Simulating DOGE

Everything you ever wanted to know about impoundment but were afraid to ask

The Department of Government Efficiency wants to cut spending. That will not be easy. No matter which party is in charge, Congress loves spending taxpayer money.

That’s not quite right. Congress doesn’t actually spend money. Instead, it enacts laws appropriating money for particular purposes. It’s the Executive Branch, controlled by the President, that spends the money.

Aha! A solution presents itself. Can’t the President just not spend the money?

This idea has come up enough times in American history to have earned its own name: “impoundment.” To make a long story short, Presidents have impounded funds from time to time throughout American history. But then President Nixon started impounding a lot more than prior Presidents. Congress responded by enacting a law in 1974 called the Impoundment Control Act, which bars the President from impounding under most circumstances.

So the President can’t impound. Unless the Impoundment Control Act is unconstitutional!

Most people do not think the Impoundment Control Act is unconstitutional. But one person who does, Mark Paoletta, will be the General Counsel of the Office of Management and Budget in the Trump administration. This means there’s a good chance that President Trump will impound funds during his administration.

It will go roughly like this: DOGE will identify a woke-looking appropriation, perhaps something to do with environmental justice or renewable energy. The President will send out a Tweet declaring that the money appropriated for this initiative will instead be used to purchase Dogecoin. Thousands on Twitter/X will pronounce this action “based.” A maelstrom of litigation will ensue. Plaintiffs will file suits in friendly venues and obtain quick preliminary injunctions from Democratic appointees. After a few months, the case will reach the U.S. Supreme Court.

In this post I’m going to do two things. First, I’ll give my own take on the legal issues. TLDR: I think the Impoundment Control Act is constitutional.

Second, I’ll try to predict what the Supreme Court will do. But rather than use my own often questionable judgment, I’ll use AI. Several months ago, I wrote a post explaining that AI is very good at reading briefs and predicting Supreme Court decisions. Unfortunately, we don’t have the briefs yet.

Not a problem! In the absence of human-created briefs, I decided to simulate the case from start to finish. I used Claude to draft full-length opening, response, and reply briefs based on a hypothetical impoundment scenario that Claude invented from whole cloth. I then used Claude to simulate an oral argument. Later in the post you can find copies of the briefs and oral argument transcript.

I then put Claude into adjudicator mode and asked it to review those AI-generated briefs and predict how the case would come out. Claude predicted a 6-3 opinion by Chief Justice Roberts upholding the Impoundment Control Act, with Justices Thomas, Alito, and Gorsuch dissenting. Claude also drafted the opinion of the Court, with concurrences by Justices Kagan and Kavanaugh and a dissent by Justice Thomas. Consider it preregistered!

(The theme of this week’s artwork is “pound.”)

The Laws of Constitutional Litigation

Under the Impoundment Control Act, if Congress appropriates money to be spent, the President must spend it. The President can’t rescind spending unless he receives permission from Congress. The President can temporarily defer spending without congressional approval for non-policy reasons (e.g., some unexpected contingency arises). But he can’t defer spending based merely on a disagreement with the policy underlying the appropriation.

Is the Impoundment Control Act constitutional?

Let’s begin by taking a look at the constitutional text, which is occasionally relevant to Supreme Court decisions interpreting the Constitution. The Appropriations Clause provides: “No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law.” This tells us that Congress has the power to enact Laws that “make” Appropriations. “Laws” are generally mandatory, so a “Law” saying “Thou shalt spend $3 billion on Program X” does sound like a Law that makes an Appropriation. Congress therefore has the enumerated power to enact a law that mandates that appropriations be spent.

Proponents of the President’s constitutional right to impound generally don’t argue that the Impoundment Control Act falls outside of Congress’s enumerated powers. Instead, they argue that Article II confers the President with the right to impound. Article II provides that “the executive power shall be vested in a President of the United States of America” and directs the President to “take care that the laws be faithfully executed.”

It may seem odd that the obligation to “take care that the laws be faithfully executed” gives the President the right to refuse to faithfully execute the Impoundment Control Act. But this reasoning ignores the First Law of Constitutional Litigation: “All propositions can be defended by reference to the Constitution’s structure.” Proponents of the right-to-impound theory point to structural separation-of-powers principles: they contend that the Constitution grants the President discretion on how to enforce the law, and the President can’t exercise discretion unless he decides how much money he will spend.

To be fair to this argument, there are, indeed, some circumstances in which the President may declare himself unbound by a statute based on considerations of constitutional structure. Justice Jackson observed in his iconic Youngstown concurrence that “[w]hen the President takes measures incompatible with the expressed or implied will of Congress, his power is at its lowest ebb, for then he can rely only upon his own constitutional powers minus any constitutional powers of Congress over the matter.” But “lowest ebb” doesn’t mean nothing.

In Zivotofsky v. Kerry, 576 U.S. 1 (2015), for example, the Supreme Court struck down a statute providing that for purposes of the “issuance of a passport of a United States citizen born in the city of Jerusalem,” the Secretary of State “shall, upon the request of the citizen or the citizen’s legal guardian, record the place of birth as Israel.” In the Supreme Court’s view, a passport’s statement that the bearer was born in “Jerusalem, Israel” is essentially a diplomatic message from the United States recognizing Israel’s sovereignty over Jerusalem. At the time, however, the President didn’t recognize Israel’s sovereignty over Jerusalem. Thus, the Secretary of State (who reports to the President) argued that he shouldn’t have to follow the statute because it interfered with the President’s authority to recognize foreign sovereigns.

Nothing in the Constitution expressly bars Congress from legislating regarding the content of passports. But the Court identified various constitutional provisions conferring powers on the President that seem in the same ballpark as recognizing foreign sovereigns, such as the power to “receive Ambassadors and other public Ministers,” to nominate our own Ambassadors, and to enter into foreign treaties. Therefore, the Court concluded, the Constitution’s structure establishes that the President has exclusive authority to decide whether to recognize a foreign country’s sovereignty. And so, the Court held, the statute was unconstitutional because it intruded on that power.

I am skeptical of the result in Zivotofsky. Justice Scalia’s dissents are often so persuasive that it is impossible to understand how anyone could have reached the opposite conclusion, and his beautifully-written Zivotofsky dissent is an exemplar of this genre. But Zivotofsky remains on the books, and it confirms that there are some things the President gets to decide without congressional meddling.

Zivotofsky implies that the President can impound funds if spending those funds would interfere with an exclusive presidential power. So if Congress enacted a statute saying “we appropriate $10,000 for the sole purpose of printing new passports saying ‘Jerusalem, Israel’ on them,” the President could decline to spend those funds. The same is true for other exclusive presidential powers: if Congress were to appropriate money for the specific purpose of pardoning a particular federal inmate (the money could be allocated to the cost of the printer paper and ink used to create the document memorializing the pardon), the President wouldn’t have to spend that money.

But setting aside unique cases of that type, neither Zivotofsky nor any other case suggests that the President has the general authority to refuse to spend appropriated funds. And there are several cases suggesting that the President lacks that power:

• Kendall v. United States ex rel. Stokes, 37 U.S. (12 Pet.) 524 (1838). After the Postmaster General refused to pay a contractor for delivering mail, the contractor persuaded Congress to pass a bill requiring the Postmaster General to pay him. The Postmaster General refused, so the contractor sued. Held: the Postmaster General was required to pay.

• Train v. City of New York, 420 U.S. 35 (1975). A federal statute appropriated money for certain environmental projects and stated that “[s]ums authorized to be appropriated … shall be allotted” by the EPA. President Nixon directed the EPA not to spend all of the money. Held: the EPA was required to spend all of the money.

• Clinton v. City of New York, 524 U.S. 417 (1998). The Supreme Court struck down the Line Item Veto Act, which would have allowed the President to cancel certain items of spending. As a practical matter, the line item veto isn’t so different from impoundment, in which the President makes a discretionary decision not to spend certain funds. In Clinton, the Court held that Congress couldn’t give the President this impoundment-like power even if it wanted to.

• Justice Scalia’s Clinton dissent. Justice Scalia’s dissents aren’t technically binding precedent but they’re so persuasive that they may as well be. Justice Scalia would have upheld the Line Item Veto Act because it’s functionally similar to impoundment. But he didn’t endorse a constitutional right to impound. After explaining that Presidents prior to Nixon frequently impounded funds, Justice Scalia observed: “President Nixon, the Mahatma Gandhi of all impounders, asserted at a press conference in 1973 that his ‘constitutional right’ to impound appropriated funds was ‘absolutely clear.’” (I do not understand the Mahatma Gandhi analogy at all but still find it hilarious.). But Justice Scalia then said: “Our decision two years later in [Train] proved him wrong, but it implicitly confirmed that Congress may confer discretion upon the Executive to withhold appropriated funds, even funds appropriated for a specific purpose.” So Justice Scalia also didn’t think the President can impound against the will of Congress.

So, we have cases saying: (1) the President is constitutionally obligated to obey a statute requiring a debt to be paid (Kendall); (2) a statute requiring the President to spend money is judicially enforceable (Train); (3) Congress can’t give the President an impoundment-like power even if it wants to (Clinton majority); (4) Congress gets to decide whether to give the President an impoundment-like power (Clinton dissent). Those cases are hard to square with the right-to-impound theory.

But we cannot forget the Second Law of Constitutional Litigation (really, of all litigation): “Every case is distinguishable from every other case.” If one is sufficiently motivated to recognize a right to impound, one can say that Kendall involved the adjudication of a private claim, Train was a statutory interpretation case, Clinton was a case about “cancellation” of a law rather than a mere refusal to follow it, and Justice Scalia’s Clinton dissent is non-binding and wrong. This would require a lot of motivation, though.

Everything that’s not illegal is mandatory

The right-to-impound argument hinges primarily on history. Before the Impoundment Control Act, Presidents dating back to Jefferson would occasionally impound funds. So, the argument goes, there’s a tradition of impoundment that the Impoundment Control Act can’t override.

This argument is the mirror image of the argument the Supreme Court recently rejected in CFPB v. Community Financial Services Association, 601 U.S. 416 (2024). In that case, the challengers argued that the CFPB (short for Consumer Financial Protection Bureau) was unconstitutional because, under the CFPB’s unique funding scheme, the President could impound.

That’s an oversimplification, but only by a bit. The CFPB is funded in an unusual way. Most agencies have to clamor for funds during the annual appropriations process. But the CFPB doesn’t—it’s authorized to draw an amount from the Federal Reserve System that its Director deems “reasonably necessary to carry out” the agency’s duties, subject to an inflation-adjusted cap that was approximately $734 million in 2022. The challengers objected to several aspects of this scheme (e.g., the fact that it was permanent), but the molten core of the challengers’ argument was that the agency got to decide how much to spend, as this back-and-forth from the oral argument illustrated:

JUSTICE THOMAS: Mr. Francisco, just briefly, I'd like you to complete this sentence. Funding of the CFPB is -- violates the Appropriations Clause because?

MR. FRANCISCO: Because Congress has not determined the amount that this agency should be spending. Instead, it has delegated to the director the authority to pick his own appropriation, subject only to an upper limit that's so -- so high it's rarely meaningful.

I disagree that $734 million is an amount that’s “so high it’s rarely meaningful”; I promise you that a federal agency can spend $734 million on staplers without any difficulty. But, the CFPB Director was indeed guilty as charged of having the discretion to decide how much he would spend up to that amount.

The Fifth Circuit held that this scheme was unconstitutional, finding it to be “the epitome of the unification of the purse and the sword in the executive—an abomination the Framers warned ‘would destroy that division of powers on which political liberty is founded.’” Sounds bad!

The Supreme Court reversed the Fifth Circuit. It reasoned that there’s nothing in the Constitution saying the CFPB can’t be funded this way. The Court also examined the historical record and noted that prior to the Founding, legislatures had significant flexibility:

In short, the origins of the Appropriations Clause confirm that appropriations needed to designate particular revenues for identified purposes. Beyond that, however, early legislative bodies exercised a wide range of discretion. Some appropriations required expenditure of a particular amount, while others allowed the recipient of the appropriated money to spend up to a cap. Some appropriations were time limited, others were not. And, the specificity with which appropriations designated the objects of the expenditures varied greatly.

Likewise, in the early history of the United States, appropriations bills “adhered to the minimum requirements of an identifiable source of public funds and purpose,” but other than that, they “displayed significant variety in their structure.” Sometimes Congress would appropriate lump sums, sometimes it would appropriate sums up to a cap, sometimes it would appropriate in other ways. Based on this historical evidence, the Court held that Congress can—if it wants to—let the President decide how much money he spends:

At the outset, the associations argue that the Bureau’s funding is not “drawn . . . in Consequence of Appropriations made by Law” because the agency, rather than Congress, decides the amount of annual funding that it draws from the Federal Reserve System. This argument proceeds from a mistaken premise. Congress determined the amount of the Bureau’s annual funding by imposing a statutory cap. … The only sense in which the Bureau decides its own funding, then, is by exercising its discretion to draw less than the statutory cap. But, as we have explained, “sums not exceeding” appropriations, which provided the Executive with the same discretion, were commonplace immediately after the founding. … Thus, we cannot conclude that Congress violated the Appropriations Clause by permitting the Bureau to decide how much funding to draw up to a cap.

With the change in Administrations, we now have the mirror-image argument. Instead of arguing that it’s unconstitutional to let the President decide how much to spend, proponents of the right-to-impound theory argue that it’s unconstitutional not to let the President decide how much to spend. This is a nice example of the First Law of Constitutional Litigation at work: the proponents are arguing that a system that the Fifth Circuit deemed a separation-of-powers-destroying “abomination” is actually mandatory under the separation of powers.

If this issue reaches the Supreme Court, I predict the Court will uphold the Impoundment Control Act. In CFPB, the Court observed: “early legislative bodies exercised a wide range of discretion. Some appropriations required expenditure of a particular amount, while others allowed the recipient of the appropriated money to spend up to a cap.” The Court inferred that Congress continues to have that discretion today.

That’s why the right-to-impound theory loses. Early legislatures sometimes “allowed the recipient of the appropriated money to spend up to a cap”—and Congress can keep doing that today if it so desires, as it has with the CFPB. Early legislatures also sometimes “required expenditure of a particular amount”—and Congress can also keep doing that today if it so desires, as it has under the Impoundment Control Act. The historical argument for the right to impound fails for the same reason as the historical argument in CFPB: it confuses “X has sometimes happened” with “X is constitutionally mandatory.” Early legislatures sometimes controlled how much the President spends, but CFPB holds that such a scheme isn’t constitutionally mandatory. Likewise, early legislatures sometimes didn’t control how much the President spends, but that’s not constitutionally mandatory, either.

The inverse major questions doctrine

The advent of the so-called “major questions doctrine” also makes me skeptical of a right to impound.

The major questions doctrine teaches that Congress, not the administrative state, is responsible for deciding national policy on issues of major political and economic importance. Thus, an agency isn’t permitted to rely on vague and general statutory language as a basis for regulating on a major question, even if that vague and general language, read literally, would seem to authorize the regulation. Instead, Congress must speak clearly before an agency may enact such a regulation.

For example, in Biden v. Nebraska, 600 U.S. 477 (2023), the Supreme Court struck down the Department of Education’s rule canceling $430 billion of student loans. The Department argued that it had statutory authority to issue the rule: a statute authorized the Department to “waive or modify any statutory or regulatory provision applicable to the student financial assistance programs” at issue, so the Department argued that it was “waiving or modifying” the requirement to pay back the principal on the loans. The Department had a decent argument based on the literal words of the statute. But, the Court held, c’mon … when Congress authorized the Department to “waive or modify any statutory or regulatory provision,” it was talking about little modifications, not canceling $430 billion worth of loans!

The major questions doctrine has spawned a vast academic literature, to which I decline to contribute. Instead, I will observe that the right-to-impound argument is essentially the inverse of the major questions doctrine. The question of how much the government should spend is the paradigmatic example of a major question. In the Impoundment Control Act, Congress provided that agencies cannot decide that major question by themselves. Whereas the major questions doctrine states that agencies can’t decide major questions even when statutes, read literally, authorize them to do so, the right-to-impound theory argues that agencies can decide major questions even when statutes ban them from doing so.

Suppose Congress enacted a law saying “the Department of Education is not allowed to cancel student loan debt under this statute,” and the Department then argued: “too bad, we get to decide how we’re going to execute the law, so we’re just not going to enforce the requirement to pay back the principal on the loans anymore.” This argument would be in some tension with the result in Biden v. Nebraska, yet that is the logical implication of the right-to-impound theory.

I understand that attitudes towards the separation of powers, on both sides of the house, shift upon a change in administration. But up to a point! I can’t understand how, in a world with the major questions doctrine, an agency gets to decide whether major federal programs, costing billions of dollars, will exist.

Now to the good part

I am pretty confident the Supreme Court would uphold the Impoundment Control Act as to most factual scenarios. But it’s always nice to ask someone smarter than you.

I asked Claude to construct a hypothetical impoundment scenario and draft Supreme Court briefs litigating the constitutional dispute. AI loves to hallucinate, so why not let its creative juices flow? You cannot just say “write the briefs” to Claude—yet—but with appropriate hand-holding and strategic prompting, you can, indeed, get it to write the briefs.

Sometimes I hear AI skeptics saying things like “AI is great for simple filings, but in complex cases, where Real Thinking is needed, it’s not helpful.” This is emphatically not the case. AI can understand and propose every argument that might plausibly appear in a Supreme Court brief. I promise you that today’s frontier models, which can ace the Putnam Exam, are perfectly capable of achieving insights such as “the President isn’t Taking Care to enforce the law when he’s not following the law.” Also, there’s a limited body of Supreme Court case law, all of which is known to AI, so AI is less likely to hallucinate in a Supreme Court brief than a lower-court brief (though it does hallucinate sometimes).

Anyway, Claude dreamed up a scenario where President Trump impounded a provision of the “Consolidated Appropriations Act, 2025” which “included $3.8 billion in funding for the Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE).” This is an amusingly plausible scenario given that the real 2023 Consolidated Appropriations Act appropriated $3.46 billion for that very program. Claude confidently described tweets, congressional hearings, a plausible-sounding plaintiff to challenge the impoundment (“Brightpath Solar Manufacturing”), a district court decision by Judge Alsup enjoining the impoundment, and a Ninth Circuit opinion affirming that decision, complete with a trenchant dissent by Judge VanDyke.

Here are the briefs:

Brief Of Petitioner Trump V

208KB ∙ PDF file

Download

Download

Brief Of Respondent Trump V

207KB ∙ PDF file

Download

Download

Reply Brief Of Petitioner Trump V

135KB ∙ PDF file

Download

Download

I then persuaded Claude to generate an argument transcript, which doesn’t pass the Turing Test but still captures the typical dynamic of a Supreme Court argument quite well:

Oral Argument Transcript Trump V

108KB ∙ PDF file

Download

Download

So how does Claude think the Supreme Court will rule? Drumroll please:

Based on my understanding of the current Supreme Court and these complex legal arguments, I predict that the Court would rule 6-3 in favor of upholding the Impoundment Control Act’s constitutionality, affirming the Ninth Circuit. Here’s my analysis:

The Majority (Roberts, Barrett, Kagan, Sotomayor, Jackson, and likely Kavanaugh):

• Chief Justice Roberts would likely write the majority opinion, as he often does in major separation of powers cases. He has consistently shown concern about maintaining institutional checks and balances and would likely view unilateral presidential impoundment authority as a significant threat to Congress’s constitutional power of the purse.

• Barrett, despite her generally conservative approach, has expressed strong textualist views about separation of powers and would likely be persuaded by the clear text of the Appropriations Clause and the long history of congressional control over spending.

• Kavanaugh, though conservative, tends to take a more moderate approach in separation of powers cases and has shown respect for established statutory frameworks. The ICA’s 50-year history and workable procedures would likely influence him.

• The three liberal justices (Kagan, Sotomayor, Jackson) would almost certainly vote to uphold the ICA based on both statutory and constitutional grounds.

The Dissent (Thomas, Alito, Gorsuch):

• Justice Thomas would likely write a strong dissent arguing for broad executive power under Article II and the Take Care Clause.

• Alito and Gorsuch would join, emphasizing historical examples of presidential impoundment and arguing for inherent executive authority to protect national interests.

Key factors influencing this outcome:

1. The Court’s institutional conservatism and reluctance to dramatically alter long-established separation of powers frameworks

2. The clear constitutional text giving Congress power over appropriations

3. The ICA’s long history of successful implementation across administrations of both parties

4. The availability of alternative mechanisms for presidential influence over spending (like the regular appropriations process)

5. Concern about creating an effective line-item veto power that the Court has previously ruled unconstitutional

The majority opinion would likely emphasize Congress’s fundamental power of the purse while acknowledging some executive discretion in implementation within statutory bounds. The dissent would focus on Article II powers and historical practice supporting broader executive authority.

And here is the Chief Justice’s majority opinion, concurrences by Justices Kagan and Kavanaugh, and Justice Thomas’s dissent. (Claude kept the opinions reasonably short - perhaps for the best!)

Supreme Court Decision Trump V

113KB ∙ PDF file

Download

Download

Note that I did this using a general-purposes, non-fine-tuned model. I didn’t use the Enterprise version of Claude, so I had a small context window and severe rate limits. I also didn’t test OpenAI’s $200-per-month o1-pro model, which I’ve heard is a great lawyer.

Thus, technology that is already available today could have produced better briefs than these—to say nothing of technology that will be available three months from now.