DEMOCRATICALLY DURABLE REGULATION

How should regulators assess the long-term consequences of their policy choices? With challenges including climate change, pandemic preparedness, and artificial intelligence all looming, regulating with a view toward the future seems more important than ever. Yet the prevailing approaches to this problem are incomplete.¹ Scholars have long debated how much to discount future interests when projecting policies’ cumulative costs and benefits (and related questions).² Some have further argued that cost–benefit analysis itself can misrepresent our moral relations to future generations.³ Though these debates have been illuminating, they have often neglected regulation’s distinctively political significance. When administrators promulgate climate regulations, for example, their choices both reflect and shape the distribution of political power. Commentators have not wholly missed these dynamics; all recognize that regulatory choices require political judgment. But while the Office of Information and Regulatory Affairs (OIRA) has institutionalized the economic analysis of proposed rules, no comparable infrastructure exists for their political analysis. Nor have scholars much envisaged it.⁴

There have been some signs of change. Immediately upon taking office, President Joseph R. Biden announced an initiative to “moderniz[e]” regulatory review.⁵ President Biden’s memorandum called for new approaches to “environmental stewardship,” to promoting “the interests of future generations,” and to “fully account[ing] for regulatory benefits that are difficult or impossible to quantify.”⁶ To carry out this agenda, the President appointed to OIRA progressive scholars attentive in their writings to the politics of public policy, including Sharon Block, K. Sabeel Rahman, and (as Administrator) Richard Revesz.⁷

These developments are promising, but they inevitably have left important matters unaddressed. For example, the memorandum sought reforms to “promote the efficiency, transparency, and inclusiveness of the interagency review process.”⁸ These three aims might all be understood in democratic terms. Yet they might also appear to sit in tension with one another: transparency might hinder efficiency; efficiency might tell against inclusiveness; and even inclusiveness might be served by diminished transparency. So understood, the “Modernizing Regulatory Review” memorandum suggests some expanded role for democracy in regulation while leaving substantial questions unresolved about what, exactly, that role should be.⁹

In this article, I take the Biden memorandum as impetus to sketch a new approach to high-stakes regulation of long-term problems, one with politics at its core. I focus on what I call “structural risks,” exemplified by the climate crisis: systemic problems that tend over time both to increase in magnitude and to become more difficult to redress. Structural risks place special stress upon conceptions of regulatory review that neglect politics. Structural risks invite—indeed, require—present regulators to make decisions that can profoundly affect future generations. Structural-risk regulation also tends to be exceptionally complex. And structural-risk regulation must also be politically stable over the long haul: it is dangerous to promulgate rules one year, only to have them reversed the next.

Considering, then, the central role of politics in regulating structural risks, I argue that there are good reasons, both principled and pragmatic, for structural-risk regulation to reflect and reinforce the core democratic values of citizens’ freedom and equality. For those who value democracy for its own sake, structural-risk regulations are profoundly important. They not only affect important, large-scale systems (including social and political systems) but ideally follow from contestable choices about what those systems should look like over time. Those who value democracy only instrumentally—say, as a means to effectuate good policy—also have especially strong reasons to care about democratic principles here. Failing to treat citizens as free and equal, I argue, is likely to undermine the trust and knowledge necessary to regulate structural risks effectively. Moreover, if regulators neglect the political consequences of their decisions, they are likely to miss important determinants of policy efficacy.

I identify, justify, and elaborate six principles for a “democratic approach” to modernizing regulatory review.¹⁰ These principles sweep widely, addressing both the process and the substance of regulation.¹¹ Though they leave room for structured experimentation,¹² the six principles support several specific reforms: first, setting systemic policy goals before developing structural-risk regulations; second, attending to new forms of public input, such as assemblies of randomly selected citizens; and third, assessing potential regulations for their long-term political durability by drawing upon social-scientific expertise beyond economic cost–benefit analysis.

The last of these proposed reforms is especially significant. Some legal scholars have certainly recognized policy feedbacks’ importance.¹³ But to my knowledge, none has yet analyzed in depth how to integrate consideration of those feedbacks into regulatory review writ large. I offer such an account. My discussion of policy feedbacks is grounded in the ideal of “democratic durability” and integrated into a broader vision of the regulatory process. Yet even those who reject this larger vision should want regulatory reviewers to consider policy feedbacks. Politics constrains public policy in ways not easily described as costs or benefits. Too-narrow devotion to the tools of economics risks missing both threats and opportunities for regulation.

***

In the analysis that follows, I take democracy, roughly speaking, to require political institutions that can be justified in terms of the freedom and equality of all citizens. Respecting citizens as equals means respecting their ability to make decisions about how political life should go—that is, their capacity for political judgment.¹⁴ This ideal is admittedly ambiguous, not least in its institutional significance. But one natural implication is that government should be equally accountable to all its citizens.¹⁵ Whether through prospective consultation, retrospective elections, or reason-giving, making government answerable to citizens’ claims seems essential to manifesting respect for their judgments.

“Democracy” is, of course, a contested concept, and some will dispute my definition.¹⁶ Nonetheless, I believe it represents shared ground among many democratic theorists. For example, some ground democracy’s value in individual autonomy.¹⁷ Whatever else realizing autonomy might require, it plausibly demands institutions justifiable (in some sense) in terms of respect for citizens’ capacity for judgment. Others ground democracy in an ideal of citizens relating as social equals, without invidious hierarchies.¹⁸ This value, too, requires institutional respect of much the same sort. Seeking to remain on common terrain, I take the basic form of the existing administrative state as given and then invoke my stipulated democratic principles to develop recommendations for reform.¹⁹

As for my focus on structural risks, that serves two main purposes. First, it offers analytical clarity, which facilitates specification of when and how democracy matters. Some saw President Biden’s memorandum as an opportunity “to fully account for the interests of future generations and … existential risks.”²⁰ “Structural risk” is a single category that can capture much (though not all) of what matters about these pressing challenges.

Second, “structural risk” offers a theoretical test case for my proposed reforms. I show that there are particularly strong reasons to embrace a “democratic approach” to regulating structural risks. Defining systems-level goals before regulating, for example, is advisable in addressing all systemic risks—but it is especially important when those risks exhibit substantial path dependency. Reinforcing democratic values in the process of choosing goals and developing regulations is worthwhile for all “significant regulatory action[s]”²¹—but particularly for those addressing risks involving profound problems of social coordination. And regulations generally should be politically durable and therefore subject to review for their potential policy feedbacks—but such review is most significant for rules likely to shape significantly the political future.

My approach might also have appeal beyond this context. If my ideas seem plausible in the context of structural risk, that is more reason to investigate their potential elsewhere. And if they do not, the countervailing considerations are likely to reveal more clearly the limits of “democratizing” administrative reforms.

The remainder of this article consists of three parts. Part I further defines and discusses structural risks. Part II lays out procedural and substantive principles for structural-risk regulation. Part III responds to some possible objections to my proposals.²²

I define structural risks as having three characteristic features:

• 1) 

  They threaten greater harm over time; i.e., they exhibit increasing returns with respect to harm.

• 2) 

  They become costlier to redress over time; i.e., they exhibit increasing returns with respect to cost of redress.

• 3) 

  They may cause significant and systemic harm.

Two concepts here are especially important: “increasing returns” and “systemic harm.” Increasing-returns processes exist when “the probability of further steps along the same path increases with each move down that path … because the relative benefits of the current activity compared with other possible options increase over time.”²³ Systemic harms are those affecting certain “regularly interacting or interdependent group[s] of items forming … unified whole[s].”²⁴ Systemic harms are never wholly localized, but instead affect interlocked processes. Because “systems” may exist at any scale, I limit the definition of “structural risk” to those threatening “significant” harm. Significance here cannot be fully defined in advance: the point is simply to pick out risks important enough that the marginal costs of implementing the reforms I describe below could likely be justified.

Together, the three features identify risks posing especially difficult problems of intertemporal political choice. Consider the climate crisis. Climate change involves greater potential damage as atmospheric greenhouse gases accumulate; its harms become more difficult to redress as new positive-feedback processes begin; and it threatens substantial harm to global economic, social, and ecological systems. By contrast, persistent pollutants, such as polyfluoroalkyl substances (PFAS), may become increasingly difficult to remediate over time, but they generally do not cause escalating harm. And though asteroid collisions threaten systemic harm, they probably do not become costlier to avoid. Decisions affecting the climate today therefore may have huge ramifications for the future, in ways decisions about PFAS or asteroids might not. These ramifications pose a normative problem, insofar as today’s policy makers ought not to leave future persons in bad positions that are difficult to redress. They also pose a motivational problem, insofar as present decision-makers perceive little interest in acting upon that moral obligation (and much pressure to act contrary to it).²⁵

Structural risks pose other distinctive challenges, as well. For one thing, because structural risks threaten systemic harm, they often exhibit nonlinearities. Systems are characteristically interdependent, so what happens in one part of a system does not usually stay there. These interactions can cascade suddenly and unpredictably, sometimes to catastrophic effect. Moreover, because the costs of redress tend to increase over time, it is preferable (all else equal) to mitigate structural risks sooner rather than later. For both these reasons, structural risks should generally be especially high priorities for regulation.

In some cases, it may be unclear whether a given risk is “structural.” For example, much depends upon how “harm” is calculated. Should policy makers choose not to discount future costs and benefits at all, then PFAS discharges might well cause more harm over time, if only because more people could be exposed to the pollutant as the years pass. Similar considerations apply to determining the cost of redress. These ambiguities suggest that almost any risk could, in principle, be characterized as “structural.” In practice, such determinations will be political, in terms of both how and why they are made.²⁶ In general, institutional context is likely to affect the production, consumption, and interpretation of information about risks.²⁷ The same holds for categorizing risks: what might seem path-dependent and systemic from one perspective may seem fluid and localized from another.

For my purposes, however, such fuzziness is not fatal. For many important risks, designating them as “structural” will be uncontroversial. However else one might characterize the risks associated with climate change or pandemics, they are clearly structural in my sense. For harder cases, scholars, politicians, and regulators can reason backward from structural risks’ policy implications. That is, they can ask whether a given problem creates particular problems of intertemporal choice, nonlinearity, and increasing cost of redress. If so, then it may be appropriate to treat that risk as structural, even if it could also be analyzed in other terms.

This subpart outlines three normative principles for the process of developing structural-risk regulations. The core theoretical idea is that regulatory procedures should reinforce democracy, for reasons both principled and pragmatic. I offer some concrete suggestions for pursuing this goal, building upon recent work in democratic theory. These recommendations are not intended to be definitive: other procedural schemes may be compatible with the broad structure I propose, and experimenting with these may prove fruitful.

Principle #1: Choices about systems-level goals should generally precede structural-risk regulations.

Regulating systemic risks involves particular epistemic challenges. The characteristics of any particular component of a system depend substantially upon the characteristics of the system as a whole.²⁸ It is therefore generally difficult to aggregate component-level choices to obtain particular systemic goals. For example, whether or not electrifying the automotive sector contributes to overall environmental sustainability depends upon the characteristics of electricity generation, lithium and cobalt mining, and much else. This does not mean marginal progress is impossible without perfect knowledge. But that progress may be inadequate for achieving broader objectives, even in aggregate. And because structural risks, in particular, threaten greater harms over time, they can overwhelm any gains made.

When considering structural risks, policy makers should ideally determine systems-level goals before identifying possible responses.²⁹ Doing so would allow them to choose more effectively among regulatory options. Concretely, this might mean asking the following questions, among others: What characterizes an environmentally sustainable economy? Or a society resilient against pandemics? Of course, these will generate further questions. Consider, for example, the internationally adopted Sustainable Development Goals, which include not only seventeen high-level goals but also a “shared blueprint” with 169 targets.³⁰ In some cases, this process of goal setting and elaboration may seem easy. Net-zero greenhouse-gas emissions, for example, is surely one necessary aspect of environmental sustainability. But some dispute the value of sustainability in the first place, the net-zero ambition, or the appropriate timeline for achieving it. Even where there is broad agreement, some choices may remain difficult, especially those concerning novel risks.

Though goal setting ought generally to precede policy making, this should not foreclose either swift regulatory action or iterative assessment of objectives. In some cases, it may be appropriate to impose even stringent regulations before setting all relevant objectives. For example, some have endorsed an “Anti-Catastrophe Principle,” according to which policy makers should adopt a “maximin” strategy in the face of (1) a potential catastrophe (2) the likelihood of which cannot be assigned a probability, but which (3) rises above some minimal threshold.³¹ In these cases, the regulatory task may consist more of preserving the status quo than of pursuing systemic change, making goal setting inapposite. In other cases, policy makers may wish to update goals set by their predecessors in light of changing conditions or new information. Setting systems-level goals may therefore be an iterative process, incorporating both new normative judgments from democratic processes and new insights from regulatory science.³²

Some generalist administrative body is necessary to lead the goal-setting process; OIRA is the most likely candidate today. Setting systems-level goals requires integrating information across existing administrative boundaries. For example, defining “pandemic preparedness” may involve considering, among many other things, global migration patterns, food and water supply, and state capacity. Goal setting also involves examining the interactions among systems: for some purposes, it may be appropriate to treat the climatic system and the public-health system as distinct; for others, less so. These meta-level questions demand resolution by an omnicompetent body capable of viewing systems as wholes. In the 1960s and 70s, scholars called for something similar: administrative bodies capable of taking “a systems or holistic approach” to public policy.³³ But no such institution emerged with more than “relatively limited” powers.³⁴ Despite its very different intellectual orientation, OIRA might now fill the gap. OIRA, of course “has come under sustained criticism almost since the inception of regulatory review.”³⁵ But it nonetheless has the distinct advantage here of an existing trans-substantive mandate. Moreover, goal setting should be responsive to the regulatory review process I describe below. Therefore, any substitute for OIRA would substantially have to replicate its functions while somehow avoiding its flaws. Reform seems the easier path.

Setting goals for structural-risk regulation would only extend existing practice. Since 1993, federal agency representatives have been required to coordinate their proposed rulemaking agendas for the coming year with the Executive Office of the President.³⁶ And twice each year, OIRA publishes the Unified Agenda, which notes agencies’ planned regulatory and deregulatory activity.³⁷ High-level, pre-rulemaking coordination structures therefore already exist. These structures could be repurposed to facilitate the creation of systems-level goals. Doing so would represent an extension of Executive Order 12,866’s stated ambition of providing for “coordination of regulations,” alongside merely “promot[ing] the President’s priorities.”³⁸

Principle #2: The process of choosing system-level goals should reinforce democratic values.

There are strong reasons, magnified by the distinctive features of structural risks, for this goal-setting process to reinforce democracy. Here again, E.O. 12,866 provides a model, with its call for “maximiz[ing] consultation and the resolution of potential conflicts at an early stage” and “involv[ing] the public and its State, local, and tribal officials in regulatory planning.”³⁹

Democracies, as I claimed above, should respect citizens’ freedom and equality. But because they concern citizens as such, they should do so systemically, not necessarily in each of their constituent institutions. What matters is whether the polity as a whole is democratic, not, say, whether the Environmental Protection Agency taken alone is democratic.⁴⁰ This means that “reinforcing” democracy through goal setting need not mean directly “democratizing” goal setting. Like Samuel Bagg and Udit Bhatia,⁴¹ I do not believe democracy must be “isomorphic,” replicating the same institutional principles in every case. In this sense, goal setting need not become a “forum for democratic agency,”⁴² even if it promotes democratic agency overall.

The principled reason for reinforcing democracy through systemic goal setting is that expertise alone is generally insufficient to justify choices among system-level goals. Regulatory decisions may sometimes be defended on technocratic grounds—for example, that they efficiently achieve certain ends. But in making systemic choices, ends are not given. These decisions require determining, say, what makes an economy “sustainable” and why sustainability is valuable in the first place. Here there is likely to be reasonable, nontechnical disagreement. Some may perceive environmental justice as partly constitutive of sustainability; others may see them as distinct. Of course, “[a]gencies are themselves the products of a democratic process,” not an alien imposition.⁴³ Regulators may therefore have sufficient normative democratic authority to make these choices. But if regulators are to answer value questions, then (all else equal) they should do so in ways that deepen democracy. That is, regulators should embrace procedures reflecting their respect for citizens’ capacity for judgment and sustaining that respect into the future.⁴⁴

Democracy’s importance here is increased by structural risks’ long-term significance. Because today’s regulatory decisions may decisively affect tomorrow’s prospects, policies made now may therefore appear as a “dead hand” reaching from the past to govern the future. To some extent, this is inevitable: every political choice affects the future, and none is strictly costless to reverse. But with respect to “major” decisions—and especially those whose significance tends to grow over time—democratic legitimacy is particularly important, in two respects.⁴⁵ First, to the extent that perceived political identity is continuous, democratic decision-making today may help future citizens see our choices as “their own.”⁴⁶ In this sense, today’s democracy may sustain desirable forms of political agency. Second, technocratic decision-making may foreclose future democratic decision-making. Institutional design itself frequently exhibits increasing returns.⁴⁷ It may therefore prove difficult to recover democratic agency once lost, and citizens may find themselves unable to change regulatory course.

These intertemporal considerations also suggest the pragmatic reason for endorsing democratic procedures: systemic stability requires both social trust and trust in government.⁴⁸ In general, complex systems are characterized by substantial uncertainty.⁴⁹ In social systems in particular, managing this uncertainty frequently requires trust: trust facilitates forbearance from conduct that may be individually rational but collectively destructive. For example, regulations concerning artificial-intelligence research, or restricting research of solar geoengineering, may be undermined by “rogue” researchers skeptical of the government’s decisions. Trust is especially important when policy makers are seeking to effect systemic change, since uncertainty is likely to be especially great during the transitional period. And after transitional periods, when decisions may be partly “locked in,” social trust facilitates citizens’ reconciliation with the status quo. Successful political institutions generally foster both social and political trust.⁵⁰ In particular, citizens who value democracy place more trust in governments they perceive as highly democratic.⁵¹ And political trust, in turn, contributes to the effective implementation of public policy.⁵²

Trust and trustworthiness are especially important in view of the epistemic challenges involved in regulation. Not only is technical expertise alone unable to answer important value questions, it is crucially limited on its own terms, as political scientist Samuel DeCanio has argued. In general, the administrative state is hindered by the “singularity” of its decisions: there is “no rival organization with the authority to produce legally binding rules that are compulsory for all members of its geographically defined territory.”⁵³ This lack of rivals means that there are “no observable policy counterfactuals within the territory the state controls.”⁵⁴ It is therefore difficult to infer the causal effects of administrative decisions and hence also difficult to assess their efficiency. These concerns are exacerbated for policies affecting complex systems, given the epistemic challenges they pose even under ideal conditions. The consequence of all this is that regulatory decisions might not earn sociological legitimacy on the basis of their causal efficacy. Instead, some other means of legitimation may be necessary. Democracy, I have suggested, can help supply it.

I offer three complementary ways that OIRA can reinforce democracy through the goal-setting process. These proposals reflect my contextual judgment, rather than any putatively general answer to the “democracy question.”⁵⁵ I aim only to illustrate how agencies might justify procedural reforms in terms of citizens’ freedom and equality, not to settle contested theoretical and institutional debates.⁵⁶

First, and most ambitiously, OIRA should open goal setting to greater public participation.⁵⁷ Here, there are opportunities to experiment, including by introducing “lottocratic” institutions. In recent years, some political theorists have argued that some political power should be shifted to assemblies of randomly selected citizens.⁵⁸ Where they have been convened, these assemblies usually gather for specific purposes and for limited times and incorporate expert presentations. The core idea is that these lottocratic bodies may better represent what the citizenry at large wants, as compared to elected legislatures. Citizen assemblies are not perfect; indeed, they remain vulnerable to capture by special interests.⁵⁹ But they may still have a limited but beneficial role within the administrative state.⁶⁰ In particular, citizen assemblies may offer administrators an additional “picture” of public opinion, one less shaped by the limitations of polling or the latest developments in partisan messaging.⁶¹ And depending upon how they are structured, citizen assemblies may facilitate deliberation between citizens and administrators, encouraging greater governmental responsiveness to citizens’ reasoning.⁶²

Second, OIRA should promote deliberation among agencies in setting systems-level goals. Interagency consultation is a familiar feature of administrative law,⁶³ and OIRA already plays a leading role in facilitating it.⁶⁴ Consultation’s specifically democratic potential, however, has been underappreciated.⁶⁵ In many statutory contexts, interagency consultation serves two main purposes: it ensures that decisions are made after discussion with those possessing relevant expertise and it delays actions that may have unintended consequences. In systemic goal setting, its function can be different. Like citizen assemblies, administrative agencies can be understood as representing the citizenry: they effect democratically determined purposes, working out their implications in ways that reflect both ordinary politics and enduring principles. Consultation and collaboration among agencies allow these fragmentary images of “the people” to coalesce. Together, they can provide a more complete picture.⁶⁶ Whereas ordinary regulatory coordination may benefit “lead agencies” as first movers, goal setting can accommodate deliberation on terms reflecting the importance of agencies’ claims and the depth of those claims’ popular support. The result would be a process that better reflects citizens’ varied commitments.

Third, and finally, OIRA should implicitly acknowledge the limits on the President’s democratic authority, even as it continues to foster forms of “presidential administration.”⁶⁷ The President can rightly claim to be “a spokesperson for public opinion” when guiding the execution of the laws—but only a spokesperson.⁶⁸ Public opinion can, does, and should manifest in other ways, as I have just suggested. Recognizing other sources of democratic authority would help constrain the President’s own claims. This should partly be a matter of role morality: OIRA staff should understand their work in terms beyond the familiar synthesis of “external political control” by the President and “internal professionalism.”⁶⁹ But recognizing the limits of “presidential administration” should also involve openness to setting systems-level goals through procedures that encourage “full and fair ventilation of reasons for action in changing contexts that resist hierarchical control.”⁷⁰ So understood, OIRA’s claim to lead in goal setting would be based not upon its political stature but upon its capacity to integrate various forms of professional knowledge and political claims.⁷¹

Principle #3: The process of developing structural-risk regulations should reinforce democratic values.

Democracy is valuable not only in choosing systemic goals but also in selecting means for achieving them. Policy making, no less than goal setting, involves contested value questions; the moral legitimacy of those decisions must be maintained over long time scales; and implementing public policy over those time scales requires maintaining social trust. In some respects, administrative law already reflects these concerns: in comparative terms, notice-and-comment rulemaking is an unusually participatory process, despite its imperfections.⁷² Regulatory review, however, can help ensure that rulemaking procedures advance democracy, beyond what is required by law.

Regulatory review already considers procedure, but not for democratic reasons.⁷³ OIRA routinely asks agencies to list alternatives to its proposed rules during notice-and-comment and to seek public comments on specific issues.⁷⁴ These requests, however, generally serve epistemic rather than moral purposes: they mostly help ensure that agencies have sufficient information. OIRA does also serve a “Frankfurterian” role, as a guardian of “fair process,” including by encouraging “participation by a large number of people inside and outside the federal government.”⁷⁵ But this role typically involves neither explicit consideration of political morality nor substantial additions to settled administrative practice.

Regulatory reviewers interested in promoting democracy should begin by critically assessing the statutory context. In general, as I noted above, the administrative state claims legitimate democratic authority through statute. But for any particular agency action, even one authorized by law, statutory authority may be more or less democratically robust. For example, a relatively recent statute vindicating citizens’ equality may provide a robust democratic basis for regulation, even if it asks administrators to answer complicated value questions. In administering this statute, regulators may not need to do much more to deepen democracy; following ordinary procedures may suffice. By contrast, an older statute inconsistent with subsequent democratic practice may be on shakier ground, even if its scope is narrower.⁷⁶ In executing these laws, regulators should be especially attentive to democratic values. Judging statutes as more or less democratic will not be easy. But lawyers can help technical experts mediate between scientific, institutional, and moral considerations.⁷⁷ And over time, regulatory reviewers may become more accustomed to deliberating in terms of democracy.

Recognizing possible democratic deficits in agency authority does not mean conceding that regulatory proposals implicate “major questions.” In West Virginia v. EPA, the U.S. Supreme Court explained that in “certain extraordinary cases,” an “agency must point to ‘clear congressional authorization’ for the power it claims.”⁷⁸ These cases are identifiable by “‘the ‘history and the breadth of the authority that [the agency] has asserted[]’ and the ‘economic and political significance’ of that assertion.’”⁷⁹ This interpretive inquiry is distinct from the normative inquiry I have recommended to administrators. To be sure, structural-risk regulations will often be of great “economic and political significance.” But even claims of regulatory authority with robust precedents may be democratically deficient if those precedents failed to advance citizens’ freedom and equality. Agencies could therefore conclude that democracy promotion is appropriate in developing structural-risk regulations, without signaling to courts that it is claiming newly broad authority.

After examining the statutory context, regulatory reviewers should ask agency staff to justify their procedural choices in democratic terms. In general, procedure and substance are “interdependent variables in the constrained optimization problem” of regulatory policy making.⁸⁰ That is, an agency’s preferred procedures depend in part upon the substance of the policies it pursues, and vice versa. This implies that regulatory review focusing on one to the exclusion of the other might miss a significant part of the picture. And this is no less true for review focused on democratic values. In practice, engaging in “democratic review” might involve OIRA asking line agencies to consider the same democratizing reforms I discussed above. Citizen assemblies and interagency dialogue might usefully complement statutory procedures in some contexts.

Two caveats apply here.⁸¹ First, OIRA should not categorically demand any particular extra-statutory procedures for developing structural-risk regulations. In shifting from setting goals to selecting means, it will matter more exactly which structural risk agencies are trying to manage. For example, solar geoengineering and artificial intelligence might both involve structural risks, but regulations for the two will look rather different. Precisely because substance and procedure are interdependent, it is difficult to make general determinations about the procedures most appropriate across regulatory domains. Citizen assemblies and interagency consultation should be salient options for promoting democracy in policy making, but they may not be appropriate in every case.

Second, though OIRA should remain a “guardian of a well-functioning administrative process,”⁸² it should not be overly demanding. OIRA is mostly staffed by regulatory generalists, while line agencies are staffed by subject-matter specialists. This means that there are important informational asymmetries between the two, and the gap cannot always easily be bridged.⁸³ In general, observers tend to note these asymmetries in technical contexts: an EPA staffer, for example, may have devoted decades to understanding some pollutant. But if substance and procedure are interdependent, then the asymmetry matters more broadly as well. Whatever advantage OIRA officials have over line-agency staff with respect to regulatory procedure is limited by their disadvantage with respect to regulatory substance, since the two are not fully separable.⁸⁴ OIRA’s capacity to democratize structural-risk regulation may therefore be limited, though still significant.

This subpart shifts from how structural-risk regulations should be developed to what they should do and offers three more principles for regulatory review. My central recommendation is that regulatory reviewers should assess (1) how new regulations are likely to alter the political environment and (2) how those political developments are likely to influence public policy. This inquiry, I argue, should be both positive and normative. In developing this recommendation, I show how emerging social-science research—apart from the economics expertise already well-represented in OIRA—can contribute to democratizing regulation.⁸⁵

Principle #4: Regulation of structural risks should be democratically durable.

Regulators should attend not only to whether policies are formulated democratically but also to whether policies are democratically durable. Roughly speaking, I use “democratic durability” to refer to the idea that policies should be resistant to unintended changes driven by external circumstances but amenable to deliberate, democratic adjustment. This, in turn, combines two related aims.

First, policies should generally resist unintentional “drift.” “Drift” refers to “changes in the operation or effect of policies that occur without significant changes in those policies’ structure.”⁸⁶ In general, drift occurs because the context in which policies operate changes. Drift may be either expansionary (e.g., the increase in Social Security Disability Insurance claimants between the 1950s and the 2010s)⁸⁷ or contractionary (e.g., the decrease in unionized workers over a similar period).⁸⁸ Expansion or contraction may sometimes be beneficial, depending upon the circumstances, and in some cases, policy makers may intentionally use leverage drift for their purposes. But when drift is unanticipated, its benefits are merely fortuitous. Policy expansion is likely to impose undue costs; policy contraction, to achieve insufficient benefits.

Second, policies should reinforce democratic values over time. Whether a political decision is democratic depends not only upon how it was made but also upon what it does and how it is implemented. In particular, even laws enacted democratically can prove anti-democratic over the long run if they undermine future democratic agency. In some cases, such as filling a wetland, democratic concerns may arise because decisions wholly foreclose some choices, no matter who would make them.⁸⁹ In others, such as authorizing presidential emergency powers, problems may result from decisions’ effects on the distribution of political power. Policies limiting democratic agency may still be desirable, all things considered.⁹⁰ But insofar as democracy is valuable, there are always pro tanto reasons to avoid such constraints.

Both aspects of democratic durability are especially important for structural-risk regulations. First, consider drift resistance. For policies to resist drift need not mean that they never change. Rather, it can also mean that they change along with the circumstances through built-in adaptation mechanisms while maintaining the same “structure.”⁹¹ Drift resistance, then, can be “dynamic” as well as “static.” Because structural risks exhibit increasing returns and threaten greater harm over time, it is especially important that structural-risk regulations be dynamically drift resistant. For example, policies causing brief, one-time reductions in greenhouse gas emissions may (or may not) be beneficial. But their efficacy will be limited because greenhouse gases will persist in the atmosphere and fossil-fuel infrastructure will remain in use. By contrast, drift resistance is less crucial for risks that do not increase or become costlier to avoid over time.⁹²

Regulations’ substance bears upon democratic values in much the same ways as does the regulatory process. Respecting citizens’ freedom and equality means respecting it over time: future generations must not be unduly burdened by the “dead hand of the past.” Structural-risk regulations—even just by virtue of their sheer significance—can entrench new distributions of power, whether formally or functionally.⁹³ To the extent these distributions prevent future citizens from being governed in ways that respect their freedom and equality, they are objectionable on democratic grounds. Moreover, structural-risk regulations’ significance heightens these objections’ importance. Making collective decisions in terms of freedom and equality is most valuable when those choices affect the very structure of society, the economy, the environment, or other large-scale systems.⁹⁴

I propose two complementary methods for pursuing democratic durability in structural-risk regulation, which I call “policy-feedback analysis” and “democratic-feedback analysis.” Both would improve regulatory review by incorporating and extending techniques now emerging in political science and other disciplines.⁹⁵

Policy-feedback and democratic-feedback analysis correspond, respectively, to resistance to drift and preservation of democratic agency. Policy-feedback analysis would ask: how are certain policies likely to affect the political context in which those policies operate, and how are those changes likely to affect the operation of those policies? Democratic-feedback analysis would likewise start with an examination of the political consequences of policy choices, but then would inquire: how are those political changes likely to promote or hinder democratic values? (Here, then, “democratic agency” refers not to any strong form of collective self-government, but only to political agency exercised through institutions justifiable in terms of citizens’ freedom and equality.) Both policy-feedback and democratic-feedback analysis would formalize some of the considerations that already inform policy development, while also encouraging long-term analyses of whether and how “new policies” will create “new politics.”⁹⁶

Policy-feedback analysis presumes that while sociopolitical dynamics are not fully predictable, the forces governing them can be sufficiently well understood to inform regulation.⁹⁷ In particular, I emphasize three major determinants of feedback effects: policy instruments (i.e., the proposals under review), initial conditions, and feedback mechanisms. These can be modeled together to develop an account of a policy’s likely effects over time.⁹⁸ Whether these “models” will be relatively formal and quantitative or relatively informal and qualitative depends upon the scientific and social-scientific “state of the art.” For this reason, particular modeling techniques may be feasible in some areas but not in others.

In focusing on policy instruments, I emphasize that feedback effects may not be uniform across a given regulation. For example, an agency may promulgate a legislative rule that includes both an ambient standard and a public-disclosure requirement for the same pollutant. Though these requirements might work in tandem, there is no ex ante reason to assume that their political consequences will be the same. Each will have its own cost–benefit structure, and each may empower different groups in different ways.⁹⁹ These effects might then interact, creating further complexities. Regulatory reviewers should therefore consider proposed regulations both as bundles of separate instruments and as interconnected systems.¹⁰⁰

Policy-feedback mechanisms and initial conditions are both capacious categories. The major mechanisms are (1) resource/incentive mechanisms, which alter the resources available to actors and/or influence which choices are available to them, and (2) interpretive mechanisms, which alter information flows and, as a result, shape how actors interpret the world.¹⁰¹ Feedback from these mechanisms may be either positive (i.e., self-reinforcing) or negative (i.e., self-undermining). Much political behavior could fall under these labels. For example, developing administrative capacity, directing funds to interest groups, or imposing criminal sanctions might all count as “resource/incentive mechanisms”; distributing propaganda, providing technical training, or changing social norms might all be “interpretive mechanisms.” Similarly, relevant initial conditions can include anything related to “resources” or “information.”

Because policy-feedback dynamics will often be complex, regulatory reviewers should seek outside input in constructing their models. Following recent work modeling sociopolitical feedbacks in the climate system,¹⁰² I propose that regulatory reviewers convene interdisciplinary workshops to identify which particular mechanisms are most germane to particular structural risks.¹⁰³ They may also glean useful information from the institutions I have recommended for setting systemic goals, including citizen assemblies and interagency consultation. With respect to “ordinary” citizens in particular, there is good reason to think that current administrative practice underrates the epistemic resources available. Under the right conditions, the cognitive diversity of “the many” can prove valuable.¹⁰⁴

Democratic-feedback analysis would involve subjecting the conclusions of policy-feedback analysis to normative scrutiny. This would involve difficult theoretical questions—most obviously, what counts as democratic or undemocratic?¹⁰⁵ And even where there is broad agreement on definitions, applications might prove vexing. For example, a policy might empower some interest groups over others but reduce material inequalities as a result. In cases like this, what “democracy” requires might be unclear.¹⁰⁶

Despite these challenges, democratic-feedback analysis would offer two main benefits. First, like cost–benefit analysis, it would force regulatory reviewers to make their considerations explicit. This process alone may be valuable: even if OIRA staff, for example, are already thinking about the democratic implications of proposed policies, they may miss certain considerations simply because they are not putting their thoughts to paper. Democratic-feedback analysis could not guarantee good normative theorizing, but it would probably secure better results than leaving moral reasoning implicit. Second, and relatedly, democratic-feedback analysis can “mak[e] bureaucracies think”¹⁰⁷ about the democratic consequences of their choices. Regulatory reviewers often conceive of themselves as relatively apolitical.¹⁰⁸ Democratic-feedback analysis would encourage them to become “political” in the specific sense of being explicitly committed to advancing citizens’ freedom and equality.

Three considerations suggest that these benefits are attainable, despite the theoretical challenges. First, administrative actors already engage in noneconomic normative deliberation. For example, Executive Order 13,563 permits agencies considering regulations to weigh “values” intimately connected to democracy, including “equity, human dignity, fairness, and distributive impacts.”¹⁰⁹ Scholarship on “administrative constitutionalism” has also shown that, long before the Executive Order, administrative actors “elaborat[ed] … new constitutional understandings” relevant to democratic values.¹¹⁰ Second, and more theoretically, advancing democracy does not require endorsing some comprehensive democratic ideal. It is enough to offer reasons for democracy’s real-world value and then to promote policies and institutions that reflect those reasons.¹¹¹ Certainly, existing regulatory practice does not presume that making decisions requires resolving every theoretical disagreement. Finally, enhancing public participation in rulemaking—as I proposed above—can help regulatory reviewers develop workable conceptions of democracy. Democracy itself would be subject to democratic deliberation.

The Administrative Procedure Act does not foreclose this deliberation. Scholars have debated whether and how judges should credit regulators’ political considerations when engaging in “arbitrary and capricious” review.¹¹² But regardless of whether they merit deference, “political” decisions are permissible—at least where they are otherwise justifiable. Indeed, in Department of Commerce v. New York, the U.S. Supreme Court explained: “Agency policymaking is not a ‘rarified technocratic process, unaffected by political considerations or the presence of Presidential power,’” and the Administrative Procedure Act does not treat it as such.¹¹³ That case, together with Department of Homeland Security v. Regents of the University of California,¹¹⁴ suggests a central concern in judicial doctrine for enhancing political accountability.¹¹⁵ To the extent that policy-feedback and democratic-feedback review encourage accountability over time, they comport with the broad thrust of recent case law.¹¹⁶

Principle #5: Regulation of structural risks should incorporate triggers for review and revision.

Though policy-feedback and democratic-feedback analyses aim to provide holistic assessments of regulations’ democratic durability, it may also be helpful for officials to have rules of thumb for regulatory design. Here is one: in general, structural-risk regulations should include triggers for their own review and revision to facilitate democratic agency. That is, regulations should include provisions requiring either new deliberation or new policy-making processes, effective upon some predetermined date or upon some other change in the regulated system.¹¹⁷

Regulatory triggers can help fill a potential lacuna in democratic-feedback analysis. Democracy, political theorist Jan-Werner Müller has explained, “has a dual nature”: “First, it requires a designated locus (and specific times) for collectively binding decision making,” and second, it needs “a place for the continuous formation of opinions and political judgments in society at large.”¹¹⁸ Democratic-feedback analysis will likely fare better at weighing the latter than the former. Regulatory reviewers may be able to predict whom a policy might empower in the public sphere but not how it will affect decision-making. Regulatory triggers provide no certainty about future decisions’ content, but they can help ensure that decisions will be made. This information will often be important to assessing a proposal’s likely democratic credentials over the long run.

Regulatory triggers, like policy-feedback and democratic-feedback analysis, would especially suit structural-risk regulations. In every domain, changing circumstances can demand new policy responses. But structural risks demand particular flexibility, in view of their significant possible nonlinearities. In some cases, more stringent regulations may urgently become necessary; in others, regulations may quickly become obsolete. In other words, the threat of policy drift is likely to loom even when regulations are designed to be politically resilient. And drift, in turn, can have significant political consequences. Nonlinear increases in risk can create “states of exception,” in which ordinary legal principles are suspended and power can be abused.¹¹⁹ Triggers can help policy regimes adjust before disaster strikes.

Current regulatory schemes often do not respond adequately to these challenges. Regulations do generally contain conditional provisions. For example, a workplace-safety rule may apply only to employers with at least 100 employees. These schemes make application of law responsive to changes in the world: when an employer hires its 100th employee, it becomes subject to the rule. But this responsiveness is often limited. It exists within a broadly stable policy framework and is given effect primarily through enforcement. Regulatory triggers today therefore do not tend to offer new forums for democratic agency. They also offer only limited protection against policy drift: insofar as regulatory triggers apply only to particular regions or parties, they may ignore the interdependencies characteristic of structural risks.

It would be preferable to formulate triggers for large-scale review and revision of regulations. These would provide “thresholds of alarm” responsive to systems-level information.¹²⁰ Regulatory thresholds should probably not be based upon new, comprehensive risk assessments—otherwise the costs of delay would be too great—but instead upon previously known indicia of systemic risk. Here the “planetary boundaries” framework in earth science is illustrative. That approach defines the “safe operating space” for humanity as a whole by identifying several quantitative “boundaries.”¹²¹ The boundary variables are interconnected, exhibiting both positive and negative feedback effects. It is possible to develop policies that use boundaries of this sort as regulatory triggers: when certain thresholds are crossed, new rules could take effect. Crucially, these new rules need not be incremental adjustments to existing rules. Substantially different regulatory frameworks may be better suited for new realities.

Regulatory triggers of this sort would be roughly akin to legislative “hammers.” For example, the “take title” provision of the Low-Level Radioactive Waste Policy Amendment of 1985 provided that if the states did not meet certain deadlines for providing for nuclear-waste disposal, the states themselves must take title to and possession of the waste and assume liability for associated damages.¹²² The “take title” provision therefore effectively created an initial, default regulatory regime and an alternative, conditional one. The former encouraged states to facilitate private waste disposal; the latter required states to handle disposal themselves. Taking inspiration from this structure, regimes governing structural risks could, for example, have triggers for switching from market-based to command-and-control instruments. Or they could extend regulations to new domains not initially covered. Regulations could also become less stringent and more flexible as needed, departing from the “hammer” model.

Principle #6: Administrators should be cautious in conducting cost–benefit analysis for structural-risk regulations.

Cost–benefit analysis may be of relatively limited use in assessing systemic-risk regulations in general and structural-risk regulations in particular. Rather than dominating regulatory review, cost–benefit analysis should be “only one input into public policy.”¹²³

Costs and benefits cannot easily be calculated for systemically significant policies. I noted earlier that because the consequences of individual decisions depend upon their systemic context, it is difficult to aggregate component-level choices to obtain systems-level objectives. For the same reason, it is not easy to anticipate the consequences of any component-level choice when the system is subject to significant change. This point is most intuitive when considering policy questions that are “so complex or large-scale that it is impossible to competently forecast the costs or benefits of a proposed regulation:”¹²⁴ these regulations may “alter the constellation of relative outputs and prices over the whole economy,” such that the rest of the world cannot be held constant.¹²⁵ But even apparently smaller choices can frustrate the policy calculus. “[S]ystem-wide costs and benefits that accumulate over time” can eventually “change the fundamental character of society.”¹²⁶

Regulatory reviewers should ask whether policies are likely either to effect system-wide change or to take effect amid system-wide change. If they are, then reviewers should take special caution in using cost–benefit analysis. By contrast, where policies would cause more incremental adjustments or where the economic and social system is likely to remain relatively stable, cost–benefit analysis may be relatively unproblematic.¹²⁷ In counseling caution in some cases, I do not embrace the categorical conclusion that “CBA is not an appropriate tool” in assessing structural risks.¹²⁸ For one thing, the risk of systemic change (even if unintended) is never zero; it is always possible that the background conditions enabling cost–benefit analysis could disappear. There is therefore no sharp, ex ante distinction between cases where cost–benefit analysis is appropriate and those where it is not. Moreover, alternative decision procedures (including those I have outlined here) are not costless.¹²⁹ Cost–benefit analysis’s limitations do not alone determine its proper role, absent consideration of what could replace it.

Regulators may sometimes be able to integrate policy-feedback and cost-benefit analysis. Policy-feedback analysis aims to anticipate the interactions between proposed regulations and their political contexts, in both directions. So understood, it is, in part, a study of regulatory efficacy.¹³⁰ In some cases, efficacy will be understood in relative terms—for example, by comparison to the intentions of a program’s creators. But in other cases, efficacy may be absolute. Policy-feedback analysis may suggest that the net benefits of some policy are likely to change over time, and these changes may be quantifiable (if only roughly). Where quantification is possible, there should be no firm boundary between policy-feedback and cost–benefit analysis. Cost–benefit analysis can provide an initial assessment of a program’s cost structure. This assessment can inform predictions about its policy feedbacks, and these, in turn, can reveal more costs and benefits.

More speculatively, regulators might also refine cost–benefit analysis in light of democratic decision-making. Cost–benefit analysis generally assigns values “through direct use of, or in analogy with, the logic of market allocation.”¹³¹ Market valuation may have merits in some contexts. But it cannot easily value goods that citizens treat “specifically” as “social concern[s],” particularly with respect to existence values.¹³² For these goods, political processes offer an alternative mechanism of valuation. Politics is, of course, flawed: it cannot reliably aggregate individual preferences with any precision, and even if it could, it would not perfectly capture the sorts of ideal, public-minded judgments that critics of cost–benefit analysis seek.¹³³ Still, regulatory reviewers may credit democratic procedures—both legislative and regulatory—as providing rough guidance on appropriate quantitative valuation. This guidance will usually be qualitative, though recent research indicates the potential for gleaning quantitative information from democratic norms as well.¹³⁴

In this part, I respond to some possible concerns. My aim here is not to insist that my proposed approach would be without drawbacks—no reforms would be—but instead to show that its problems would be manageable.

The democratic approach could not easily be administered by current regulatory reviewers alone. Some of what I have proposed, such as considering the democratic credentials of agency procedures, would extend OIRA’s existing roles. But others, such as convening citizen assemblies or engaging in policy-feedback review, would be more novel. These tasks would require expertise not currently well-represented within OIRA. OIRA currently has career staff with graduate degrees in “economics, law, policy analysis, statistics, and information technology,” as well as various “technical fields.”¹³⁵ Yet these subject areas, for all their value, do not exhaust insight into the distinctively political considerations I have described here. Without adequate personnel, the federal government would likely lack sufficient state capacity to undertake the full panoply of democratizing reforms.¹³⁶

Staffing should not be viewed as an insuperable barrier. President Biden’s “Modernizing Regulatory Review” memorandum directs the OMB Director to “consider ways that OIRA can play a more proactive role in partnering with agencies” to pursue beneficial regulations and to “identify reforms that will promote the efficiency, transparency, and inclusiveness of the interagency review process.”¹³⁷ These goals do not require OIRA reform, but they are highly open-ended. Increasing OIRA’s staff and broadening its professional diversity can be an effective strategy for meeting the President’s objectives. “Modernizing Regulatory Review” affords a rare opportunity to reflect upon OIRA’s structure: OIRA has consolidated its essential regulatory role over the last four decades, and it could now effectively manage the expansion of its ranks. In short, there is good reason not to treat the institutional status quo as fixed. “Modernization” need not consist solely of adopting new analytical techniques.

If the composition of OIRA and other agencies should not be taken as given, the challenge is determining who should join their ranks. Part of the answer lies in administrators’ academic backgrounds. Staff trained in political science, political economy, and political theory, for example, might have much to contribute to regulatory review. Yet these disciplines as currently constituted have their own limits. For example, political scientist Jacob S. Hacker notes that his field is ill-prepared to embrace “a policy mindset that takes seriously imbalances of power in the political economy, grapples with the challenge of building durable governing capacities, and focuses on the political as well as economic effects of public policies.”¹³⁸ For now, it remains to identify personnel with the personal, professional, and intellectual attributes necessary for pursuing democratizing reforms.

Worries about personnel also relate to a broader concern: large-scale administrative reforms require effective institutionalization. So far, I have been discussing political constraints much as though they applied only to policies administered, and not to administration itself. But OIRA, and the administrative state more broadly, is not insulated from the struggle for power.¹³⁹ Nor is it immune to subtler pressures stemming from “the nonrational pressures that arise out of the administrative process,”¹⁴⁰ including those related to “group identification, status, and relationship networks.”¹⁴¹ Dynamics political, cultural, or otherwise might generate resistance within agencies (including OIRA) to my proposed reforms.¹⁴² Even some who view my ideas sympathetically might think their development futile.

This objection points to the need for implementation to extend beyond OIRA. Just as the “economic style of reasoning” in public policy ascended only over decades,¹⁴³ so, too, will institutionalizing alternatives take time. In part, this is a matter of how the reformers approach their task. To make changes in regulatory review “stick,” explains scholar and former OIRA official Bridget C.E. Dooling, “some combination of internal development, interagency review, external peer review, and public comment may help firm up the basis for the changes.”¹⁴⁴ But though I have emphasized the promise of “Modernizing Regulatory Review,”¹⁴⁵ any single presidential initiative is likely insufficient to support substantial change. Deeper reform would benefit, for example, from the support of higher-education institutions¹⁴⁶ and professional networks.¹⁴⁷ These, too, offer no panacea, not least because universities and professions have their own durable cultures.¹⁴⁸ But they are nonetheless important forums for the larger project of making democratic equality more salient in regulation.

The democratic approach involves potentially burdensome new procedures. In regulatory review, no less than in other administrative contexts, procedure is no panacea.¹⁴⁹ Perhaps most importantly, the “ossification” associated with bureaucratic process might prove greater than whatever benefits that process brings.¹⁵⁰ This concern is especially significant with respect to structural risks, given that their harms can quickly multiply. And it is exacerbated still further by the current combination of political polarization and growing “anti-administrativism:”¹⁵¹ if a new rule is delayed, the next presidential administration may be unwilling or unable to carry forward its promulgation.

“Ossification” is a serious concern, but the process costs of the democratic approach need not be too high. Most important here is what the democratic approach does not involve: expanded judicial review. Some have argued that judges should accord or withhold deference based upon the democratic quality of agencies’ decision-making processes.¹⁵² But whatever the judiciary’s proper role, it does not determine the executive’s. I have argued only for new forms of review by OIRA staff, not by judges, and the former would not lead to new litigation delays. Moreover, much as OIRA’s staff need not remain fixed, neither need its procedures. OIRA could still review agencies’ decisions not to regulate.¹⁵³ If necessary, the OIRA Administrator could waive review of significant regulations to prevent harmful delay.¹⁵⁴ And over time, OIRA could collaborate with agencies to develop more efficient review practices.

Concerns about capture mainly take two related forms. First, some might object to “politicizing” regulatory review. Under the previous presidential administration, cost–benefit analysis was “subject to flagrant manipulation.”¹⁵⁵ Further politicizing regulatory review—even in democracy’s name—may seem like a risky proposition, especially in the context of structural risks. Indeed, one advantage of regulatory review might lie precisely in stabilizing otherwise-contentious regulatory processes with the technical language of costs and benefits. Second, some might be more open in principle to politicization but worry that mechanisms meant to promote democracy would in fact facilitate regulatory capture.¹⁵⁶ Multiplying the channels for influence over public policy might only serve the powerful.

Capture is indeed a concern, but not one the democratic approach would likely exacerbate. It is true that my proposals would introduce novel considerations into regulatory review. But there is little new danger in politicizing what has always been political. OIRA had an anti-regulatory bent for decades, notwithstanding strong policy arguments to the contrary.¹⁵⁷ If anything, the democratic approach might counteract certain forms of politicization by embedding within OIRA pro-democratic and pro-regulatory commitments at odds with the office’s historical tendencies. For similar reasons, my proposals should not be particularly vulnerable to capture. The democratic approach would not propagate procedural safeguards against abuse of state power. Rather, it would aim to channel that power toward particular substantive purposes, relying mostly on administrators themselves. To the extent my suggestions would create new opportunities for influence, they would be designed to resist the influence of special interests. Corruption is always possible, but I believe the reforms I outline would not make it worse.

More generally, the framework I have developed is flexible. In proposing that administrators consult assemblies of randomly selected citizens, for example, I suggested they do so for only limited purposes.¹⁵⁸ But my underlying point was that the process of developing structural-risk regulations should reflect and reinforce citizens’ freedom and equality. If it turns out that “lottocracy” has more potential to accomplish this than I have recognized, I have no basic objection to expanding assemblies’ role. Similarly, I argued that administrators could help promote democratic equality by influencing regulations’ design.¹⁵⁹ It is possible that some might construe “democracy” to support empowering certain private interests over the public interest.¹⁶⁰ Nothing could prevent all error. But this only underscores the importance of addressing the problems of personnel and institutionalization. Ultimately, sufficiently powerful coalitions, both within and outside the administrative state, must come to support reform. This is a serious challenge for implementation but not for the theoretical position that sustaining democracy requires attending to the political effects of public policy.

Scholars have argued (and argued) about cost–benefit analysis. They have grappled, too, with the “democracy question”¹⁶¹ in the administrative state. And they have certainly subjected OIRA to scrutiny. Yet so far, they have largely neglected democracy’s place in regulatory review.

I have offered a new approach. I have argued that, in confronting some of the planet’s most important challenges, democratic values can and should be embedded in regulations’ design. My proposals are hardly radical: they presume that federal regulators will continue to have an essential policy-making role. But they would require a reorientation: from instrumental to normative reasoning, from economics to political science. To guide this shift, I have refined the concepts we use to analyze important policy challenges, identifying a category of “structural risks.” I have outlined a new process for confronting these challenges, from setting broad goals to choosing regulatory specifics. I have shown how recent studies of policy feedback can inform the regulatory process. And I have proposed that an ideal of “democratic durability”: resistance to exogenous, nondemocratic change—and not just “efficacy,” conventionally understood—should be regulators’ lodestar when addressing structural risks.

This article has also raised questions for future research. Some are more conceptual: How should we understand the history of efforts to reform the administrative state from within, and what can that history tell us about reform efforts today?¹⁶² How should we understand democracy’s place in regulatory review outside the context of structural risks? What would democratic durability require given various other conceptions of democracy? Other questions are more institutional: How should courts treat regulators’ efforts to ensure democratic durability? How should regulators negotiate among different conceptions of democracy? What practical obstacles impede the pursuit of democratic durability? Beyond these lie still other questions of application and of methodological refinement.

I have not endeavored to answer all these. Rather, I have aimed to suggest new avenues for theory and practice. Democracy, I hope I have shown, need not be at odds with rational regulation. Indeed, in a society committed to citizens’ freedom and equality, it should be at regulation’s very core.