In his piquant memoir, Streetwise, Lloyd Blankfein, former CEO of Goldman Sachs, shares an arresting anecdote about the administrative state. The SEC sued Goldman for allegedly misleading investors about the risks of mortgage-backed securities. Blankfein considered the suit meritless but settled for $550 million. “You can’t litigate against your regulator. It’s like a Star Trek episode: an alien controls the Enterprise’s bridge and threatens to shut off life support unless the ship surrenders.”

　　The dilemma for the regulated firm is that even if it wins the suit, its regulating agency has many other ways to make life unpleasant. It can sue the firm about other matters. It can issue regulations particularly unfavorable to its business. And it can do this without personal cost to the regulators, because the expenses of suits and regulations, successful or not, are borne by taxpayers, ignorant of the merits—indeed even the existence—of their actions. As a result, citizens and corporations, even wealthy ones like Goldman Sachs, may be coerced into settlement.

　　This burden of the administrative state is difficult for courts to ameliorate on their own. The Roberts Court has issued rulings that have leveled the playing field between the regulator and the regulated at trial. In Loper Bright, for instance, the Court eliminated the interpretive edge accruing to the government by eliminating Chevron deference, under which the government could put any “reasonable” interpretation on a statute, even if it was not the best interpretation. In SEC v. Jarkesy, the Court held that regulators could not seek legal remedies for matters that had a common-law analogue before their own administrative judges, but had to prove their case instead to a jury of disinterested citizens. In Axon, the Court held that the targets of agency enforcement could bring structural challenges to the nature of the agency (as, for instance, when the agency heads were improperly appointed or insulated from control by the president) upfront rather than be forced to endure the expense of the proceeding before they could be heard on such salient constitutional issues. All of these changes have made the administrative trial process fairer, but they do little to address Blankfein’s concern about whether it is rational to seek a trial given the leviathan that modern administration has become.

　　Regulators have many arrows in their quiver. Even if one is errant, they can easily aim at their target with another. Their current regulations can lead to enforcement actions against a wide range of company actions. They also have the authority to make new rules, and the policy choices they are authorized to make can hurt some companies in an industry more than others. And according to the Administrative Procedure Act, such choices are only lightly policed, struck down only if “arbitrary and capricious. The question for a company is thus never just “can we win this case,” but “how can we get on with the people who control our future?”

　　Administrative settlements should be seen as a distinct aspect of the administrative state. Such settlements often establish compliance systems, reporting duties, or restrictions that influence industry norms. As a result, policy can be shaped not only through rulemaking or judicial decisions, but also through pressure, negotiation, and agreements.

　　One might respond that this is no different from the plea bargaining that happens in criminal law. The argument goes that defendants settling cases this way are just acting rationally in the shadow of the law, giving up the right to go to trial in return for some reduction in sentencing. But whatever the arguments for plea bargaining, settlement in the administrative context is different in certain respects. First, because both the company and the agency are repeat players, destined or doomed to interact over the decades, they are not acting only under the shadow of the law applied to the current dispute, but also under the shadow of the future exercise of agency power. The target of enforcement fears not only today’s complaint, but tomorrow’s exam, approval, license, merger review, or enforcement sweep. Second, a plea bargain is narrowly focused on sentencing. Federal prosecutors are limited to recommending jail, house arrest, or probation. Administrative agencies can impose a broader range of conditions that can affect industry norms. Any sentencing recommendation from the prosecutors is advice to the judge, who retains authority to impose his own sentences within wide limits of discretion. In contrast, judges’ ability to reject administrative settlements is far more circumscribed.

　　A government that can coerce a settlement from a firm the size of Goldman Sachs can do still more to smaller firms and ordinary citizens.

　　Administrative government by settlement has not gotten as much attention as it deserves. We do have two fine conceptual articles. Lars Noah wrote on “administrative arm-twisting,” arguing that agencies can use their leverage over the regulated to get more favorable settlements than they could otherwise have and sometimes with terms that the law would not otherwise allow. Matthew Turk has shown how “regulation by settlement” creates a distinct form of agency regulation that lacks the judicial oversight and other safeguards that accompany adjudication and rulemaking. But I know of no work that has comprehensively catalogued the settlements of major agencies and how they shape the law.

　　This aspect of the administrative state provides a counterpoint to the notion that the Roberts Court is undertaking a fundamental counterreformation of administrative law. Axon, Jarkesy, and Loper Bright equalize the playing field at the backend when a case gets to trial. But administrative settlements show that asymmetric power comes at the front end, due to the agencies’ pervasive influence over regulated enterprises. In short, strengthening judicial administrative review may make less of a difference if the agency has other means to make access to that review unbearably costly.

　　The point here is not to vindicate Goldman or Blankfein in this particular matter or do special pleading on behalf of Wall Street. It is instead worrying as a friend of liberty that the administrative state, by its nature, can act on the basis of power rather than by rule. Montesquieu—and his students among the American Founders—long ago sought to protect liberty through the separation of powers. But regulation by settlement affronts the separation of powers in two respects. First, insofar as agencies have used other powers, such as rulemaking, to gain advantages in enforcement, they enhance the executive power of prosecution by combining it with traditionally legislative powers. Second, insofar as civil settlements escape substantial judicial review, the executive power faces little check from the traditional power of the third branch. Firms less powerful than Goldman will face even greater coercion from this relaxation of the separation of powers.

　　There are ways to ameliorate the problem. The first, recommended by Turk, is to substantially strengthen judicial oversight of settlements. As with class action settlements, they should receive public hearings. If any agency seeks to regulate through settlements, settlements should become more transparent. And Turk suggests that agencies should be required to consider the collective effect of their settlements. An agency is charged with regulating an industry, and even in individual settlements, industry-wide focus must be maintained.

　　While those suggestions are welcome, they are more focused on ensuring the deliberative rationality of settlements than on protecting against the coercion about which Blankfein complains. Noah urges Congress to constrain agencies’ ability to use powers beyond the specific enforcement action to twist the arms of the regulated and wants to ensure that settlements cannot pursue ends that the agency could not otherwise obtain.

　　In my view, this is an area also ripe for executive action. President Trump has improved the fairness of administrative actions in a variety of ways. For instance, in one executive order, he required that agencies keep all their guidance documents in a single searchable database so citizens can better predict how agencies will act. In another order, he required agencies to inform the regulated community in advance of new theories of jurisdiction or liability to avoid unfair surprise. President Trump could now issue an executive order directing agencies not to retaliate against any target of an enforcement action for going to court to resist, and requiring that the agency have legal authority for any conditions imposed in any settlement. The president can tie the hands of his subordinates in the interest of justice for American citizens.

　　When the retired CEO of one of our premier financial firms says that even his company could not bring the government to court on what it believed to be a meritless claim, we should worry that the administrative state can threaten the liberties of all. A government that can coerce a settlement from a firm the size of Goldman Sachs can do still more to smaller firms and ordinary citizens.

　　This is also a realist lesson about the limits of the Court’s power. The administrative state is not truly constrained merely because the Supreme Court announces elegant doctrines about statutory interpretation and the right to a jury trial. It will be constrained only when ordinary parties can refuse agency demands without fear that the agency will make them pay a price in the future. For that, we need congressional or, at the very least, presidential action. And we will get it only if the people demand that the government proceed by rule rather than by pressure.