in which government will be able to react to crises. A major justification for administrative policy making is that agencies respond more quickly to changing conditions than do legislatures,Footnote 48 and part of what allows them to do so is the power to issue ad hoc, situation-specific orders rather than the general rules that Hayek preferred.
Hayek feared the abuse of power that ad hoc administrative orders make possible. A leading model,Footnote 49 however, illustrates a competing consideration which is usually decisive in practice: delegation to the executive should increase during crises, despite the increased risk of abuse, when and because the benefits are greater still. The more serious the crisis, the greater is the anticipated benefit if the executive uses its enhanced discretion to adopt utility-improving policies. Some level of executive abuse is the necessary by-product of other goods; hence, the risk of abuse should be optimized, not strictly minimized.
To be sure, the main Hayekian comparison is between ‘government’ or ‘the planner’ and markets, not between legislatures and the presidency or bureaucracy.Footnote 50 But because the administrative state has developed a mechanism – delegation to the executive and to agencies – that speeds up the pace of governmental response, the government-to-markets comparison is more favourable to government on the dimension of rapidity of adjustment than it would otherwise be. In so far as Hayekians emphasize the speed of adjustment to changing conditions, delegation to bureaucracies is itself part of the solution. And judging by the experience of the Second World War, the more extreme the crisis, the more bureaucratic adjustment will emerge as the only feasible option.
Let me sum up the claims so far. It is a cartoonish oversimplification to claim that the administrative state relies upon abstract technocratic knowledge, whereas the market relies upon local practical or tacit knowledge (metis). Critics who say things of this sort are unfamiliar with the actual institutions of the administrative state, which are variegated and complex and which contain far more in the way of novel institutional forms than is dreamt of in the critics’ philosophy. The administrative state itself builds in mechanisms which generate, aggregate and exploit local knowledge and which adjust rapidly to crises and changing circumstances. Even if there are no trade-offs between the generation and exploitation of local and global knowledge so that local knowledge is a good which should be strictly maximized, the choice between markets and administration on epistemic grounds is far more difficult than the Hayekian argument implies.
The organization of the bureaucracy
I turn now from the boundaries of the governmental firm – the scope of the administrative state vis-à-vis the market – to the internal organization of the bureaucracy given an administrative state of a certain scope. Suppose that the legislature delegated some domain of jurisdiction to the administrative state. Conditional on the assignment of that set of regulatory tasks, how should the bureaucracy be organized? I will suggest that the trade-off between local and global knowledge is central here as well. As we will see, however, although the questions are partly inspired by the Hayekian concern with dispersed local knowledge, the answers need not be Hayekian at all, in so far as the relevant variables and trade-offs counsel in favour of a more centralized administrative state than Hayekians might find desirable.
One might, of course, also come at the problem the other way around, by asking what tasks ought to be assigned to the bureaucracy, conditional on a certain internal organization. I will pursue the first formulation because it poses questions that have explicitly been analyzed in epistemic and Hayekian terms in the literature on industrial organization, which may usefully be arbitraged into the bureaucratic setting, mutatis mutandis. Although most of my examples and concrete details will come from the American administrative state, I believe that the trade-offs are universal and that similar questions arise in the organization of any advanced industrial democracy.
Synoptic or contextual regulation
A major fault line in the theory of the American administrative state involves the question of whether regulation should be synoptic or contextual. These are my terms rather than banners carried by the participants in the debates themselves, yet I hope that they capture the core of the disagreements in a way that hooks up with neo-Hayekian analyses of the administrative state. An example is James Scott's Seeing Like a State, which contrasts the synoptic regulation characteristic of high-modernist administration with local knowledge and metis, or tacit practical knowledge which central experts find it costly to acquire. Quite obviously, the contrast between synoptic and contextual regulation is a highly stylized one, whereas in reality there is a continuum between the two extremes, and everything is a matter of degree.
Synoptic regulation: Justice Breyer and OIRA. In the American debates, a leading proponent of synoptic regulation is Justice Stephen Breyer of the United States Supreme Court. In academic writings, Breyer has championed an approach to risk regulation that in effect requires regulators who possess global knowledge.Footnote 51 On this view, regulation should achieve an overview of all socially or economically relevant risks, should attempt to arrange them in order of priority and should regulate them just to the point at which the net social costs of regulation are equal to the benefits, but no more. The antithesis of synoptic regulation is uncoordinated, socially wasteful regulation by a myriad of myopic agencies and bodies. A decentralized regulatory apparatus of this sort will suffer from three major problems: ‘tunnel vision’, a kind of obsessive myopia in which agencies attempt to eliminate the last 10 per cent of the particular risk within their jurisdiction, even if the costs of doing so far exceed the benefits; ‘random agenda selection’, in which uncoordinated agencies devote resources to regulating risks on grounds other than a ranking of expected social benefits; and ‘inconsistency’, in which uncoordinated agencies regulate similar risks differently or different risks similarly.Footnote 52 Breyer's institutional prescription is a centralized overseer – a coordinating body perched atop the American administrative state, composed of politically insulated technocrats and charged with attempting a global assessment of the relative priorities of various risks.
The closest real analogue to Breyer's council of technocrats is a powerful body called the Office of Information and Regulatory Affairs (OIRA).Footnote 53 On paper, OIRA is merely a body subordinate to the Office of Management and Budget (OMB), itself a division of the Executive Office of the President. Yet, a bipartisan series of presidential orders – starting with President Reagan and continuing through President Obama – has given OIRA broad powers to coordinate and oversee regulation by the line agencies, such as the Environmental Protection Agency.Footnote 54 OIRA requires agencies to submit annual regulatory plans, with a view to coordinating the agendas of different agencies, but its most salient and controversial power is that it reviews major rules proposed by agencies to ensure that the quantified social benefits – defined in economic willingness-to-pay terms – exceed the quantified social costs. Although, in theory, OIRA also may issue informal requests or more formal ‘prompt letters’ that nudge agencies to promulgate regulations where regulation would produce net social benefits, in practice, the main effect of OIRA review is probably de-regulatory or anti-regulatory, blocking proposals by line agencies whose quantified net benefits cannot be demonstrated to be positive. I will return to this point later in this section.
Viewed in its best light, OIRA is a centralized mechanism for aggregating and coordinating information that is dispersed around the bureaucracy. Through formal review and informal consultation, OIRA draws together packets of information scattered through the line agencies and attempts to piece them together into a comprehensive picture of inter-related risks and their relative priority. There is a separate question – to be taken up shortly – about whether epistemic aggregation and coordination of this sort might be accomplished through decentralized mechanisms; for now, the point is just that epistemic coordination is the major good that OIRA supplies.
Contextual regulation: OIRA's critics. If Justice Breyer is the most visible proponent of synoptic, highly centralized and coordinated regulation, there is an array of critics who call for more contextual, decentralized forms of regulation, entrusted principally to the line agencies which specialize in particular regulatory problems. Although the resulting debates are multi-faceted, a prominent strand is epistemic. The critics suggest that OIRA intervention often makes things worse, rather than better, because OIRA lacks contextualized local knowledge about particular regulatory domains and problems. In one account, evaluations of OIRA's work by the personnel of line agencies
are almost uniformly negative … In one midlevel [agency] employee's opinion, ‘they [OMB personnel, i.e., OIRA] don't know what they are talking about, and they don't care.’ One EPA analyst complained that OMB analysts spend too little time with any single regulation to become sufficiently educated to contribute much to the agency's analytical efforts. Agency officials frequently observe that OMB analysts lack sufficient expertise to understand highly technical questions that often arise in agency rulemaking. [One agency's] scientists, for example, note that OMB analysts often ‘venture their opinions on items of industrial hygiene and epidemiology when they are not qualified to be giving opinions.’ Addressing OMB's attempt to affect agency carcinogen policies, Congressman John Dingell complained about OMB's ‘extensive effort to second-guess the scientific and technical judgments of federal agencies in the highly complex area of cancer risk assessment.’Footnote 55
The main thrust of these critiques is that OIRA is under-specialized. The downside of its purportedly comprehensive vision and comprehensive agenda planning, across different agencies and regulatory risks, is that OIRA analysts may not understand particular risks in context. To some degree, OIRA might attempt to solve its under-specialization problem in either of two ways: (1) soliciting comments from line agencies during the OIRA review process or (2) expansion which internalizes subject-specific expertise, perhaps by hiring experts in relevant domains. The first is a routine feature of OIRA's decision making, and as for the second, an OIRA administrator in the George W. Bush administration, John Graham, hired experts in a range of disciplines.Footnote 56
Even in principle, however, neither approach can provide a complete solution because part of line agencies’ context-specific knowledge is practical, tacit and non-transmissible. The problem is Hayekian: part of what agencies know they learn by doing in the course of implementing congressional or presidential commands on the front line.Footnote 57 The tacit practical knowledge of line agencies is a form of metis, knowing how rather than knowing that, and will by the nature of the case be inaccessible to experts in OIRA, no matter how technically specialized, who have not themselves worked through the myriad complexities of implementing general statutory commands and policies. No amount of meetings between line agencies and OIRA personnel or hiring of new OIRA personnel will convey this tacit local knowledge to the centre. Neither communication with OIRA nor internalization of expertise within OIRA can fully substitute for the local knowledge of line agencies.
It follows that there is a genuine trade-off between the benefits of synoptically rational priority setting and coordination, on the one hand, and local knowledge, on the other. A structurally analogous trade-off is recognized in the literature on team theory and the industrial organization of firms.Footnote 58 The trade-off can be analyzed either in incentive terms or in epistemic terms or both; in its epistemic aspect, the trade-off is sometimes traced explicitly to Hayek and in effect restates the trade-off between local and global knowledge. ‘Decentralized organizations have a natural advantage in adapting decisions to local conditions, since the decisions are made by managers with the best information about those conditions. However, such organizations also have a natural disadvantage since the manager in charge of one division is uncertain about the decisions made by others.’Footnote 59 The debate over synoptic versus contextual regulation, over global knowledge of the risk-regulation agenda and global priority setting versus localized problem-specific expertise and practical knowledge, sounds in exactly these terms. Despite the pretensions of both camps, neither can claim the mantle of full rationality or true expertise; rather,
bounded rationality affects both decentralized and centralized decision making. In a centralized setting, bounded rationality manifests itself in a ‘one size fits all’ policy. In a decentralized setting, bounded rationality manifests itself as a lack of awareness of synergies [across subdivisions].Footnote 60
Mutatis mutandis, the same holds for the organization of the risk-regulation bureaucracy. Although synoptic regulators such as Breyer emphasize the failures of line agencies to set globally rational priorities in risk regulation, contextual critics argue that OIRA's decision procedures, particularly quantified cost-benefit analysis based on willingness to pay, amount to a one-size-fits-all approach that sometimes ignores the full complexity of local regulatory problems.
Coordination: vertical or horizontal?
Given this large-scale trade-off between local and global knowledge in risk regulation, can we make any progress on the design of the regulatory bureaucracy? As the team theory and firm organization literature also emphasizes, it is a mistake to state the trade-off as pitting the benefits of coordination against the benefits of local adaptation or tailoring. Coordination may be accomplished through centralized oversight or else through decentralized mechanisms. Just as Hayek argued that the price system might bring about decentralized coordination of economic activity, so too it is possible that lateral communication between or among line agencies might bring about decentralized coordination of risk-regulation agendas and priorities. Indeed, assuming that agencies have common fundamental preferences, the greater the benefits of inter-agency coordination, the greater is the incentive for line agencies to coordinate horizontally.Footnote 61
So the real trade-off is between centralized coordination through an OIRA-style oversight body and decentralized coordination through various mechanisms of inter-agency collaboration and communication. The devices and mechanisms available for this purpose are myriad, ranging from simple talk between agency officials to the relatively formal ‘memorandum of understanding’ which sets forth the agencies’ priorities, jurisdiction or cooperative agenda. Of course, some mix of centralized and decentralized coordination will almost certainly be optimal in any given regulatory environment; it is unlikely that either of the extremes, total centralization or total decentralization, will prove desirable. But the desirable set of arrangements may well be skewed in one direction or another in any given environment.
To get traction on these questions, we need to identify the conditions under which a relatively more or less centralized regulatory bureaucracy will prove to be optimal. (Or, if ‘optimal’ is too ambitious, we may think in terms of marginal or incremental movements from the status quo; given some extant organization of the regulatory bureaucracy, would we want to nudge it towards greater centralization or the opposite?) The industrial organization literature teems with models, some of which focus on trade-offs between local and global knowledge, some on incentive compatibility and some that attempt to combine the two questions by endogenizing the incentives for actors in the bureaucracy to acquire information.Footnote 62 Limiting myself as far as possible to the epistemic side of these questions, I will attempt to distil and extract some common themes from the models and offer a list of variables that push in one direction or the other. When all is said and done, I believe, the American administrative state is distinctly unpromising terrain for decentralized coordination by line agencies.
Duration in office and turnover of personnel. As emphasized by a pioneering study of vertical and horizontal coordination in American and Japanese business firms,Footnote 63 horizontal coordination flourishes in the presence of long-standing informal relationships among actors in different divisions or offices. Relationships of this sort reduce the costs of communication, allow repeated interactions which support long-term commitments and engender trust and deepen horizontal knowledge across units. In a stylized contrast, Japanese firms typically display low inter-firm mobility but high intra-firm mobility on the part of workers, who may stay at a single firm for a whole career but move frequently among the firm's internal units; American workers show the opposite pattern. The Japanese pattern encourages informal horizontal relationships and thus supports horizontal coordination across divisions and units, while the American pattern hampers informal relationships and thus supports vertical coordination by firm management.
Even a passing familiarity with the American administrative state suggests that low duration in office and high turnover make the horizontal coordination model inapposite. Although the civil service is staffed by career employees, most line agencies have several layers of political appointees at the top, who serve for a few years and then return to the private sector or go on to an unrelated government post.Footnote 64 The American administrative state is under-professionalized in comparative perspective, and given the role of political appointment to leadership posts, the chiefs of line agencies typically have little chance to form enduring relationships of repeated cooperation and trust. All this raises the costs of horizontal or decentralized communication and thus raises the costs of horizontal transmission of local knowledgeFootnote 65; thus, top-down coordination of regulatory priorities through an overseer such as OIRA becomes relatively more attractive.
Spillovers and communication costs. Suppose that the risks agencies regulate are highly inter-related. To the extent that this is so, the action of any particular agency may have large spill-over effects on the work of other agencies, yet the first agency may not realize this because its knowledge is local. In an epistemic framework, the spill-over issue is not or not mainly that one agency fails to internalize the costs that it is inflicting on other agencies; the issue is that the agency producing spill-overs is myopic and is unaware of the effects of its behaviour.Footnote 66
Just as ‘externalities’ arise from the transaction costs of bargaining,Footnote 67 so too spill-overs arise from the costs of communication. The smaller the costs of horizontal communication between or among agencies, the easier it is for agencies to inform one another of these spill-overs and coordinate horizontally. This implies that the greater the number of agencies involved in regulating any particular risk, the higher are the costs of communication and the greater is the attraction of a centralized coordinator, who can act in part as a clearinghouse for information and in part can enforce bureaucracy-wide standards that reduce the informational load on line agencies.
Here too, structural characteristics of the American administrative state make a centralized coordinator attractive. For reasons good and bad, Congress has parcelled out regulatory jurisdiction to a staggering variety of agencies, many of which have overlapping jurisdiction to regulate the very same risk or else risks that are causally and scientifically inter-related.Footnote 68 In the response to the Deepwater Horizon oil spill in the Gulf of Mexico, over a dozen U.S. federal agencies played a role, ‘making decision making slow, conflicted and confused’ and producing an overall ‘lack of coordination’.Footnote 69 The relevant agencies included the Environmental Protection Agency, the Coast Guard, the National Oceanic and Atmospheric Administration, the Minerals Management Service, and the list goes on; the jurisdiction of these agencies is defined in crosscutting terms. This is not the sort of issue for which OIRA review is relevant, but it illustrates in extreme form the chronic problem that communication costs increase exponentially as the network of related agencies and inter-related risks becomes ever more dense.
Asymmetrical bureaucracies. Another determinant of the relative costs of vertical and horizontal coordination is the presence or absence of symmetry across bureaucratic units. Symmetry can be defined either in terms of size, in terms of organizational form and procedures or in terms of the timing of decisions; asymmetrical organizations are ones in which units are of unequal size, in which units are structured differently and make decisions through very different procedures or in which one unit moves first and the other follows rather than the two units engaging in simultaneous decision making.
In a leading model of the trade-offs between centralized and decentralized coordination, asymmetries tend to push in favour of centralization.Footnote 70 Although that model gives a mix of epistemic and non-epistemic reasons, there is a straightforward intuition that asymmetries will generally make informational coordination more difficult. Other things being equal, larger organizations and first-mover organizations will generally prove more myopic in the local-knowledge sense both because their own problems and organizational imperatives will loom larger and because the first mover will necessarily have less information than will the second mover about which policies have been adopted and what effects those policies produce. Generally speaking, the more unlike coordinating units are, the higher are the costs of transmitting information among them and the less the leadership of one will understand the problems of the other.
If this intuition holds, then the American administrative state is singularly unpromising terrain for horizontal coordination. The American bureaucracy is characterized by exorbitant variability and heterogeneity of size, scale, function and institutional organization. Its bewildering alphabet-soup of agencies and its menagerie of agency forms imply that very often unlike will have to coordinate with unlike, resulting in miscommunication and coordination failures. An obvious answer is bureaucratic consolidation and reorganization, which is occasionally tried, with mixed results. But, conditional on the current welter of agencies persisting, the high costs of horizontal communication among unlike entities implies the need for some centralized coordinator.
Overall, I believe that extant models of the trade-offs between centralized and decentralized communication imply a robust role for centralized oversight of the American administrative state. None of these considerations enable to us to make pinpoint prescriptions about how centralized that oversight ought to be or what form it ought to take; abstract considerations of the sort I have adduced cannot tell us whether a body such as OIRA is a good idea or what the precise bounds of its authority ought to be. But they can tell us, at a minimum, that conditional on assuming no large-scale simplification of the American bureaucracy and no large-scale retrenchment of its delegated regulatory jurisdiction, fully decentralized coordination is a deeply unpromising strategy. As Breyer and the other proponents of synoptic regulation argue, the problem of locally myopic agencies does imply a need for centralized oversight – although the analytical path needed to reach this conclusion is different from the one which Breyer and other synoptic regulators usually follow. Whatever its failings, OIRA aggregates and deploys information which is dispersed around the bureaucracy; plausibly, decentralized or horizontal coordination among agencies cannot accomplish the same ends.
A central planner to protect free markets?
I conclude by attempting to bring the two parts of the discussion together. The distinction between the scope of the governmental firm, on the one hand, and its internal organization, on the other, has the effect of pitting two Hayekian premises against one another. In the modern administrative state, a centralized overseer of the bureaucracy is an indispensable protector of free markets, yet the same local-knowledge arguments that Hayekians deploy to support market freedom in the first place also may be deployed to undermine centralized oversight of the regulators.
Hayekians seek to protect the market from unnecessary regulatory intervention by agencies in the name of local knowledge, dispersed information and decentralized coordination by economic actors. The principal way to do this would be to limit the delegated regulatory jurisdiction of the agencies altogether, as discussed in the first part of this chapter. Yet, given the large-scale delegation of regulatory jurisdiction to line agencies – an entrenched feature of the American administrative state – some Hayekians argue for a centralized overseer of the bureaucracy, a guardian of the regulatory guardians. The overseer might apply cost-benefit analysis or instead some other decision procedure, but its main aim will be to take a synoptic perspective that cuts across regulatory domains, coordinates risk-regulation priorities and agendas with a view to social welfare and corrects for the myopia of line agencies which becomes obsessed with their regulatory missions and thereby over-regulate.
For concreteness, let me focus on an article by Susan Dudley, the administrator of OIRA in the second term of the George W. Bush administration.Footnote 71 Dudley's principal concern is that line agencies are charged with ‘single-issue missions’ that structure their ‘perspectives’ and cause them to suffer from Breyer's ‘tunnel vision’.Footnote 72 By contrast, Dudley says that the value of OIRA lies in its ‘cross-cutting perspective and its focus on understanding tradeoffs and consequences’.Footnote 73 For Dudley, the main risk of agency myopia is over-regulation of the market, and the promise of OIRA's synoptic perspective is to check over-regulation. She explicitly invokes Hayek in order to warn against ‘substituting the judgment of government regulators for the decentralized wisdom of crowds’.Footnote 74 Her peroration asks,
[H]ow can we overcome this fatal conceit [a Hayekian catchphrase]Footnote 75 and raise awareness of the Hayekian insight that decentralized market processes are better able than centralized government to focus dispersed information – information that no one individual (not even a regulator) can obtain – and convey it efficiently to market participants?Footnote 76
But this combination of views suffers from a latent tension. The Hayekian argument from local and tacit knowledge that Dudley uses to support OIRA's role as a check on over-regulation, and (indirectly) as a defender of free markets, is the very same argument that line agencies, and the critics who support them, urge against centralized oversight by OIRA. In this line of criticism, OIRA is itself the synoptic regulator which relies on abstract technocratic expertise to issue general guidelines and to set system-wide priorities. And, in the critics’ view, OIRA does poorly because it suffers from the very same informational deficits that afflict ‘the planner’ in Hayek's argument: no matter how well motivated, OIRA lacks the local, partly tacit and practical knowledge of particular regulatory problems which would be needed to make optimal decisions.
The hidden pre-condition for this tension to arise is that OIRA usually intervenes to block putatively myopic agencies from creating new regulations which hamper free markets. In principle, OIRA might also nudge agencies to intervene where welfare-enhancing regulation is called for. OIRA occasionally employs the device of ‘prompt letters’ which nudge line agencies to regulate where the apparent benefits of regulation exceed the apparent costs. Yet, there is broad consensus that in practice OIRA rarely does this;Footnote 77 agencies do not lie awake worrying that OIRA will force them to regulate where they do not want to. Dudley and other Hayekians want an OIRA which is vigilant to weed out ‘unnecessary’Footnote 78 regulation but rarely discuss the opposite case.
The logical implication of their views is a centralized synoptic overseer of the bureaucracy – a sort of Hayekian central planner of regulation. The patent tension in this ideal is not a contingent peculiarity of Dudley's argument which might be corrected. Rather, it is a structural problem for Hayekians, who must come to terms with the existence of a massive regulatory bureaucracy and who will have to decide whether the local-knowledge argument for protecting markets from unnecessary regulation can be squared with the global-knowledge argument for a centralized regulatory overseer.
For my part, I do not think the two halves of Dudley's view can be reconciled. The most promising approach would be a second-best argument: the Hayekian ideal is a genuinely free market, but conditional on the creation of a massive regulatory bureaucracy, a centralized overseer which in effect constrains unnecessary regulation is the attainable second best. The problem with this justification is that it is ideological, in the sense that it treats the same causal argument differently depending on the political valence of the context in which the argument arises. If one subscribes to the original Hayekian argument that decentralized tacit knowledge justifies a robust regime of free markets, then one ought also to agree with OIRA's critics that decentralized line agencies will enjoy a comparative informational advantage – relative to the centralized overseer – on the question of which regulations are actually necessary. Valid second-best arguments, like other arguments, must rest on consistent causal premises, and it is hardly obvious that the Hayekian justification for a centralized regulatory overseer satisfies this minimal-validity condition.
This is a point about the consistency of Hayekian justifications for the centralized overseer. It is not an objection to the overseer itself, which is in my view justifiable on other grounds. It is perfectly consistent to believe both the following claims: (1) the benefits of centralized epistemic coordination in the economy as a whole and the ability of representative bureaucracies to generate local knowledge together imply a broad scope for the administrative state, and (2) the benefits of centralized epistemic coordination within the bureaucracy and the costs of horizontal coordination among agencies together imply a robust role for an OIRA-style overseer. Hayekians who desire a centralized overseer to defend free markets subscribe to point (2) but not point (1) – even though the argument for point (2) also implies support for point (1). A practical difference between these two (combinations of) views is that the non-Hayekian approach implies that OIRA ‘prompt letters’ to spur regulation might well be justifiable depending upon the circumstances, while on the Hayekian approach, such letters would always be suspect, absent special circumstances such as monopoly.
Conclusion
The distinction between local and global knowledge is essential for understanding both the scope of the administrative state and its internal organization; the epistemic organization of the administrative state ought to be a central agenda item for prescriptive legal and political theory. Hayek's views are directly relevant to these questions, but his views have also turned out to be largely untenable. The market is necessarily an imperfect aggregator of information, including local knowledge; on the other side of the ledger, the administrative state, although itself highly imperfect, amounts to a complex machine for aggregating and exploiting local and practical knowledge, both through democratic representation and through delegated policy making by representative bureaucracies. Given large-scale delegation, the trade-off between local and global knowledge and the high costs of horizontal coordination among agencies implies a robust role for centralized oversight of the bureaucracy by an institution such as OIRA. Paradoxically enough, free-market Hayekians applaud such an institution precisely because it is a centralized overseer whose effect is to protect markets from regulatory intervention. Whether this applause is theoretically consistent is at best unclear; I believe it to rest on shifting, inconsistent and ideologically inflected causal premises.
