Rethinking Preliminary Remedies

Eric Brooks^[1]*

Abstract

It is universally assumed that courts, when picking a preliminary remedy, should consider more than the legal merits. They also should consider factors like the “equities,” “public interest,” and “irreparable harm.”

But that assumption is mostly wrong. The idea behind it is that at the preliminary stage, the merits are too uncertain to give courts the full guidance they need. Yet this is not true. For example, judges are often unsure of the merits even after a final hearing. But few think that judges may therefore filter their final decisions through an equitable balancing test. And even if uncertainty did distinguish preliminary from final hearings, it would not justify current doctrine. By factoring in the equities, current doctrine allows courts to replace the substantive law with their own freelance efficiency analysis. But whatever effect a court’s uncertainty should have on its preliminary decisions, the effect cannot plausibly be to change the body of governing law. By calling for this, current doctrine is unsound.

What should replace it? To answer that, this essay focuses on what truly distinguishes preliminary hearings: not uncertainty, but the fact that the court will get to reconsider its decision later. To take fullest advantage of this chance, a court should consider not only the merits of its preliminary choices, but also those choices’ option value. Maximizing option value is a more constraining aim than the goals of current doctrine. This essay explains how to weigh option value against the merits to create a more accurate preliminary-remedy test.

Introduction

National Federation of Independent Business v. Department of Labor, OSHA hit the news last year for holding that OSHA’s vaccine-or-test rule was illegal.^[2] But to a certain crowd, the bombshell was what the Supreme Court did next.

The Court was considering whether to stay the OSHA rule pending full judicial review of it.^[3] Under textbook remedies law, this meant that it was supposed to consider more than whether the rule was lawful. Because stays are an “equitable” remedy, the case law tells us, the Court also should have weighed three other factors: “whether the applicant will be irreparably injured absent a stay”; whether “the stay will substantially injure the other parties interested in the proceeding”; and “where the public interest lies.”^[4]

In NFIB, though, the Court junked the last three factors. After finding that the rule exceeded OSHA’s statutory powers, it said:

The equities do not justify withholding interim relief. We are told by the States and the employers that OSHA’s mandate will force them to incur billions of dollars in unrecoverable compliance costs and will cause hundreds of thousands of employees to leave their jobs. For its part, the Federal Government says that the mandate will save over 6,500 lives and prevent hundreds of thousands of hospitalizations.

It is not our role to weigh such tradeoffs. In our system of government, that is the responsibility of those chosen by the people through democratic processes.^[5]

And with that: “The applications for stays . . . are granted.”^[6]

This ending was startling, at least to American lawyers interested in remedies: “Did the Supreme Court Overrule Equity?” asked one;^[7] might “Equity [be] dead (for stays),” pondered another.^[8]

To anyone else, however, NFIB’s coda should not have been terribly shocking. After all, when a court decides after a full hearing that an agency’s rule is illegal, it vacates the rule, with rare exceptions not relevant here.^[9] It may not keep the illegal rule alive because it thinks that the rule is nonetheless “equitable” or in the “public interest.” So once the Supreme Court found that the vaccine-or-test rule was illegal, the rule was doomed. It would have been strange if the Court had nevertheless kept the illegal rule on life support for a few months based on a “public interest” rationale that would soon become irrelevant. (And if that does not sound too bad to you, you probably would not love the reverse—if the Court had found the rule lawful, but then stayed it anyway because it thought the rule bad policy.)

NFIB was an easy case for ignoring the equities because the Court seemed sure that the vaccine-or-test rule was illegal.^[10] But its move cannot be cabined. In this essay, I will show how the principle behind NFIB implies a stronger result: that the Court should have ignored the “equities” and “public interest” even if it had not been sure of the merits.

In fact, the equities and public interest should never factor into a court’s preliminary decision, at least when considering these factors would also be verboten after a full hearing. Considering them at only the preliminary stage is not a benign procedural rule. It is a way to change the governing substantive law. At best, it tells courts to apply different bodies of law in different months, depending only on the vagaries of the court’s litigation calendar. At worst, it lets a court ignore the law in favor of its own policy views.

For instance, imagine if the NFIB Court had not been sure of the merits, and had thus gone with the equities. Then, for many months after January 2022, the government’s right to impose a vaccine-or-test rule would have hinged on a policy assessment—perhaps something like the arbitrary-and-capricious review that administrative decisions often get.^[11] But then one day—whatever day the district court got around to publishing its final decision—the right would instantly turn on a different legal framework, pure statutory interpretation. The textbook reason for treating the preliminary and final stages so differently is that during the earlier stage, the law was uncertain. But even if the court’s degree of uncertainty should factor into the preliminary analysis somehow, it is hard to see why the way it should matter is to change the governing body of law.

In fact, as I show in Part I, uncertainty does not even distinguish the preliminary and final stages of a case. Sometimes courts are unsure about the merits even after a full hearing. But that does not license them to throw up their hands and decide the case based on their sense of fairness. Instead, we expect them to make their best guess of the law and pick a final remedy accordingly. If they do not, we might accuse them of thwarting democracy. Uncertainty thus cannot be a reason to treat preliminary and final remedies differently.

In Part I, I also expose a more technical problem with current doctrine. Even if it were proper to weigh the equities or public interest at the preliminary stage, courts would need to do so in a way that is internally coherent. But the near-consensus theory of how to weigh these factors is unsound. The problem, in short, is that the theory gives those factors only half weight when, to be coherent, it would need to give them full weight.

While this criticism is separate from my first, the two problems are entwined. By giving the equities and public interest only half weight, the modern approach masks (but does not erase) its disrespect for the substantive law. That might be why the problem has gone largely unnoticed. But once we see the technical flaw, we will see that if the equities and public interest are to play any role in preliminary decisions, they should play a much larger role: the merits should drop out entirely. Because most judges would balk at that, it will seem more appealing to instead junk the other two factors.

But if the equities and public interest are out, should we treat preliminary and final remedies identically? No. I will show why in Part II. A court that wants to minimize the deviation between its rulings and the substantive law should want to minimize this deviation over the whole future, not just the deviation between now and the final hearing. Thus, when a court knows that it will have a second crack at a case, it should prefer (all else equal) a preliminary remedy whose effects can be reversed. So when deciding on a preliminary remedy, courts should consider not only the merits, but also the remedy’s option value, the room it leaves the court to reverse course later.^[12]

Maximizing “option value” might sound suspiciously like minimizing “irreparable harm,” a prong of current preliminary-remedy doctrine. But these aims are only cousins. Maximizing option value is a more constraining objective. While it still forces a court to consider more than just the merits, it frees the court from deciding whether a remedy’s effects count as harms. Plus, there turns out to be a significant class of cases in which all a court’s remedial choices will have nearly the same option value. In these cases, option value should fall out of the equation.

When option value washes out, we will end up with a startlingly simple test: courts should award a preliminary remedy exactly when they would issue a full remedy on the same record. In other words, in many cases, courts should simply decide the merits. There should be no separate test for stays or preliminary injunctions.

This would be a big change. But if a court’s job is to apply the substantive law, not to reinvent it for one stage of a case, then current doctrine is far off.

I. Our Flawed Approach to Preliminary Remedies

The modern approach to preliminary remedies has a critical flaw. Courts and scholars assume that because the law has not been “determined authoritatively,” courts should get more leeway to base their decisions on factors like the “equities” and “public interest” that would otherwise be forbidden.^[13] The intuition is that at the preliminary stage, the law is still uncertain, so we need these other factors to guide our decision. In this Part, I hope to erase that intuition. Along the way, I will also expose a serious technical flaw in the near-universal academic theory of preliminary remedies.

A. Prelude: How Courts Approach Final Decisions

First, consider how courts make final decisions. Once a plaintiff convinces the court that she is right on the law, she wins and gets a remedy. If she is wrong on the law, she loses. Either way, the court may not change who wins based on its own sense of the “public interest.” As the Supreme Court has put it, “[o]nce Congress, exercising its delegated powers, has decided the order of priorities in a given area, it is . . . for the courts to enforce them when enforcement is sought. . . . [T]heir choice is not whether enforcement is preferable to no enforcement at all.”^[14] This has been the official rule for centuries.^[15]

Until 2006, I could have relied on research by Douglas Laycock to make an even stronger claim: not only did a plaintiff who was right on the law get the win, but she also had her choice of permanent remedy. “Injunctions [we]re routine.”^[16] In federal cases, however, that is no longer quite true. In the patent case eBay Inc. v. MercExchange, L.L.C.,^[17] the Supreme Court, “[a]pparently by accident,”^[18] seemed to make it harder for federal courts to issue permanent injunctions. Before issuing an injunction, the Court said, courts should consider factors like the public interest and balance of hardships.^[19]

But while eBay has been cited often, the case has not “led to a great many differences in outcomes outside the patent arena.”^[20] The eBay Court itself did not seem to think that it was transforming the law.^[21] The Supreme Court still does not “routinely march[] through the four eBay factors in injunction cases.”^[22] And “[s]tate courts have mostly ignored” the case.^[23] So while the dust is still settling, there seems to be a large swath of cases in which, once a plaintiff wins after a full trial, she continues to have an easy time getting an injunction or injunction-like remedy.^[24] Courts will not routinely deny injunctions out of their own sense of the public interest.

This method of judging reflects a political theory. The idea is that the substantive law fully encodes all the relevant social values. If that is right, then there is no need to consider the public interest separately; the result that best serves the public interest is, by definition, whatever the substantive law dictates. So central is this axiom to many domains of law—especially statutory and constitutional law—that it comes up mostly in the alleged breach. One of the harshest things you can say in a dissent is that the majority “rewr[o]te the law simply because of [its] own policy views.”^[25] For then you are accusing the majority of “rob[bing] the People of the[ir] most important liberty”—“the freedom to govern themselves.”^[26]

Not everyone agrees that this axiom is coherent, that courts follow it, or that courts should.^[27] I do not mean to enter that debate. Instead, given how central the axiom is to how courts (at least) talk about the law, I will assume it and ask: Is it consistent with official remedies doctrine?

To start, note that courts do not claim to abandon the axiom after a full hearing just because the merits are still uncertain. For instance, once a court takes a side in a circuit split, it does not then consider whether, given the closeness of its legal call, the merits might be outweighed by the equities.^[28] Or take a case like American Bankers Ass’n v. National Credit Union Administration.^[29] The district judge had to decide whether a credit agency’s regulation complied with a federal law. She thought that the rule was “troubling,” “jarring,” and a “barely reasonable interpretation” of the statute.^[30] But “barely reasonable” is still reasonable. So the rule stood.^[31] And although the judge seemed to doubt the rule’s wisdom, she certainly did not suggest that, given the closeness of her statutory analysis, she should factor her policy doubts into whether to issue a remedy. She just followed the merits.

Every day in every court, judges make close calls. And once they do, they treat the call as a certainty and issue a remedy.^[32]

B. Flaw One: Weighing the Equities Wrongly

That is not remotely how courts approach preliminary remedies.

First, some housekeeping: By “preliminary remedy,” I mean any remedy that the court will have a chance to reconsider. This definition covers preliminary injunctions and agency stays. It also covers stays of lower-court decisions pending appeal. I will focus mainly on injunctions, which get the most attention. But courts purport to treat stays similarly^[33]—or at least they did until NFIB.

Before a federal court grants a preliminary injunction, it must consider not only the merits but also the “balance of equities,” the “public interest,” and the irreparable harm the plaintiff would face absent an injunction.^[34] State courts consider similar factors.^[35] The “equities” and “public interest” mean, roughly, the good or bad effects that an injunction will have on the world. Courts often classify effects on the parties as “equities” while treating effects on nonparties as part of the “public interest.” But these factors overlap, and I will use “equities” and “public interest” interchangeably to mean a remedy’s effects.^[36] (“Equities” can also cover other issues, like a party’s wrongful litigation behavior. I have no objection to considering the equities in this sense, and I will not discuss this issue any further.)

Courts are often vague about how they weigh the injunction factors.^[37] But there is a near academic consensus over how they should. As John Leubsdorf first argued, a court should issue a preliminary injunction when the injunction would minimize the expected amount of “legally unjustified irreparable harm.”^[38] Judge Richard Posner rewrote that test into a formula: courts should issue an injunction when

p*H_p>1-p* H_d,

where p is the probability that the plaintiff will win on the merits after a full hearing; H_p is the irreparable harm the plaintiff will face without an injunction; and H_d is the irreparable harm the defendant will face with one.^[39]

The Leubsdorf-Posner test’s “conceptual goal,” Justice Thomas Lee wrote, has become the “triumphant, dominant theory of preliminary injunctions.”^[40] “Today courts routinely apply a standard that is essentially consistent with Leubsdorf’s framework,”^[41] and one state has even expressly adopted the formula.^[42] In all, the test “finds almost no detractors or competitors in the academy or in federal jurisprudence.”^[43]

But the Leubsdorf-Posner test is flawed. It misunderstands the connection between “legally unjustified” and “harm.” It imagines that we—society, Congress, common-law courts, or whoever else you credit with writing the substantive law—first divide harms into the “justified” and “unjustified” sorts, and then try to minimize the latter.^[44] If that were true, then it would make sense to craft legal rules that minimize only legally unjustified harm.

But that is not how lawmaking works. Harms are not bad because they are legally unjustified. (If they were, there would be a trivial way to prevent all badness: abolish all laws!) Harms are bad because they are, well, harms; that is what “harm” means. Legally justified homicides are still tragedies.

Ideally we would identify and eliminate all harms. But we cannot. And thus (our axiom goes), the lawmaker weighs harms against each other, decides which are the most serious, writes laws to stop those harms, and lets some of the lesser harms through as necessary evils. The harms that it has banned are now, by definition, “legally unjustified.” The lesser harms it lets through are “legally justified.”

“Legally justified” and “legally unjustified” are thus conclusions about which harms our legal doctrines have ended up allowing, after our imperfect attempt to minimize them. These categories cannot be inputs to a legal rule. By treating them that way, the Leubsdorf-Posner test is circular.

A different way to state this problem is that the test makes a hash of two incompatible ways to pick a remedy. The first way is the merits approach—the way that courts usually make final decisions. They simply decide which party has the better view of the merits. Or, in Posner’s notation, they ask whether

p>(1-p).

This test makes sense if you think that the law encodes all the relevant social values. If so, and if a party has the better view of the law, then giving her a remedy minimizes the expected social harm, again by definition. A judge should use this approach if she agrees with NFIB that once the public has decided on the law “through democratic processes,” “[i]t is not [a court’s] role to weigh [the policy] tradeoffs.”^[45]

We can also posit a second, radically different way to pick a remedy. If a judge mistrusts the substantive law’s guidance, she can directly measure costs and benefits and decide for herself whether a remedy would be in the “public interest.” In Posner’s notation, she would issue the injunction when she thinks that

H_p>H_d

Two academics, Richard Brooks and Warren Schwartz, have proposed something like this. Until a case is over, they would award an entitlement to whoever would use it most efficiently—law be damned. For example, if we are in a contract dispute over who should own certain land, the court would temporarily give the land to whoever would get the most value from it, without even trying to interpret the contract.^[46]

The Leubsdorf-Posner test takes the merits test and the efficiency test and literally multiplies them together. But although the two tests answer the same question, they are on separate scales and use incompatible methodologies. They cannot be jumbled so casually.

Imagine if a scientist did something similar. To decide whether a man is obese, a doctor might ask whether his weight over his height squared (his BMI) is more than 30 kilograms/meter². Or she might ask whether his waist size is more than 40 inches.^[47] But she would not likely ask whether his waist size times his weight over his height squared exceeds 30×40 kilogram-inches/meter². The variables are all in the right direction. But because they are on different scales, the combined formula is meaningless. If the hodgepodge proved useful, that would be a nonobvious result calling for an explanation.

Leubsdorf and Posner do not have a good explanation for their own multiplication attempt. Their formula is superficially plausible because, in law and economics, it is common to multiply a welfare value by a probability. Take tort law’s Hand Rule. It asks whether the cost of a precaution is less than the benefit times the probability the precaution will be needed—that is, whether the precaution has positive expected value. If so, then not taking the precaution is negligent.^[48] The Hand Rule is coherent because under economic theory, the welfare of an uncertain event may be discounted by its likelihood, to get an expected utility.^[49] Leubsdorf and Posner, however, want to multiply the welfare of a (possibly certain) event by the chance it is found legal. Yet they offer no theory for why these variables may be so combined.

Might there be a better way to combine them? I have not found one. The problem is that most conceivable methods (for instance, averaging them) would first require us to get the variables on the same scale. But it is not clear how a court can translate legal probabilities into social utilities.

If a court tries to mix them, it may get a headache. Take the Third Circuit case Novartis Consumer Health, Inc. v. Johnson & Johnson-Merck Consumer Pharmaceuticals Co.^[50] Novartis sued J&J for false advertising under the Lanham Act; it claimed that J&J’s “Night Time Strength” medicine worked no better at night.^[51] Under the Lanham Act, Novartis needed to prove (1) that J&J used a deceptive label and (2) that Novartis was “likely to be damaged” as a result.^[52] Novartis sought a preliminary injunction ordering J&J to stop using the label.^[53]

The district court granted the injunction.^[54] A key part of its reasoning was that the label would likely harm Novartis by diverting sales from the rival.^[55] So far, so good. But under modern doctrine, the court had to consider this issue in a most convoluted way. First, under the “likelihood of success on the merits” prong, it had to consider, not whether Novartis would likely suffer harm, but—because the Lanham Act has a “likely” of its own—whether it was likely that Novartis would likely be harmed. And then it had to return to this issue under the “harm” prong: it had to separately measure how much Novartis would be harmed absent an injunction.^[56]

Because the two “harm” questions get at the same real-world issue—will J&J’s label divert sales from Novartis?—you might think that the court could analyze them in tandem. The district court seemed to think so: when it measured Novartis’s harm for purposes of the harm prong, it cited cases that interpreted “damage” under the Lanham Act.^[57]

But that was wrong. As the Third Circuit explained on appeal, the two “harm” questions “differ[ed].”^[58] So the court could not mix and match case law.^[59] To analyze harm for the “harm” prong, the district court should have stuck to cases interpreting the injunction standard.^[60] In other words, to measure the potential harm from a Lanham Act violation, the district court was wrong to consult the definition of “harm” under the Lanham Act.

That is bizarre. But it is also, under current doctrine, correct. The merits and harm prongs are parallel tracks that combine only at the end. While they ask the same question, they use different tools and are on different scales. So courts must analyze them separately. In Novartis, where “harm to one of the parties” was an explicit element of both the merits and the preliminary-injunction test, the folly of this approach seems especially clear. But every case has this problem. Even when “harm” is not a merits element, it is always in the background. The whole point of elements is to specify which acts the law deems harmful. So whenever a court marches through the elements, it is deciding whether the defendant did something harmful. When it also does a freelance harm analysis, it is double counting.

C. Flaw Two: Considering the Equities at All

The modern approach to preliminary remedies is not just internally flawed. It also degrades the substantive law and is incompatible with how courts approach final decisions.

I will start on less controversial ground: at the very least, the efficiency test (Brooks and Schwartz’s approach) is mistaken. Most judges and lawyers would agree with that. But current doctrine has the same flaw.

1. Why we should reject the efficiency test. Consider how it would apply in a simple property case. We are fighting over who owns a piece of farmland. You say that you inherited the land from your uncle, but I claim that your uncle sold it to me before he died. While we are equally good farmers, you value the land more because you are sentimentally attached to it.

Yet property law cares not one bit. How we value the land, economically or sentimentally, has nothing to do with who will get to occupy it permanently. When the court makes a final decision about who owns the land, it will ignore your feelings.

But say the final decision is two years away. The court still might need to make a preliminary decision about who will control the land in the interim. The issue might come up, for example, if I am occupying the land, but you move for a preliminary injunction to get me off.^[61]

If the court follows the efficiency approach, it will grant your motion. It will let you control the land for the next two years because you value it more.^[62] It will thus rule for you based on a factor that is irrelevant to the substantive law, while ignoring every factor that property law does care about. If we are cynical, we might say that the court has replaced the law with a policy preference (its sympathy for you). If we are more generous, we might say that the court is still doing law, just a different kind. After all, there is another body of law—tort law—that sometimes does tell judges to award an entitlement to who values it most.^[63] So another way to describe the court’s reasoning is to say that it replaced property law with a sort of tort law.

But land ownership is not supposed to be governed by tort law. We could have formed a society in which courts give land to whoever values it most. Yet for many good reasons, we have long rejected that rule.^[64] And we still reject it: after a final hearing, our case will be governed by property law. Imagine if the court ruled for me after the final hearing but stayed its order to give you two final years on your family’s land. The appeals court would swiftly reverse this power grab. Normally, throwing out the substantive law—even temporarily—is forbidden.

Since the efficiency approach tells courts to repeatedly throw out the substantive law during one stage of the litigation calendar, it should have a compelling explanation. But Brooks and Schwartz do not offer one. Their excuse for ignoring the merits is that until the end of the case, there is “legal uncertainty.”^[65] Thus, they say, the merits have not been “determined authoritatively,” so “[a]nalysis predicated upon one party or the other having the entitlement is simply not helpful.”^[66]

This excuse does not hold up. To start, just because guidance is uncertain, that does not make it “simply not helpful.” A prediction that a party will likely win on the merits says something about what the right outcome now should be, just like how a 65% forecast of rain says something about whether you should bring an umbrella. You would not ignore the forecast because the weather has not yet been “determined authoritatively.”

Worse still, the “uncertainty” excuse does not even distinguish preliminary from final hearings. We can discuss two kinds of uncertainty. First, a court may not know who should win on the merits. Second, it may not know who will win. These questions are related but different: if the factfinder is only 75% sure that I am right on the merits, but knows it will learn nothing more, then it can be certain I will win.

Can either type of uncertainty explain current doctrine? No. The first (uncertainty over who should win) is not even unique to preliminary hearings. Sometimes, a court will still be gravely unsure of the merits after a final hearing: the lawyers are bad, key evidence is missing, or there is a circuit split on a controlling issue. In other cases, conversely, the court will be sure of the merits at the preliminary stage. This will often be true in administrative cases like NFIB, which can be decided on a closed record.^[67]

Even when the record is open, if the stakes are high enough, a court can learn the facts quickly. In a recent high-profile immigration class action, the plaintiffs filed 664 pages in support of their preliminary-injunction motion; the government replied with 214.^[68] In another case, “the preliminary injunction hearing lasted 5 days; the judge heard 4 witnesses and admitted 89 exhibits.”^[69] Some courts even recognize “a strong right to a [preliminary-injunction] hearing when facts are disputed.”^[70] So in important matters, a court can learn more at the preliminary stage than it will ever learn in a normal case. No doubt the law and facts will be clearer on average after a full hearing. But this is not an inherent difference that can justify using different bodies of law for different stages of a case.

You might respond that, while there is uncertainty at both stages, the uncertainty at the preliminary stage is deeper. The parties may have not yet uncovered vital evidence, identified their key witnesses, or even settled on their theories. Perhaps. And when this is true, a court trying to guess the merits is in a tough spot. But that is not the merits’ fault. If the court is missing facts that could upend its view of the merits, those facts could just as easily change which outcome seems more efficient. Efficiency reasoning cannot make up for missing facts. Like its rival legal reasoning, efficiency reasoning tells the court how to act on the facts once it knows them. If the facts are a coin flip, then neither kind of reasoning works well.

But might efficiency reasoning turn on different kinds of facts, facts that tend to show up earlier in a case? Not under our premise that the substantive law encodes the relevant social values. On that view, if a fact seems relevant to the efficiencies but not at all to the merits, then we have misunderstood the efficiencies. For instance, it might seem efficient to give the land to whoever values it most. But this may be false, since doing so might undermine expectations—a point exposed by the refusal of property law to weigh that factor. If you instead reject our premise and think that the substantive law leaves gaps for judges to fill, then you might want to peek at the efficiencies. But you should also want to do that when the merits are certain. There still would be no reason to treat one stage of the case differently.

A second kind of uncertainty does distinguish the two stages. By the final decision, the merits might be as misty as ever, but there is no longer any doubt how the court will rule. Before the end of the case, by contrast, the court can change its mind. This is the uncertainty that counts under the Leubsdorf-Posner test. It is real and important. I will return to it in Part II. But this uncertainty cannot justify what I am considering now: using different bodies of substantive law for different stages of a case.

There are many reasons why we might prefer one body of law to another.^[71] In land-use cases, maybe property law is better than tort law because it encourages people to invest in their land. Or maybe the law should be more sensitive to how people value the land. But whatever the answer is, it is not plausible that the right body of law to govern land disputes would depend on whether a particular court has already issued a final decision in a particular case. For that to matter, you would need to believe that in the preliminary world, a court’s remedial decision will bring costs and benefits that the normal substantive law cannot adequately capture—and yet, after the final decision, something in the world will change, so that the substantive law will now account for all the relevant factors.

Yet this ad hoc premise is incompatible with how most courts understand their role. In final decisions, courts declare what the law has always meant, who has always deserved an entitlement.^[72] Final decisions are still practically important because they announce that conclusion to the world. But they do not change who deserved the entitlement the whole time. So the publishing of a decision cannot be a reason why the body of law governing an entitlement would switch. If property law is the right doctrine now, then it was the right doctrine before. To replace it with efficiency analysis, even for a few years, would disrespect the social values that we thought important enough to encode into property law. We should therefore reject the efficiency test.

2. Current doctrine has the same flaw. Few people believe in the efficiency test. You probably did not need much convincing to reject it. But the modern, universal approach to preliminary remedies has the same flaw.

The modern approach does not replace the substantive law with pure tort law. Instead it replaces the law with a substantive-tort hybrid. The modern rule has the court evaluate the merits, measure the efficiencies, and then, if those factors cut different ways, go with whichever it feels more strongly about. Yet just like how society did not pick tort law to govern our farmland dispute, neither did it pick a tort-property hybrid. So the hybrid approach also cleaves a disconnect into the substantive law. It too imposes a different legal regime at different times.

Again, most experts justify this result by citing uncertainty. (Posner: because courts are “act[ing] on an incomplete record”;^[73] Laycock: because at the preliminary stage “no wrongdoer has been finally identified.”^[74]). But again, this uncertainty does not distinguish preliminary hearings from final ones: if you thought that the modern test is justified by “uncertainty,” you would need to apply it after many final hearings.

Indeed, the modern test—the Leubsdorf-Posner rule or its four-pronged judicial cousins—is just the efficiency rule lite. The efficiency rule says to do utility balancing always. The modern approach says to do it sometimes, and to other times follow the merits. When it ends up agreeing with the merits, it adds nothing and wastes time. When it says to reject the merits, it is the efficiency rule, with all that rule’s problems. Because the modern test follows the merits sometimes, it respects the substantive law somewhat more often than the efficiency test does. But it is more arbitrary: as we saw in the last Part, it does not pick between the merits and efficiencies coherently.

Courts rarely admit that the other preliminary-injunction factors have led them to overrule their best guess of the merits. In my research for this essay, I found no clear examples of that. But just because courts do not explicitly override the substantive law, that does not mean they follow it. Instead, the common move is to declare the law “close” and then ignore it, breaking the tie with other factors. At that point, the court might as well be using the efficiency rule.

Take Comanche Nation v. United States, a real-life version of our land dispute.^[75] The Fort Sill Apache Tribe wanted to run a casino on tribal land.^[76] The federal government was poised to let it.^[77] But the Comanche Nation objected, claiming that it controlled gaming rights on the land.^[78] So it sued the government to block approval of the casino. It moved for a preliminary injunction.^[79] The main legal issue was the meaning of some federal regulations and a treaty.^[80]

The court granted the injunction.^[81] But all it said about the merits was that they “deserve[d] more deliberate consideration” later.^[82] The ruling turned instead on the balance of harms. The court weighed whether allowing gambling would be a good idea. On the one hand, an injunction would cause the Apaches “economic loss from the inability to conduct [their proposed] gaming activit[y].”^[83] But this harm, the court felt, was outweighed by the increased risk of “criminal activity” on the land and the cost to the Comanches of losing their “regulatory authority over the property.”^[84] Since the balance of harms favored the Comanches, the court ordered the government to hold up approving the casino.^[85]

Maybe the judge was right; maybe the casino was a bad idea. I do not know. But federal judges normally do not get to block a casino because they feel that one tribe’s economic benefits are outweighed by another’s loss of power, plus the risk of crime. Comanche Nation settled, but suppose it had reached a final judgment. It is impossible to imagine that the court, though concluding that the Comanches had no right to block the casino, would have ruled for them anyway because of its crime and governance concerns. It is equally hard to imagine the reverse—the court’s finding that the Comanches had the right to veto the gaming, but ignoring that right because the Apaches needed the money.

No: from the date of the final judgment onward, the casino’s fate would turn solely on a point of Indian law. Yet because the case was at one particular stage of the litigation schedule, control over the land turned on factors that were (for all we know) irrelevant to the merits. For a while, the court denied the Apaches development rights based on an assertion of sovereignty that may have been baseless.

3. What if the substantive law itself cares about efficiencies? The Comanche Nation case is extreme. A court’s public-interest balancing will not always be as disconnected from the merits as the pros and cons of gambling are from the meaning of a treaty. Sometimes, like in the Novartis case, the substantive legal standard will itself call for a sort of interest balancing. And even when it does not, courts can do a merits-sensitive interest balancing: they can (Leubsdorf suggests) “determine whose interests [to] . . . consider by reference to the substantive law.”^[86]

But mining the law for social values is hard, and courts are apt to get it wrong. Consider another Indian law case, Adams v. Vance.^[87] Alaskan Inupiats have traditionally hunted bowhead whales.^[88] But the International Whaling Commission decided to ban hunting them for a year.^[89] The United States was a member of the Commission. While it was not required to acquiesce to the Commission’s ruling, the Secretary of State decided to.^[90] Some Inupiats sued him, claiming that he had “violated the [federal government’s] trust obligation to the Eskimos.”^[91] They moved for a preliminary injunction ordering him to ignore the Commission.^[92] The district court enjoined him, but the D.C. Circuit reversed.^[93]

It reversed even though it thought that the Inupiats had “presented serious questions of law indicating a substantial case.”^[94] In fact, it agreed with them that “the United States owes trust obligations to native Americans.”^[95] But it did not try to figure out what those obligations were. Instead, like in Comanche Nation, the court skipped to the “harm” prong. It vacated the injunction because it thought that the injunction was bad policy. The injunction, it felt, would be an affront to our allies and “substantially endanger the [foreign policy] interests of the United States.”^[96] Because this harm outweighed the tribe’s interest in hunting the bowheads, an injunction was improper.^[97]

Like in Comanche Nation, this is policy weighing. But there is a difference. Here the policy issues seem more relevant to the merits. The “trust obligation” asserted by the Inupiats was an ill-defined common-law right.^[98] A safe bet is that a court, in fleshing out this right, would consider the government’s foreign-policy concerns. So even if the D.C. Circuit had stuck to the merits, it could have reasonably considered that issue.

But it could not have stopped there. If the court had been deciding the merits, it would have needed to then relate those foreign-policy concerns back to the doctrines that define the trust obligation. Those doctrines were crucial, because they were how the substantive law encoded the strength of the Inupiats’ interests. If the court had analyzed the merits, it would have needed to answer questions like: Does the trust obligation cover subsistence hunting? Does it have an exception for foreign relations? Those questions might be hard. But they would force the court to properly weigh all the competing social values.

Yet because the case was at a preliminary stage, the court skipped all that. Instead, it did only an intuitive interest balancing. And as a result, it probably missed some important values. For example, when it considered the Inupiats’ interests, it focused on whether they would run out of food. To the court’s relief, the Inupiats had “contingent reserves and [were promised] assistance from the United States” for the next year.^[99]

But this discussion frames the Inupiats’ interest in the narrowest possible light. The federal trust obligation is not just a way to make sure that they have food. It is also, arguably, a duty to honor the dignity of our country’s “domestic dependent nations.”^[100] And besides—as the opinion mentioned at the top, but by the end forgot—the “hunting of the bowhead has been the vital element of a millennia-old Eskimo culture.”^[101] So when the Secretary of State banned bowhead hunting, he may have injured the Inupiats in a way not reducible to their utility from hunting food. Or maybe, in the eyes of the law, he had not. That depends on the scope of the trust obligation. We do not know, because the D.C. Circuit did not try to find out. Under the preliminary-injunction test, it did not have to.

You might reply that the problem here is not with the test but with the application: the court missed an equity, the Inupiats’ cultural and dignity interests. But courts will always miss something. They will never be able to reverse-engineer every relevant social value encoded in the law. They should stop trying and simply apply the law.

The bowheads fared better than Yogi, the “575-pound Canadian black bear” who starred in the remedies classic Lakeshore Hills, Inc. v. Adcox.^[102] A homeowners association sued a resident for keeping Yogi on his property.^[103] The Illinois Appellate Court affirmed a preliminary injunction.^[104] According to the court, the association was right on the merits.^[105]

The court should have stopped there. But under Illinois law, it had to consider the public interest too. The injunction was in the public interest, the court explained, because living in a cage near humans is “contrary to the nature of bears.”^[106] While that does not seem relevant, surely the case law needs more musings on the nature of bears, not fewer. Let’s give Adcox a pass.

* * *

Perhaps some judges—Posner?—would love to replace all stages of a case with an open-ended efficiency analysis.^[107] But they could never get away with that. Under the modern preliminary-remedies doctrine, however, they get the next best thing.

Why do few bat an eye at this? There may be two reasons. First, the modern doctrine limits efficiency analysis to a stage when the merits are often uncertain. We have seen, however, that this is not a real difference.

Second, even in the preliminary phase, the modern rule does not tell courts to apply the efficiency test always. Instead, it toggles between efficiency and the merits. This makes the rule seem more like normal law; four-prong tests are the litigator’s bread and butter. But as cases like Comanche Nation show, the modern rule can easily become the efficiency test whenever a court wants it to. The court can simply breeze through the substantive law, declare the merits close, then use efficiency reasoning to break the tie.

But mixing merits and efficiencies is incoherent. They are different ways to answer the same question. They cannot be sensibly weighed. Instead, courts should commit to one or the other. If they want to decide a case using efficiency reasoning, they should admit upfront what they are doing. They should also admit that, under the same logic, they would be allowed to use efficiency reasoning in final decisions. But only a few judges would accept that.

Maybe those few judges are right. Maybe all decisions should be filtered through equitable balancing. But whenever courts would not consider the equities and public interest after a full hearing, they should not consider those factors at the preliminary stage either.^[108] The principle behind NFIB does not gain or lose force over time. Either courts may “weigh such tradeoffs” throughout a case—or they never should.^[109]

II. Preliminary Remedies Rethought

The merits are in. The public interest and equities are out. We have one more factor to consider: irreparable harm. It would be convenient if we could discard this factor too. Then there would be no special preliminary-remedy test; a court would simply decide the merits.

But that would often be absurd. Say a criminal defendant has been sentenced to death. Shortly before the execution, however, he comes forward with a novel and forceful argument for why the death penalty is unconstitutional. He asks for a stay. Still, the judge is not quite convinced; if she had to make a final decision now, she would narrowly let the execution go forward.

But she need not make a final decision now. She could grant the stay. Suppose she thinks that if she lets the defendant develop more evidence, there is a 60% chance that she will stick with her view but a 40% chance that she will find the death penalty unconstitutional. Should she grant the stay?

Obviously. And for the obvious reason: the execution is, in any sense, an “irreparable harm.” Under any plausible set of values, this harm outweighs the harm of delaying the execution.

This example shows that, at the preliminary phase, the merits do not provide enough guidance on their own. We need a second factor. And it needs to be something like “irreparable harm.” But my thesis is that courts should not weigh harms. So here is the challenge: Is there a way to keep the “irreparable” while axing the “harm”?

There is. Instead of asking about “irreparable harm,” a court should measure the more precise factor of option value—how much leeway the court has to switch remedies later.^[110] “Option value” and “irreparability” are related concepts. But they are not the same. In this Part, I will explain what option value is, how courts should measure it, and how this factor should merge with the merits into a new preliminary-remedy test. I will then show how this new test imposes more discipline on courts than does the modern rule. Most dramatically, in many cases—though not the death-penalty one—the test does indeed reduce to simply deciding the merits.

A. Courts Should Account for Option Value

There is another story we can tell about the death-penalty case—one that does not use the words “harm” or “benefit.” I will model the decision of a trial judge. While my model should mostly hold for appellate judges, they may need to consider other issues—that they act in panels, that they set precedents—outside the scope of this essay.

Assume the judge has no interest in weighing the pros and cons of capital punishment. Her only goal is to get the world to look as close as possible to the world intended by the substantive law. After a final hearing, she does that by making her best guess of the law and imposing a remedy accordingly.^[111] This is the right approach even if she is uncertain about the merits. Her uncertainty means that she might get the decision wrong and not conform the world to the law’s true intent. But that is a risk whatever she decides, so her best bet is to go with her best guess of the merits.

That was after a final hearing. If the judge has this same degree of uncertainty after the preliminary hearing, she is in better shape. Then she still has a chance to gather more information and refine her views. Yet her preliminary options do not give her equal use of that chance. If she lets the execution go forward, she is forever committed to not awarding a remedy. Her decision for the preliminary period (Period 1) also locks in the remedy for the longer permanent period (Period 2). If she stays the execution, however, she keeps open either option for Period 2. So staying the execution increases the probability that the Period 2 world will match what she ultimately finds the substantive law to dictate.

This strategy is not free. It requires her to impose a remedy in Period 1 that she believes is legally unwarranted. For perhaps years, the government will be thwarted from an act that (she believes) it most likely has the right to do. But the judge’s goal is not to minimize deviance from the substantive law during any one period. It is to minimize the overall deviance. So she must balance the two periods. And if she stays the execution, then under most plausible assumptions, the increase in expected conformance in Period 2 outweighs the decrease in Period 1.

To tell this story, we did not need to classify any effects as benefits or harms. We did not say anything like “death is extremely bad.” While that is surely true, it is redundant. The social value that “death is extremely bad” is already baked into the substantive law, which allows the death penalty in only the most extreme cases (and possibly never, if courts end up crediting the novel argument). If, in addition to considering the substantive law, the court also explicitly considered the badness of death, then it would double-count the issue.

Instead of saying that “death is bad,” our story relied on a structural asymmetry between the two preliminary choices. Not only is death bad, it destroyed the judge’s option value by ending the case. Her other choice preserved most of that value. This asymmetry is what justified overriding the merits. And unlike the equities or public-interest factors, the structural asymmetry is a genuine way to distinguish preliminary from final decisions. It matters only during the preliminary hearing: there is no value to keeping your options open once you have made your final decision.

My point here is not that a goal of preliminary remedies is preserving option value. That is no secret. I am making a stronger claim. When courts talk about irreparability, they tend to have in mind “two separate ideas.”^[112] One is “the idea of option value”; the other is that certain losses like death are hard to value or incommensurable with money damages.^[113] My point is that irreparability should matter only in the “option value” sense.

No doubt some harms are incommensurable. But “sometimes there is no avoiding the need to weigh seemingly incommensurable values.”^[114] We must resolve hard tradeoffs one way or the other, if only by not acting. Once we do, our answer is encoded into the substantive law or lack of one. Courts should not override these answers using a procedural vehicle. That is why, as we saw in the last Part, courts should not routinely factor the equities and public interest into their decisions. But if courts could still launder in their policy preferences through the “irreparable harm” prong, we would be back to square one. Courts should thus consider “irreparable harm” only in the sense of “loss in option value”—that is, “irreversible deviation from the world intended by the substantive law.”

B. How to Measure Option Value (In Theory)

But “consider option value” is a vague command. What specifically should courts do? Here is one possible answer (that I will reject): If any irreversible deviation from the substantive law is an irreparable harm, then perhaps current doctrine is missing harms. So maybe courts should start calling more effects “harms.”

Two academic proposals do this; they find categories of harms that are overlooked by current doctrine. First, Douglas Lichtman claims that judges tend to code certain effects as “benefits” to one party rather than “harms” to another, and then disregard those effects.^[115] But as he explains, if these benefits were not intended by the law, then they are really harms (because, for example, they skew incentives).^[116] Thus, Lichtman argues, the preliminary-injunction test needs extra prongs for “irreparable benefits” to the plaintiff and defendant.^[117]

Ofer Grosskopf and Barak Medina show that the concept of “irreparable harm” must be expanded in another way. Under the traditional view, harm is “irreparable” only if it cannot be compensated later (for example, with damages).^[118] But this, they note, misses a crucial point. Unless a harm was purely distributional, compensation does not undo the harm; it just changes who bears the costs. The global welfare has fallen just as much. Yet current doctrine ignores compensated harms. Instead, they argue, courts should treat all “harms that represent deadweight-loss as irreparable” and “take into account their entire amount, regardless of the availability of an ex-post remedy.”^[119]

These proposals are improvements on their own terms. But they still ask courts to label effects as benefits or harms, while adding more epicycles. They are moving in the right direction but on the wrong track.

We should back up and try a different one. Recall why we need to look beyond the merits at all. It is because of the structural asymmetry: that certain remedial options become irreversible earlier than others. Since that is what ultimately matters, what if courts forgot about benefits and harms, and simply measured the asymmetry directly?

Here is how they can. Suppose a court is considering two remedies, A and B. Option B could be not issuing a remedy. Over time, these remedies will create two different worlds. But those differences will not be set in stone immediately. Instead, they will accrue over time. (An effect “accrues” when it can no longer be reversed, not necessarily when it is experienced. If someone dies, his loved ones might grieve for decades, but that grief is now bound to happen. So the “grief” effect accrues at the moment of death.) The court should study the accrual profile of each remedy. To do this, it should estimate what percentage of the remedy’s effects will accrue between now and when the final decision will be issued. As we will see, besides assessing the merits, this is all that the court needs to do.

To see how this step works, we can return to the execution example. I will posit stylized facts to show what courts should be aiming at in theory. In the next section, I will discuss what we can realistically expect of them. Suppose that at time zero, the court is deciding whether to stay the execution. In a year, the court will make a permanent decision. If it stays the execution, there is a 60% chance that it will switch remedies later. Assume that there will be no appeal and that we care only about what happens over the next ten years.

We can now imagine two worlds, A and B. These worlds are extremes. In World A, the execution never happens: the court issues the stay at time zero, decides in a year that the execution would indeed be unconstitutional, and so permanently vacates the sentence. In World B, by contrast, the execution happens immediately. At the end of a decade, Worlds A and B will be different in many ways. Imagine that there are three:

1. In World A only, the government will have expended $200,000 a year to imprison the defendant.

2. In World A only, there is a .01% chance each year that the defendant will escape prison forever.

3. And in World A only, the defendant will not be executed. You might think that this is good or bad, depending on whether you think that capital punishment is just or barbaric.

But whatever the judge decides at time zero, the world does not necessarily turn into World A or World B immediately. Those Worlds correspond to what would happen if her initial decision is left undisturbed for the full decade.

Suppose, for instance, that the judge stays the execution. This puts the world on pace to look like World A. But only slowly. Take difference one, the money. In World A, the government spends $2 million more than it spends in World B. Yet during the first year after the stay, it spends only $200,000 more than it would have spent in the first year of World B. If, at the one-year mark, the judge switches remedies and allows the execution, the government will spend only that $200,000. And in the end, the world will end up looking like a blend of the two Worlds, but weighted 90% toward World B. In other words, after 10% of the time, only 10% of the financial effect accrued, because this effect accrued continuously, as it was experienced.

The next difference—the risk of escape—also accrues continuously, though for a different reason. This effect is probabilistic. If the court stays the execution for one year, there is a .01% chance that the defendant will escape. If the execution is permanently canceled, the chance that he will escape during the decade is about .1%: ten times greater.^[120] Thus, 10% of the risk accrues during the first 10% of the period. This effect is different than the financial effect, because in any realized world, it will not accrue continuously (the defendant either will escape one day or never will). But from the court’s perspective at time zero, that does not matter. The key is that, on expectation, the escape risk will fall by 90% if the court later ends the stay. So again, even if the court originally picks World A, the escape risk can (on expectation) later shift 90% of the way toward World B’s.

By contrast, the third difference—the execution itself—does not accrue at all if the judge stays the execution. Whatever justice or cruelty the death penalty causes, it will cause the same in a year.

After the court considers all these factors, it should estimate the overall percentage of the effects that will accrue before the final decision. In our idealized model, the court might estimate that overall, 5% of the effects of staying the execution will accrue in the first year.

Next, the court should consider its other option, allowing the execution. This remedial choice has nearly the opposite profile: all the effects accrue immediately. Once the execution happens, the level of retribution, the amount of money spent jailing the defendant, and the escape risk will become permanently fixed. Even if the court issued an advisory opinion in a year announcing that the execution was a mistake, those effects would not change. So if the court denies the stay, 100% of the effects of that decision will accrue before it can reverse course.

In sum, preliminary remedy A (the stay) will be 5% set in stone, preliminary remedy B (no stay) will be 100% set in stone, and, as stipulated before, if the court picks A it is 60% likely to switch remedies. And that is it—the court now has all it needs to decide the stay motion. To get its answer, the court should compare its options on the merits, discounting them by option value.

Judges can do this intuitively. They do not need to use a formula, and you do not need one to understand this essay. Still, writing out a proposed legal rule as a formula can be a good way to discipline your thinking. It forces you to articulate precisely how each factor of a test matters. And some readers may find it helpful. The rest of this section is for them.

Imagine (stepping away from the death-penalty example) that a court picks remedy A over remedy B. Is the expected benefit of this decision positive or negative? Suppose first that remedy A turns out to be the right remedy after the full hearing. So the remedy stays in place forever. This is the World A that I defined before. Here World A is the best possible world. Certainly, it is much better than World B, in which the wrong remedy B is in place forever. Suppose that World A is better than World B by 1 unit of social value. The 1 is arbitrary; as we will see, it cancels out.^[121]

It does not follow, however, that preliminary remedy A is better than preliminary remedy B by 1 full unit of social value. After all, even if the court picks preliminary remedy B, the world is not locked into World B. The court can partially change course after the full hearing. For instance, if only 10% of B’s effects accrue before the full hearing, then the world can still end up looking 90% like World A. And thus, preliminary remedy A does not get full credit for saving us from World B. It gets only 10% credit—the portion of World B that otherwise would have been locked in. More generally, when A turns out to be the right remedy for the long term, the advantage of preliminary remedy A over preliminary remedy B is 1*Ω_B, where Ω_B is the percentage of B’s effects that would have accrued before the court issued a final decision.

Next, imagine that preliminary remedy A turns out to be wrong after a full hearing. In this scenario, the ideal world would have been World B, in which remedy B was always in place. And the worst possible outcome would be World A. We can suppose that World A is worse than World B by 1 unit of social value. Again, the 1 is arbitrary. The key point is that it is the same 1 from two paragraphs ago. In other words, we are assuming that all deviation from the substantive law is equally bad: being in World B when the law prescribed World A is just as bad as being in World A when the law prescribed World B. This is exactly what courts must assume if their only goal is to hunt for deviations from the substantive law, rather than weigh costs and benefits for themselves.

But once again, just because World A is worse than World B by 1 unit of social value, it does not necessarily mean that preliminary remedy A is worse than preliminary remedy B by the full unit. Instead, it is worse only to the extent that its effects are locked in. So in this scenario, the harm of picking A over B is 1*Ω_A, where Ω_A is the percentage of A’s effects that accrue before the court can issue a final decision.

Finally, let p be the probability that the court will pick remedy A after a full hearing.

We are done. As we just saw, if a court picks preliminary remedy A, then with probability p its decision will work out and result in a net benefit of 1*Ω_B. On the other hand, with probability (p), the decision will not work out and will cause a net harm of 1*Ω_A. The court should pick preliminary remedy A when the expected benefit outweighs the expected harm: when

p* Ω_B>(1-p)Ω_A.

As promised, the 1s canceled out. In our death-penalty example, we had p = 40%, Ω_A=5%, and Ω_B=100%. The left-hand side is much greater (and, in a capital case, will be greater under most plausible assumptions). So the stay is proper.

* * *

My test and Judge Posner’s take the same functional form. But mine has an appealing feature that his lacks. Remember the final­-judgment “formula”: p>(1-p). That formula is a limiting case of mine. We can recover it by noticing that after a final decision, Ω_A=Ω_B=1. There is no option value because the remedy will not be revisited either way.

Conversely, my test is the natural generalization of how courts make final decisions. It simply adds in the one variable—the structural asymmetry—that truly distinguishes preliminary and final decisions. Judge Posner’s formula does not have this feature. To derive the final-judgment formula from his, we would need to assume, implausibly, that after the final decision, the harm of ruling for the defendant is exactly the harm of ruling for the plaintiff.

C. How to Measure Option Value (In Practice)

Law is much easier when you can assume that all the inputs are round numbers. But the real world is messy. In the death-penalty example, I ignored many complications. I ignored important effects like deterrence. I assumed that the effects would linger for only ten years; a real judge would need to guess that.

And I elided the hardest part of the analysis. For option A, I stipulated that 10% of the financial and escape effects, but 0% of the retributive effect, accrued. I then stipulated that overall, 5% of A’s effects accrued. That averaging was arbitrary. In the real world, estimating this percentage will be a hard call. The court cannot take a simple average; it must decide how important each effect is. And this, alas, calls for a value judgment.

The good news is that the court still does not need to measure harms and benefits. It thus can avoid taking a directional view. For example, the court need not say whether the death penalty is moral or immoral; it just needs to realize that either way, the moral implications are weightier than the financial ones. But how much weightier? There is no mechanical answer.

Given how messy the real world is, does switching to an option-value approach accomplish anything? It does. First, in a significant class of cases (though not in the death-penalty example), all the remedial options that the court is considering will have the same accrual profile. And then the optionality effects cancel out. In these cases, courts can decide preliminary remedies exactly by deciding the merits.

In many cases, for example, both remedial options will have effects that all accrue continuously, at least approximately. We have already seen two examples of continuous effects: effects that accrue as they are experienced and uniform probabilistic effects. Many other sorts of emotional, dignitary, and constitutional harms will work like this. For instance, it will often be reasonable to assume that muzzling someone’s speech for one year is one-tenth as bad as muzzling it for ten years.

As long as the court’s remedial choices have effects that all accrue continuously, then optionality washes out. Take Adcox, the case about Yogi the declawed bear. There, all the significant effects were likely continuous. If the court enjoined Yogi, the main effect would be his needing to move somewhere else. But this was likely reversible; if, after 10% of the relevant timeframe, the judge reversed course, Yogi could still enjoy 90% of the benefits of living at his old home. (And Yogi’s owner could enjoy 90% of his company, and so on.) By contrast, if the court had not enjoined Yogi, the main risks were that the bear would scare, annoy, or hurt his neighbors. But these effects were roughly uniform, so they would have accrued continuously too; if Yogi were forced to move after 10% of the relevant period, then the neighbors would spend 90% less time in fear, and the expected risk of an accident during the whole period would fall by 90%.

Since the optionality effects in Adcox canceled out, the court should have simply considered the merits. True, if the court had gotten the merits wrong and took 10% of the relevant timeframe to fix its mistake, then 10% of the bad effects of that decision would have been locked in. But that risk was symmetric. So the judge’s best bet was to pick the remedial option that was most likely right—to just decide the merits.

Adcox is one paradigm of when optionality cancels out. The other is when all the relevant effects accrue immediately. Think of a First Amendment case. Activists are planning to protest a visiting leader next week. The court must decide whether to enjoin the police from breaking up the protest. So the preliminary hearing is it. By the time a full hearing happens, the subject of the dispute will be over. Since the court gets only one chance to decide the case, there is no value to keeping options open. The court should just decide the merits. (Perhaps the protestors could still get damages. But remember Grosskopf and Medina’s point: those damages would not stop the First Amendment harm from happening, they would just change who bears the cost.)

Cases like these are where my proposal has the most bite.^[122] In others, like death-penalty stays, the court’s options will have differing accrual profiles. And then my test will not simplify nicely. The court will need to make hard judgment calls. Its reasoning will be less elegant.

But that is just a limit of law. Law is not math; every preliminary-remedy test—the modern rule, the efficiency rule, and mine—will have hard calls. Given that, we should pick the test that is analytically correct, the one that nudges courts toward the right reasoning. If optionality and accrual profiles are indeed the right way to think about preliminary remedies, then courts should adopt my test for the same reasons that they should adopt any other correct legal doctrine.

At least my test is more determinate than current doctrine. In any close case, whether a remedy causes net benefit or net harm will be purely in the beholder’s eye. Should there be a casino? Should tribal rights trump the whales? Where should Yogi live? Good luck getting a consensus on any of these. By contrast, even the most ardent supporter of the death penalty could not say with a straight face that a quick execution leaves the court with more option value than does a stay.

Conclusion

This essay is about how courts should decide cases when they will have a chance to reconsider. They should start with what would be decisive after a full hearing, the merits. And then they should fold in option value, the variable that truly distinguishes preliminary from final hearings.

Though my argument has focused on injunctions and stays, it applies just as well to other preliminary terrains. Consider the pretrial detention of criminal defendants. Before the trial, a court may detain a defendant if it finds that he is dangerous.^[123] But to keep him imprisoned after the trial, the government of course must prove beyond a reasonable doubt that he committed a crime.^[124]

The Supreme Court has held that the lower pretrial standard satisfies due process. In United States v. Salerno, it reasoned that “[t]he government’s interest in preventing crime by arrestees is both legitimate and compelling.”^[125] But as the dissent pointed out, this explanation is puzzling. The government also has a compelling interest in preventing crime by acquitted defendants. Yet few think that the government could “continue to hold [an acquitted] defendant in detention based upon its showing that he is dangerous.”^[126] Under Salerno, then, “an untried indictment somehow acts to permit a detention . . . which after an acquittal would be unconstitutional.”^[127]

This essay can help us think through that debate. The dissent’s argument is a variant of my point in Part I. At the preliminary-detention hearing, the judge makes the same remedial decision (will or won’t the defendant be detained) that she and the jury will make after a full trial. But during the preliminary hearing, the judge uses a different substantive standard, dangerousness rather than guilt. It is hard to see why the substantive standard for the full trial—one from the Constitution, no less—should be suspended for one stage of the case. While confining dangerous people may be a vital policy goal, so is safeguarding liberty, and the Constitution has decided how to balance these aims. Why should Congress or the courts get to reweigh them?

The majority gave no answer.^[128] But this essay offers a partial one. Preliminary-detention hearings are different than full trials, not because the Constitution has less force during them, but because they feature structural asymmetries like the ones we explored in Part II. This cuts both ways. If a defendant is detained, but this proves to be the wrong remedy (he is acquitted), his time in prison cannot be undone. Yet if he is let out on bail and commits a violent crime, that harm also accrues forever. Because the due-process standard for full trials does not account for these points, the same constitutional principle might yield a different rule for the pretrial stage.

Salerno should have pointed this out and then asked: Is the “dangerousness” standard what we get if we take the trial standard and adjust it for optionality effects? The answer is probably no, not least because the standard gives no weight to guilt beyond the minimal probable cause the government needs to indict anyone.^[129] But I will leave the correct standard for future research.

The merits-plus-option-value framework may be useful in other preliminary domains too, from immigration detention to bankruptcy stays. In some areas (like bail reform), the current standard is set by statute. But courts set the injunction standard themselves.^[130] When they do so, NFIB reminds them to respect the substantive law even at the preliminary stage. This principle cannot be cabined to that case’s facts. If courts take it seriously, they should entirely rethink preliminary remedies.

1. * I am grateful to Blake Himebaugh and the staff of the Washington University Law Review for their thorough editing; to Allison Douglis, Emmett Gilles, Jordan Goldberg, Louis Katz, Ben Kornfeld, Chase Robinett, and Garrett West for crucial insights; to Judge Stephanos Bibas, for alerting me to important counterarguments and for many other things too; and to Daryl Levinson, for making me rethink remedies, preliminary and otherwise. ↑

2. . 142 S. Ct. 661, 663 (2022) (per curiam). ↑

3. . Id. at 664. ↑

4. . Nken v. Holder, 556 U.S. 418, 434 (2009) (quoting Hilton v. Braunskill, 481 U.S. 770, 776 (1987)). The last two factors “merge when the Government is the opposing party.” Id. at 435. So (in the NFIB dissent, for instance), they are sometimes considered together. See 142 S. Ct. at 671 (Breyer, Sotomayor, and Kagan, JJ., dissenting). ↑

5. . NFIB, 142 S. Ct. at 666 (majority opinion) (emphasis added) (citations omitted). ↑

6. . Id. ↑

7. . Richard Re, Did the Supreme Court Overrule Equity?, Re’s Judicata (Jan. 14, 2022, 6:01 AM), https://richardresjudicata.wordpress.com/2022/01/14/did-the-supreme-court-overrule-equity/ [https://perma.cc/553Z-H9NF]. ↑

8. . His answer was no. Will Baude, Balancing the Equities in the Vaccine Mandate Case, Volokh Conspiracy (Jan. 14, 2022, 11:24 AM), https://reason.com/volokh/2022/01/14/balancing-the-equities-in-the-vaccine-mandate-case/ [https://perma.cc/RUZ4-4R34]. ↑

9. . Am. Bankers Ass’n v. Nat’l Credit Union Admin., 934 F.3d 649, 673 (D.C. Cir. 2019). ↑

10. . NFIB, 142 S. Ct. at 664–66. ↑

11. . See Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins., 463 U.S. 29, 46 (1983). ↑

12. . For a similar definition, see Kenneth J. Arrow & Anthony C. Fisher, Environmental Preservation, Uncertainty, and Irreversibility, 88 Q.J. Econ. 312, 313 (1974). ↑

13. . Richard R.W. Brooks & Warren F. Schwartz, Legal Uncertainty, Economic Efficiency, and the Preliminary Injunction Doctrine, 58 Stan. L. Rev. 381, 406 (2005). ↑

14. . United States v. Oakland Cannabis Buyers’ Coop., 532 U.S. 483, 497–98 (2001) (citation and quotation marks omitted). ↑

15. . See 1 William Blackstone, Commentaries *91 (“[I]f the parliament will positively enact a thing to be done which is unreasonable, I know of no [judicial] power that can control it.”). ↑

16. . See generally Douglas Laycock, The Death of the Irreparable Injury Rule 5 (1991). ↑

17. . 547 U.S. 388 (2006). ↑

18. . See Douglas Laycock & Richard L. Hasen, Modern American Remedies 441 (5th ed. 2018). ↑

19. . eBay, 547 U.S. at 391. ↑

20. . Laycock & Hasen, supra note 17, at 441, 445. ↑

21. . Id. at 446 (quoting eBay, 547 U.S. at 394, which described its test as flowing from “traditional principles of equity”). ↑

22. . Id. at 448. ↑

23. . Id. at 446. ↑

24. . See id. at 456. ↑

25. . Bostock v. Clayton County, 140 S. Ct. 1731, 1824 (2020) (Kavanaugh, J., dissenting). ↑

26. . Obergefell v. Hodges, 576 U.S. 644, 714 (2015) (Scalia, J., dissenting). ↑

27. . See, e.g., Jerome Frank, What Courts Do in Fact (pt. 2), 26 Ill. L. Rev. 761 (1932). ↑

28. . See, e.g., Guzman Chavez v. Hott, 940 F.3d 867, 875, 882 (4th Cir. 2019), rev’d sub nom. Johnson v. Guzman Chavez, 141 S. Ct. 2271 (2021). ↑

29. . 306 F. Supp. 3d 44 (D.D.C. 2018), rev’d, 934 F.3d 649 (D.C. Cir. 2019). ↑

30. . Id. at 63–64. ↑

31. . Id. at 65. ↑

32. . See Douglas Lichtman, Irreparable Benefits, 116 Yale L.J. 1284, 1300–01 (2007). But see B. Braun Melsungen AG v. Terumo Med. Corp., 778 F. Supp. 2d 506, 525 (D. Del. 2011) (declining to award a permanent injunction in part because the case presented “close calls”). ↑

33. . Nken v. Holder, 556 U.S. 418, 434 (2009). ↑

34. . Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20 (2008). ↑

35. . For the rundown, see Kirstin Stoll-DeBell, Nancy L. Dempsey & Bradford E. Dempsey, Injunctive Relief 50–72 (2009). ↑

36. . See M. Devon Moore, Note, The Preliminary Injunction Standard: Understanding the Public Interest Factor, 117 Mich. L. Rev. 939, 950–52 (2019); see also supra note 3. ↑

37. . John Leubsdorf, The Standard for Preliminary Injunctions, 91 Harv. L. Rev. 525, 525–26 (1978). ↑

38. . Laycock, supra note 15, at 122; see Leubsdorf, supra note 36, at 541–42. ↑

39. . Am. Hosp. Supply Corp. v. Hosp. Prods., Ltd., 780 F.2d 589, 593 (7th Cir. 1986). ↑

40. . Thomas R. Lee, Preliminary Injunctions and the Status Quo, 58 Wash. & Lee L. Rev. 109, 154, 157 (2001). ↑

41. . Joshua P. Davis, Taking Uncertainty Seriously: Revising Injunction Doctrine, 34 Rutgers L.J. 363, 368 n.20 (2003) (collecting cases). ↑

42. . See Douglas Laycock, Modern American Remedies 446 (4th ed. 2010) (citing Packaging Indus. Grp., Inc. v. Cheney, 405 N.E.2d 106, 111–12 (Mass. 1980)). ↑

43. . Lee, supra note 39, at 157. Some courts also ask whether an injunction would alter the status quo, but “[t]hey almost [n]ever explain how this standard relates to the four-part test.” Laycock, supra note 41, at 448. I agree with the critics who find this standard “unsound,” and I will not discuss this side issue any further. Lee, supra note 39, at 110. ↑

44. . Leubsdorf, supra note 36, at 541 (“Not even all irreparable harm, but only irreparable harm to legal rights, should count.”). ↑

45. . Nat’l Fed’n of Indep. Bus. v. Dep’t of Lab., OSHA, 142 S. Ct. 661, 666 (2022). ↑

46. . See Brooks & Schwartz, supra note 12, at 403. Actually, they want the law to force the parties to honestly reveal whether the harms outweigh the benefits, rather than having the court decide. Id. at 393. But for us that detail is not important. ↑

47. . Obesity Definition, Harv. T.H. Chan Sch. Pub. Health: Obesity Prevention Source, https://www.hsph.harvard.edu/obesity-prevention-source/obesity-definition/ [https://perma.cc/S49F-CRF3]. ↑

48. . United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir. 1947). ↑

49. . See generally R.A. Briggs, Normative Theories of Rational Choice: Expected Utility, Stan. Encyc. Phil. (Aug. 15, 2019), https://plato.stanford.edu/entries/rationality-normative-utility/ [https://perma.cc/4AS7-KDBD]. ↑

50. . 290 F.3d 578 (3d Cir. 2002). ↑

51. . Id. at 583, 585. ↑

52. . Id. at 586 (quoting 15 U.S.C. § 1125(a)). ↑

53. . Id. at 585. ↑

54. . Id. at 586. ↑

55. . Id. at 596. ↑

56. . Id. at 595. ↑

57. . Novartis Consumer Health, Inc. v. Johnson & Johnson-Merck Consumer Pharms. Co., 129 F. Supp. 2d 351, 367 (D.N.J. 2000). ↑

58. . 290 F.3d at 595. ↑

59. . Id. ↑

60. . Id. ↑

61. . Assume for now that we are both judgment-proof: if the court gets it wrong, the losing party cannot recover damages later. Later, we will see why this assumption is unimportant. See infra text around note 118. ↑

62. . Brooks & Schwartz, supra note 12, at 403. ↑

63. . See United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir. 1947). ↑

64. . See Guido Calabresi & A. Douglas Melamed, Property Rules, Liability Rules, and Inalienability: One View of the Cathedral, 85 Harv. L. Rev. 1089, 1092 (1972). ↑

65. . Brooks & Schwartz, supra note 12, at 406. ↑

66. . Id. ↑

67. . See Fla. Power & Light Co. v. Lorion, 470 U.S. 729, 743–44 (1985). ↑

68. . See Redacted Motion for Preliminary Injunction by Al Otro Lado, Inc., Al Otro Lado, Inc. v. McAleenan, 423 F. Supp. 3d 848 (S.D. Cal. 2019) (No. 3:17-cv-02366); Redacted Response in Opposition re Redacted Motion for Preliminary Injunction, Al Otro Lado, Inc. v. McAleenan, 423 F. Supp. 3d 848 (S.D. Cal. 2019) (No. 3:17-cv-02366). ↑

69. . Laycock, supra note 41, at 466 (discussing Lektro-Vend Corp. v. Vendo Co., 660 F.2d 255 (7th Cir. 1981)). ↑

70. . Id. at 467. ↑

71. . See generally Calabresi & Melamed, supra note 63. ↑

72. . See Rivers v. Roadway Express, Inc., 511 U.S. 298, 313 n.12 (1994). ↑

73. . Am. Hosp. Supply Corp. v. Hosp. Prods., Ltd., 780 F.2d 589, 593 (7th Cir. 1986). ↑

74. . Laycock, supra note 41, at 447. ↑

75. . 393 F. Supp. 2d 1196 (W.D. Okla. 2005). ↑

76. . Id. at 1201. ↑

77. . Id. at 1202. ↑

78. . Id. ↑

79. . Id. ↑

80. . Id. at 1212. ↑

81. . Id. ↑

82. . Id. ↑

83. . Id. at 1211. ↑

84. . Id. at 1210–11. ↑

85. . Id. at 1212. ↑

86. . Leubsdorf, supra note 36, at 549. ↑

87. . 570 F.2d 950 (D.C. Cir. 1978). ↑

88. . Id. at 952. ↑

89. . Id. at 952, 957. ↑

90. . Id. at 952. ↑

91. . Id. at 953. ↑

92. . Id. It was labeled a temporary restraining order, but the court treated it like an injunction. ↑

93. . Id. at 954. ↑

94. . Id. at 955. ↑

95. . Id. ↑

96. . Id. at 957. ↑

97. . Id. ↑

98. . See Reid Peyton Chambers, Judicial Enforcement of the Federal Trust Responsibility to Indians, 27 Stan. L. Rev. 1213, 1233–34 (1975). ↑

99. . 570 F.2d at 957. ↑

100. . Cherokee Nation v. Georgia, 30 U.S. (5 Pet.) 1, 17 (1831); see Chambers, supra note 97, at 1215–16. ↑

101. . 570 F.2d at 952. ↑

102. . 413 N.E.2d 548, 549 (Ill. App. Ct. 1980). ↑

103. . Id. ↑

104. . Id. at 550. ↑

105. . Id. ↑

106. . Id. ↑

107. . Adam Liptak, An Exit Interview with Richard Posner, Judicial Provocateur, N.Y. Times (Sept. 11, 2017), https://www.nytimes.com/2017/09/11/us/politics/judge-richard-posner-retirement
     .html [https://perma.cc/G9Q7-TX8H] (“‘I pay very little attention to legal rules, statutes, constitutional provisions,’ Judge Posner said. ‘A case is just a dispute. The first thing you do is ask yourself—forget about the law—what is a sensible resolution of this dispute?’”). ↑

108. . Cf. Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 33 (2008) (denying a preliminary injunction for reasons of public policy, but also stating that an injunction would not be proper even after a final hearing). ↑

109. . Nat’l Fed’n of Indep. Bus. v. Dep’t of Lab., OSHA, 142 S. Ct. 661, 666 (2022). ↑

110. . See Arrow & Fisher, supra note 11. ↑

111. . I am ignoring the inevitable appeal. The judge will need to decide whether to stay her final decision pending appeal. She should use the same principles I am discussing here. ↑

112. . Cass R. Sunstein, Irreparability as Irreversibility, 2017 Sup. Ct. Rev. 93, 95. ↑

113. . Id. at 95, 110–11. ↑

114. . Nat’l Pork Producers Council v. Ross, 143 S. Ct. 1142, 1168 (2023) (Roberts, C.J., concurring in part and dissenting in part). ↑

115. . Lichtman, supra note 31, at 1300. ↑

116. . Id. at 1290. ↑

117. . Id. at 1289–90. ↑

118. . Leubsdorf, supra note 36, at 551. ↑

119. . Ofer Grosskopf & Barak Medina, Repairing (the Doctrine of) Irreparable Harm 1–4 (Feb. 10, 2008) (unpublished manuscript), https://ssrn.com/abstract=1091730 [https://perma.cc/YA2S-Y6UY%5D. ↑

120. . Not quite because .01% × 10 = .1%.The probability of his escaping in later years is less than .01% per year, because there is a chance he has already escaped. The proper calculation is 1 – (1 – .01%)¹⁰ which is extremely close but not exactly equal to .1%. But for small probabilities that are uniform over time, the multiplication approximation is excellent, and the effect will accrue roughly continuously. And when an event can happen more than once, multiplication is exactly right. ↑

121. . Here I assume that the final decision is correct. Although it might not be, we can ignore this complication. Since the judge’s decision is final, then (ignoring appeals) she will never know if she got it wrong. The best she can ever say is that her decision is most likely right and will have positive value on expectation. The 1 here represents this positive expected value. ↑

122. . But these cases may raise another concern. Some psychologists have found that “a decision maker who must revisit an earlier decision” may feel “locked in to that earlier decision,” even when the facts change. Kevin J. Lynch, The Lock-In Effect of Preliminary Injunctions, 66 Fla. L. Rev. 779, 779 (2014). Lynch fears that judges will display this cognitive bias when they consider the merits a second time. Id. His antidote is a preliminary-injunction test that balances the merits against other injunction factors, so that courts will consider the pure merits question only once. Id. at 810. My test will normally fit that bill—but not when option value washes out.

     It is not clear whether even then, lock-in would be a real concern. There have been no studies of whether judges are psychologically locked into their preliminary decisions; the main study discussed by Lynch involved undergraduates pretending to make investments. Id. at 785–86, 811. And the one study “that comes close to addressing [judicial] lock-in” hints that there is no effect. Id. at 811–12.

     Still, I hope that someone studies the issue empirically. If there is a large lock-in effect, that may call for a reform. Not necessarily to the injunction test: if my merits approach is analytically right, it would be odd to purposely cloud judges’ reasoning with irrelevant factors, in the hope that they will make better decisions later. The better answer might be a procedural reform, like assigning a new judge for the next stage of the case. Already this can happen in appeals courts. See, e.g., E. Bay Sanctuary Covenant v. Biden, 993 F.3d 640, 660 (9th Cir. 2021). But I leave this issue for future work. ↑

123. . See, e.g., 18 U.S.C. § 3142(f). ↑

124. . See In re Winship, 397 U.S. 358, 361 (1970); Kansas v. Hendricks, 521 U.S. 346, 358 (1997) (approving civil commitment of dangerous people, but only if there is “some additional factor, such as a ‘mental illness’ or ‘mental abnormality’”). ↑

125. . 481 U.S. 739, 749 (1987). ↑

126. . Id. at 763 (Marshall, J., dissenting). ↑

127. . Id. at 764. ↑

128. . Id. at 748–49 (majority opinion) (discussing other laws that allowed confinement without a conviction, but citing none that used the “dangerousness” standard). ↑

129. . Id. at 750. ↑

130. . E.g., eBay Inc. v. MercExchange, L.L.C., 547 U.S. 388 (2006). ↑