GAO’s role in appropriations oversight

The new Trump administration is staking out aggressive views of the president’s power over the purse. Challenging the traditional understanding that it is Congress, not the president, who wields that power, the new administration is, among other things, refusing to spend vast swaths of appropriated dollars; spending money in ways that appear to go beyond what Congress appropriated; trying to attach novel conditions to spending in support of the president’s priorities without clear statutory authorization; and dismantling the website on which Congress has required centralized decisions about disbursements of funding to agencies to be posted. These early actions are in line with the plan articulated by senior administration officials at the end of the first Trump administration to “defend the president’s authority to spend money at any time and in any manner that he determines appropriate.”

Many lawsuits challenging these actions have already been filed, and many temporary restraining orders and preliminary injunctions are currently in place. Beyond courts, however, an additional institution is likely to play an important role in evaluating the administration’s compliance with appropriations law: the Government Accountability Office (GAO). GAO issued decisions disagreeing with the Trump administration multiple times on appropriations issues between 2017 and 2021. As the new administration tries to assert vast amounts of power over federal dollars, it is an opportune moment to take stock of GAO’s role in appropriations law and to make sense of this role in interbranch relations.

Appropriations law is a subfield of administrative law, the general body of law governing how agencies operate and interact with the president, Congress, and courts. The typical appropriations law question asks whether an agency can spend the amount of money it wants to spend in the way that it wants to at a particular time—often called the purpose, time, and amount requirements.

While most administrative law questions are resolved in court or in the shadow of judicial review, appropriations law questions are mostly not. This is at least in part because there are preliminary hurdles to getting into court that tend to be more difficult in the appropriations context. For example, there is often no clear party injured by an agency’s decision to spend money in a certain way who would have standing to bring a lawsuit. In addition, much appropriations language provides only general directions to an agency, with more specific decisions “committed to agency discretion by law” and thus not judicially reviewable. The Trump administration is arguing that these hurdles exist in many of the current court challenges.

Historically, then, most appropriations law questions have been resolved in the shadow of political, not judicial, review, making appropriations law a prime site for the study of the congressional-executive relationship. Since its creation in 1921, GAO has played an important support role to Congress in appropriations oversight. GAO investigates appropriations controversies brought to its attention; publishes a treatise on federal appropriations law; provides appropriations law trainings; offers informal advice on appropriations law questions; and issues formal appropriations law decisions.

As part of broader research for a book on how agencies have traditionally executed appropriations laws that one of us is writing, we have spent time over the last 18 months reading and coding these formal appropriations law decisions. Here, we report our top 10 findings from a preliminary dataset covering the 126 decisions evaluating executive branch conduct over the last 12 years—that is, conduct undertaken by the Biden administration, the first Trump administration, and the second Obama administration (January 20, 2013, through January 19, 2025).¹

To be sure, this dataset does not exactly compare apples to apples, as it includes the second of two contiguous terms for Obama and one term each for Trump and Biden, respectively. Nonetheless, it allows comparison across four years for each of the three most recent presidents. While we are continuing to review and code decisions back through previous administrations for the larger book project, this preliminary dataset provides a useful lens to help contextualize what we anticipate will be many GAO decisions responding to the current Trump administration’s appropriations law actions.

Unpacking the nuanced way these decisions are requested, investigated, and decided illustrates the complexity of interbranch relations and the dynamic relationship between different components of Congress and the executive branch across administrations. The bottom line: While it would be valuable for GAO to be able to complete its investigations and publish its conclusions more quickly so that Congress and agencies can respond in a timely way, GAO nonetheless plays a critical role in providing oversight as agencies implement the appropriations Congress entrusts to them to carry out the government’s work. GAO has done especially important work in recent years evaluating a rise in arguments that the executive branch has impounded, or refused to spend, appropriated dollars. And contrary to arguments that GAO is biased when it issues decisions that one political party does not like, the evidence shows instead that GAO interprets the law and applies the law to the facts in a nonpartisan, impartial way. 

GAO is a congressional support agency, but just shy of half (62/126) of the appropriations law decisions came from requests made by Congress. Agencies themselves requested a GAO opinion on something they had done or wanted to do in just over 41% (52/126) of the decisions. In other words, during the 12 years reflected in the dataset, agencies themselves turned to GAO for a formal legal opinion on an appropriations matter with almost as much regularity as Congress did.

At first blush, this turn to GAO by agencies seems surprising. Why would agencies affirmatively reach out for oversight, especially in light of ongoing instruction from the Office of Legal Counsel that GAO opinions are merely advisory rather than binding?

It might make sense if agency requests came largely from an Office of Inspector General (OIG), as inspectors general might be expected to turn to GAO in its capacity as a fellow watchdog over federal funds more than core agency offices entrusted with spending those funds would. Indeed, a healthy percentage of requests, 42.3% (22/52), did come from OIGs. Yet that still leaves the vast majority of agency requests coming from other offices; 25% (13/52) came from an agency’s Office of General Counsel, while 32.7% (17/52) came from a variety of other sources, such as certifying officers in military installations or senior officials at smaller agencies. What explains this voluntary request for oversight from another branch?

It turns out that agencies, in asking for GAO appropriations law decisions, are relying on a longstanding statutory provision that allows them to do so in order (as GAO explains) to obtain “a measure of protection and counsel” against individual agency employees being held personally accountable for repaying “illegal or improper payments.” Congress has provided that those decisions are “conclusive” on GAO in settling accounts. Therefore, if GAO has blessed a certain use of an appropriation, the agency official who signs off on the spending can do so without fear that GAO in a subsequent audit will determine that repayment is required. Agencies thus have an incentive to obtain GAO appropriations law decisions, even though the Office of Legal Counsel treats those decisions as merely advisory.

In addition to publishing appropriations law decisions in response to congressional requests and agency requests, GAO can also issue certain appropriations law decisions on its own initiative. In our dataset, GAO initiated decisions acting on its responsibilities under the Impoundment Control Act or the Antideficiency Act in 8.7% (11/126) of the decisions. These two laws, which one of us has called the Power of the Purse statutes, help make Congress’ spending choices mandatory in both directions. Under the Impoundment Control Act, the executive branch cannot ordinarily obligate or expend less than Congress has appropriated, while under the Antideficiency Act, the executive branch cannot obligate or expend more than Congress has appropriated. Congress has tasked GAO with reporting requirements under each law.

While the data do not show any statistically significant difference in the frequency of agency requests by administration, the GAO-initiated inquiries all involved the Trump administration. Given that administration’s novel constructions of the Antideficiency Act and Impoundment Control Act, as discussed more below in Findings #7 and #9, this fact is not surprising.

Finally, in addition to congressional requests, agency requests, and GAO-initiated decisions, our dataset also includes one GAO decision issued in response to a federal labor union’s request for reconsideration. Initial decisions are published only in response to members of Congress or agency officials or on GAO’s own initiative. Still, GAO protocol allows either the initial requester or any “other entity with a stake in a recent decision” to request reconsideration based on an “alleged error of fact or law or new relevant and material information” that was not available at the time of the initial decision. Although our dataset of 126 decisions includes three requests for reconsideration, only one was from an outside entity, underscoring the primacy of the other requesters in the universe of GAO appropriations law decisions. That GAO upheld all three of its initial decisions in these reconsideration decisions also helps explain why few outside entities request reconsideration; such requests do not appear to be likely to produce a different outcome, given the narrow standard for reconsideration.

One might expect that GAO appropriations law decisions would be primarily requested by the appropriations committees themselves as a form of oversight as compared to other congressional committees. But authorizing committees requested GAO appropriations law decisions almost as frequently—22 requests (35.5% of congressional requests), compared to 26 requests from appropriations committees (41.9%). The Budget committees, in turn, requested four decisions (6.5%), while members of Congress acting in individual (rather than committee) capacities called for decisions 10 times (16.1%). The executive branch’s compliance (or lack thereof) with appropriations law requirements from Congress thus seems to be of broader interest. (There appears to be no meaningful difference in the number of requests originating in the House or Senate.)

In keeping with the idea that appropriations law can be a vehicle to address more substantive issues, decisions do not merely reflect minor questions about proper accounting. We coded for three different levels of what we called “severity.” Decisions received the lowest rating, a 1, if they dealt purely with accounting issues and the underlying issue did not appear to implicate any real questions of policy. Decisions received a rating of 3 if they dealt with high-profile substantive decisions on which a presidential candidate might campaign or that might be the focus of a major news cycle. In between these poles, decisions received a rating of 2 if the issue bore some hallmarks of mere accounting but also had some indicia of greater policy or political import, such as if there was some external news coverage about the underlying controversy. With the exception of noting if the request came from congressional leadership or included a lengthy list of congressional signatories, which might be relevant to a rating of either 2 or 3, our severity rating focused on the substance of the issue rather than the identity of the requester.

While 50 decisions (39.7% of the total) reflected issues that we coded at the lowest level of severity—the typical such case involved whether an agency could use its appropriation to pay for bottled water in a particular extenuating circumstance—the rest involved issues of greater import.

We coded another 44 decisions (34.9%) at the intermediate level; these decisions involved, for example, whether the Obama administration EPA’s use of social media in connection with its “Waters of the United States” rulemaking violated a prohibition on using appropriations for publicity or propaganda; whether the Trump administration’s Interior Department properly paid for parts of the Fourth of July events on the National Mall; and whether the Biden administration’s USDA properly used its Commodity Corporation Credit funds to award certain climate-related grants.

We coded 32 decisions (25.4%) at the highest tier, including whether the Obama administration’s Department of Defense violated the Antideficiency Act when it transferred five individuals detained at Guantanamo to Qatar, given notification requirements in the agency’s appropriations law for that year; whether the Trump administration’s Department of Defense violated transfer restrictions when it moved $2.5 billion from a drug interdiction account to build a wall at the southern border after an appropriations dispute between the president and Congress failed to produce the funding the president sought; and whether the Biden administration’s Department of Homeland Security violated the Impoundment Control Act by pausing construction at the wall consistent with a presidential proclamation. These decisions illustrate how Congress can use GAO appropriations law decisions as an oversight tool for addressing broader issues.

The severity rating appears to vary depending on the type of requester. Overall, agency requests involved lower-level issues, with an average severity rating of 1.3, compared to an average of 2.2 for congressional requests. This variation is highly statistically significant (p = 0.00). For its part, decisions initiated on GAO’s own authority are also strongly associated with higher severity levels, with an average severity rating of 2.9 (p = 0.00). The pattern of congressional requests focusing on more weighty matters than agency requests is consistent no matter which administration’s conduct is at issue, as is GAO’s severity level.

Within congressional requests, authorizing and oversight committees requested decisions about more consequential topics than appropriations committees, with an overall severity rating of 2.4 for the former, compared to 1.9 for the latter. These differences are marginally statistically significant (p = 0.07). While the severity ratings are higher for Budget Committee requests (all severity level 3.0) and lower for requests made by individual members (an average severity rating of 2.1), these variations are not statistically significant, perhaps due to sparse data.

Severity in the issue underlying congressional requests appears to be increasing over time. The average severity level in congressional requests for decisions involving the Obama administration was 1.9, while it was 2.3 for the Trump administration and 2.6 for the Biden administration. These differences are statistically significant (p = 0.006 for the Trump administration being higher than the Obama administration and p = 0.032 for the Biden administration being higher than the Obama administration).

Most of the congressional requests for appropriations law decisions investigating the Obama and Biden administrations came solely from Republicans, whether they were in the majority or the minority—18 out of 26, or 69.2%. In turn, most of the congressional requests for appropriations law decisions investigating the Trump administration came solely from Democrats, again whether they were in the majority or minority—27 out of 33, or 81.8%. These relationships between the administration and the political party of congressional requesters are statistically significant (p < 0.001).

This partisan reality makes sense, especially given the previous finding that 60% of all the appropriations law decisions in this dataset went beyond mere technicalities to involve substantive policy or political debates. These findings also reflect GAO’s openness to requests from ranking members of committees (not just chairs) and from individual members. They are also consistent with previous scholarship concluding that the majority party uses GAO more in times of divided government and that GAO provides a power-balancing mechanism for the minority party in Congress. (We discuss the bipartisan requests in Finding #5 below.)

What is perhaps more surprising is that there appear to have been fewer Republican requests for decisions about the Biden administration (five) than about the Obama administration (13). To be sure, the number of Biden-era decisions in our dataset is not complete. GAO has issued two decisions involving Biden-era appropriations issues since Inauguration Day, the cut-off date for the dataset discussed here, and if the timing of GAO decisions issued about the two previous administrations is any indication, additional decisions about the Biden administration will continue to be issued over the next year or two. There is no public docket for pending decisions, though, so there is no way to know how many requests about Biden-era conduct are in the queue. It is possible, however, that the lower number reflects some Republican disenchantment with using GAO because of its findings on impoundment and the Trump administration’s appropriations law conduct, as discussed further below in Findings #9 and #10.

Twenty-one of the 62 congressional requests, or more than a third, were made solely by Republicans or solely by Democrats but involved requesters from both the House and Senate—that is, they were not bipartisan, but they were bicameral. Of these, eight came from Republicans and 13 from Democrats, again reflecting the pattern of heavier use by Democrats. While some of these requests involved technical matters like whether the CFTC had properly accounted for various leases across three administrations, others involved more weighty matters, such as Democrats requesting decisions about whether the Trump administration violated the Antideficiency Act by keeping various agency activities running during the 2019 shutdown or Republicans requesting a decision about whether the Obama administration could properly use appropriations to pay for the Affordable Care Act’s risk corridor program.

Notably, more than half of the all-Republican or all-Democrat bicameral requests came from authorizing/oversight committees (n=11) compared to only a quarter of these requests coming from appropriations committees (n=5), a variation that is statistically significant (p = 0.029). This distinction might suggest that requests for GAO appropriations law decisions are a particularly important signal of partisan opposition to executive branch conduct for authorizing/oversight committees. In turn, one of these one-party bicameral requests came from budget committees and four came from members of Congress in their individual (rather than committee) capacities. These latter variations are not statistically significant, again perhaps due to sparse data.

While the overwhelming majority of congressional requests were made solely by Democrats or solely by Republicans, eight requests (12.9%) were bipartisan. In contrast with bicameral decisions requested by only one political party, none of the bipartisan requests came from authorizing/oversight committees. Instead, seven came from appropriations committees, and one was from the budget committees. This variation among congressional committee type with respect to bipartisanship is statistically significant (p = 0.020). These bipartisan requests appear to reflect institutional concerns about Congress’ power of the purse rather than partisan oversight, so perhaps it makes sense that the committees with the greatest stake in that issue operate in the most bipartisan way in the dataset. Half of the bipartisan requests were also bicameral, furthering the sense that Congress was acting in a fully institutional sense against the executive branch.

The decisions in the dataset that were issued in response to bipartisan requests arise in three different contexts. For two of the decisions in the dataset, GAO began its investigation in response to directions in appropriations law text, a full committee report to an appropriations law, and/or a joint explanatory statement to an appropriations law that instructed GAO to investigate whether an agency’s current practice on using its appropriations in a particular way complied with relevant restrictions. Five grew out of requests from appropriations subcommittee chairs and ranking members for GAO to investigate whether an agency was complying with a limitation or notification requirement in an appropriations law. The last, from the House Budget Committee’s chair and ranking member, asked GAO to interpret a provision of the Impoundment Control Act that the Trump administration was then contesting. Of these bipartisan requests for decisions, five were made about the Obama administration, three were made about the Trump administration, and none were made about the Biden administration.

Use of this tool by co-partisans of the president also exists, though it is slightly less common than bipartisan requests—six requests, or 9.7%. As with bipartisan use, co-partisan use also appears to be declining, although again, the small numbers prevent us from saying this with statistical confidence. During the Obama administration, three of the 21 congressional requests for decisions (14.3%) came from Democrats, while during the Trump administration, three of the 33 congressional requests for decisions (9.1%) came from Republicans—and none of the five congressional requests for decisions came from Democrats during the Biden administration.

The co-partisan requests vary in their political significance. One in the Obama administration and one in the Trump administration came from members of Congress acting in their individual (rather than committee) capacities about a matter of home-state interest. The other Obama-era co-partisan decisions came from the ranking members of a relevant authorizing/oversight committee about substantive issues but not weighty matters of interbranch disputes. 

However, the two remaining Trump-era co-partisan requests came from the chair of either the House or Senate Appropriations Committee asking about matters of interbranch consequence—the Social Security Administration’s compliance with a reprogramming consultation and notification requirement in its appropriations law and Homeland Security’s compliance with the Impoundment Control Act. These types of co-partisan inquiries thus run the gamut from protecting home-state constituents to protecting a diffuse national public to protecting Congress’ own prerogatives over appropriations.

We identified four types of agencies at issue in these decisions, which we categorized by referring to the organizational chart in the U.S. Government Manual and the list of independent commissions surveyed by Neal Devins and David Lewis in their award-winning law review article on independent agencies: those in the Executive Office of the President (EOP), cabinet departments, independent commissions, and non-cabinet executive branch administrations.

Using this coding, 7.9% (10/126) of the decisions in our dataset involved the Executive Office of the President (such as OMB and OIRA); 56.3% (71/126) involved cabinet departments or their subcomponents (such as the Departments of Defense, Agriculture, and Transportation); 23.8% (30/126) involved independent commissions (such as the Commodity Futures Trading Commission and Consumer Product Safety Commission); and 15.1% (19/126) involved non-cabinet executive branch administrations (such as the EPA and Social Security Administration). (Percentages add up to more than 100% because several decisions involve multiple types of agencies.)

Requests from agencies and from Congress tend to vary by type of agency, however. Of the congressional requests, 64.5% (40/62) involved oversight of a cabinet agency, while only 50% (26/52) of agency-initiated requests came from cabinet agencies. In turn, 38.5% (20/52) of requests from agencies came from independent commissions, but only 12.9% (8/62) of requests from Congress involved oversight over independent commissions. This variation is statistically significant (p = 0.005). (There are smaller differences for executive branch administrations—19.4% (12/62) of requests from Congress involved oversight over executive branch administrations, and 11.5% (6/52) of requests from agencies came from executive branch administrations—and this variation is not statistically significant.)

It may be that Congress focuses its attention on appropriations law conduct of cabinet agencies as closest to presidential activity, while independent commissions (especially smaller ones) are less well-resourced internally on appropriations law than cabinet agencies are and need additional guidance from GAO or feel less bound by executive branch guidance disfavoring outreach to GAO. Not surprisingly, given that guidance, no entity within the Executive Office of the President requested a decision in this dataset, in contrast to six congressional requests and four decisions issued on GAO’s own initiative that involve the EOP.

The agency that had the most decisions in this dataset–12 out of 126 (9.5%)—was the Defense Department, including questions involving spending on individual bases, within different components, and at the centralized department level itself. Since almost half of all discretionary spending is defense spending, it is not surprising that the greatest number of decisions involved DOD in some way.

Eleven agencies had between five and 10 decisions: the Commodity Futures Trading Commission, Agriculture, Interior, Commerce, Energy, Homeland Security, Housing and Urban Development, Treasury, Health and Human Services, EPA, and OMB. Nine more agencies had between two and four decisions: Transportation, the Election Assistance Commission, the Consumer Product Safety Commission, the General Services Administration, Labor, the Office of Navajo and Hopi Relocation, the Smithsonian Institution, the Social Security Administration, the U.S. Postal Service, and the U.S. Chemical Safety and Hazard Investigation Board. In turn, 18 additional agencies had only one decision. The overall impression is of breadth rather than depth.

GAO protocol involves obtaining an “agency explanation of pertinent facts and its views on the law” before issuing a legal decision. In the vast majority of cases, agencies in our dataset complied with GAO’s request for this material, and GAO described without comment the back-and-forth between the agency and GAO in sharing such explanations and views. Only three times—all in the Trump administration—did an agency decline to provide the requested information to GAO.

In the first of these instances, the Democratic ranking members of two authorizing/oversight committees had asked GAO to investigate whether phone calls the Interior Department Secretary had made to the senators from Alaska violated the anti-lobbying restriction applicable to all government funds or any other restriction on Interior’s appropriation. The issue involved an apparent threat to withhold federal funds from Alaska after Republican Senator Murkowski’s vote against repealing the Affordable Care Act doomed that effort. Interior confirmed that phone calls took place but declined to provide any information about the content of those calls, either to GAO or to the agency’s inspector general. Consequently, GAO explained, “we lack the requisite facts on which to base a legal opinion,” and closed the matter without resolving it.

In the two subsequent instances, GAO issued a decision notwithstanding the failure of the agency to provide information—EPA about the agency’s tweet praising confirmation of its deputy administrator over Democratic opposition, and Interior again, this time about the agency’s continuing to operate during the government shutdown of 2018–19. In each case, GAO was able to rely on publicly available information to gather sufficient facts to respond to the congressional requesters, a combination of Democratic ranking members of authorizing/oversight and appropriations committees. But it did so without the benefit of hearing the agency’s own analysis and views.

In principle, GAO could file a lawsuit to obtain relevant agency records and information under a statutory provision authorizing a civil action where an agency refuses to provide that material. In 2002, however, in the one case in which GAO tried to use this provision, a district court held that GAO had no standing. While that decision had certain unique features that conceivably would not apply more broadly—for example, it involved a lawsuit to obtain records from the vice president, a constitutional officer, rather than an agency—GAO has tended to proceed informally rather than in the courts since then, even though a bipartisan Congress reaffirmed its authorizing of such a lawsuit in 2017.

Six additional decisions are also relevant to the issue of agency non-cooperation: decisions in which GAO reported to the president of the Senate and the speaker of the House that agencies had failed to report to Congress Antideficiency Act violations that GAO had previously identified. Three of these grew out of the Trump administration’s activities during the 2018–19 shutdown. In these cases, agency non-cooperation with GAO moved beyond individual acts to a government-wide policy dispute between the Trump administration’s OMB and GAO. OMB’s general counsel notified agencies that they should no longer report GAO-identified violations to Congress unless OMB and the agency agreed that a violation had occurred. GAO responded that if agencies fail to report, GAO will not only report the violation to Congress but will also report agencies’ own failure to do so. While the Biden administration reverted to the pre-Trump OMB view that agencies would generally report GAO-identified violations even when they disagreed that a violation had taken place, this interlude reflects the contingency of agency cooperation with GAO.

It can take GAO well over a year to issue a decision from the date it receives a request letter, an overall average of 435.8 days. The speed with which GAO is able to finalize a decision appears to vary by how serious the underlying issue is, however; GAO publishes decisions on the weightiest matters more quickly than decisions on more mundane issues. Matters with severity rating 1 take 570.4 days on average, while matters with severity rating 2 take 420.0 days on average, and matters with severity rating 3 take 255.7 days on average. Controlling for administration and requester party, the differences in publication times across severity levels are statistically significant (p = 0.054 for the difference between severity rating 2 and 1, and p = 0.002 for the difference between severity rating 3 and 1).

Developing the record and researching the law take time, of course; as GAO protocol explains, it solicits agency views through a “development letter”; reaches out for the views of other entities (such as congressional committees) with a stake in the matter; holds additional conversations with agency staff as needed; and applies its own relevant decisions as precedent. This careful work, accompanied by thorough reason-giving, is key both to providing requesters with useful answers and bolstering GAO’s own institutional legitimacy.

At the same time, the length of time from request to decision arguably limits the effectiveness of these decisions as an oversight tool. This may be especially the case for congressional requests, when presumably the requesters would like an opportunity to revise appropriation language or take some other action in response to agency action it wishes to change. It is also especially the case when the decision comes out after an administration has changed.

Still, the decisions can sometimes play a role in timely congressional oversight, particularly when GAO publishes a decision about a serious, time-sensitive matter relatively quickly. For example, in 2018, GAO took only 40 days to issue two decisions rejecting the Trump administration’s efforts to interpret the Impoundment Control Act to let it run out the clock on the fiscal year and thereby achieve an impoundment in effect, enough time to factor into the negotiations around the full-year appropriations for that fiscal year. (To be sure, those decisions resolved exactly the same legal question, just raised by two sets of congressional requesters, and the legal question did not depend on factual development, which is the typical source of delay.) GAO also took only 78 days to issue a decision concluding that the Trump administration’s efforts to withhold funding for Ukraine in August of 2019 violated the Impoundment Control Act, resulting in publication before the first impeachment trial, which centered on the permissibility of those efforts. 

There were zero decisions about the Impoundment Control Act during the four years of the Obama administration in our dataset, but eight decisions applying that Act to actions taken during the first Trump administration and two applying that Act to actions taken during the Biden administration. Early in the first Trump administration, there appeared to be some interest in institutional protection of Congress’ power of the purse, in keeping with the bipartisan nature of the congressional sponsorship of and votes for that Act when it was originally passed in 1974. That initial interest appears to have given way to more partisan inquiries.

Three inquiries in 2017 and 2018 into the Trump administration’s potential impoundment conduct stemmed from GAO’s own instigation, a Republican request, and a bipartisan request, respectively. GAO found that the administration had violated the law in all three instances. In 2017, GAO investigated the potential impoundment of some Energy Department funding during a period in which President Trump’s budget request to eliminate those funds was pending. GAO concluded that the agency had, in fact, impounded in violation of the Act but had ultimately released the funds after GAO’s investigation began.

Next, also in 2017, the Republican chair of the Senate Committee on Appropriations asked GAO to investigate the potential impoundment of Homeland Security’s FY17 Coast Guard funds during a period when it was unclear whether Congress would rescind those funds in accordance with President Trump’s budget request for FY18. Once more, GAO concluded that the agency had improperly impounded the funds, though again, it had ultimately obligated them during the investigation.

And in 2018, the Republican chair and Democratic ranking member on the House Committee on the Budget requested GAO’s views on a legal argument the Trump administration was then making about the Impoundment Control Act, which was to assert that the Act allowed an administration to delay spending close to the end of the fiscal year, effectively rescinding the funds without any affirmative action by Congress. GAO rejected this interpretation, concluding that “[t]he statutory text and legislative history of the ICA, Supreme Court case law, and the overarching constitutional framework of the legislative and executive powers provide no basis to interpret the ICA as a mechanism by which the President may unilaterally abridge the enacted period of availability of a fixed-period appropriation.”

The remaining five impoundment decisions about the Trump administration were either prompted by Democratic inquiries or developed under GAO’s own investigative authority under the Impoundment Control Act. The Trump administration saw mixed success in these decisions. A 2018 bicameral inquiry from the Democratic vice chair and ranking member of the Senate and House Committees on Appropriations, respectively, on the same issue of interpretation of the Impoundment Control Act raised by the bipartisan House Committee on the Budget produced the same result against the Trump administration. But a 2019 Democratic inquiry from House members acting in their own individual capacities into whether the National Weather Service had improperly impounded when it had a delay in filling many open positions produced a GAO decision finding no violation.

As for arguably the most politically salient of the impoundment decisions—inquiries driven by GAO itself under its own statutory authority into the Trump administration’s withholding of funds to Ukraine, conduct connected to President Trump’s first impeachment inquiry—GAO concluded that OMB’s withholding of DOD’s Ukraine funds was improper but that OMB’s withholding of State’s Ukraine funds was permissible. (The latter decision saw a reconsideration to correct a factual mistake but reached the same conclusion.) The difference was based on GAO’s interpretation not of the Impoundment Control Act but of the underlying authorities for the different pots of Ukraine money, even though OMB’s actions in reapportioning the funds during the same time period were identical. GAO determined that DOD’s appropriation provided no discretion to the agency in providing the funding, whereas the State Department funding did; the underlying actions were thus a permissible “programmatic delay” in State’s case. 

Both investigations into the Biden administration’s potential impoundment activity were driven by Republican inquiries into whether the administration was improperly slow-walking the execution of funds appropriated to build a barrier at the southern border. Building such a wall had been a Trump administration priority but was opposed by the Biden administration. In response to a proclamation President Biden issued on the first day of his term directing officials to pause both construction and obligation of funds, more than 100 Republican members of Congress in both chambers asked GAO to investigate potential impoundment. Using the same logic as in the State Ukraine decision, however, GAO concluded that there had been no improper impoundment because the underlying statute allowed discretionary space to make some decisions with respect to the funding, and there was no intent to thwart a congressional spending policy. A 2024 update based on a request by Republican leadership in the House Budget Committee produced the same result: GAO determined that despite the Biden administration’s clear policy preferences against the wall, there was no evidence of improper delay or withholding.

With the new administration challenging the Impoundment Control Act on both constitutional and policy grounds, this trend of GAO’s inquiry into impoundment issues is likely to continue.

We coded four different possible outcomes: Consistent with Law (that is, GAO said that what the agency did in the past or what it wanted to do in the future was permissible); Not Consistent with Law (GAO said that what the agency did in the past or what it wanted to do in the future was not permissible); Mixed (GAO said that some past or future conduct was permissible but some was not); and Unable to Issue a Decision (GAO was unable to obtain enough facts from the agency to issue a decision).

Of the 44 decisions evaluating agency conduct during the four Obama administration years in our dataset, 27 (61.4%) resulted in a finding that the agency’s action was Consistent with Law, while 17 (38.6%) resulted in a finding that the agency’s action was Not Consistent with Law. Of the 63 decisions evaluating agency conduct during the first Trump administration, 27 (42.9%) resulted in a finding that the agency’s action was Consistent with Law, while 30 (47.6%) resulted in a finding that the agency’s action was Not Consistent with Law. The difference in the Consistent/Not Consistent with Law rate between these two administrations is not statistically significant.

Of the 15 decisions evaluating agency conduct during the Biden administration in our dataset, however, 11 (73.3%) resulted in a finding that the agency’s action was Consistent with Law, and only two (13.3%) resulted in a finding that the agency’s action was Not Consistent with Law, a statistically significant difference (p = 0.03). This balance may well change as the number of decisions issued about Biden-era conduct increases over time.

As for mixed decisions, there were no such decisions evaluating agency conduct during the Obama administration, five for the Trump administration, and one for the Biden administration—although these variations were not statistically significant. And there were only two cases in which GAO could not issue a decision—one involving the Trump administration’s Interior Department, as mentioned above, and the second involving choices that the Treasury Department under Biden had yet to make.

There are no significant variations in outcome by administration depending on whether an agency or Congress made the request, perhaps due to sparse data.

However, the final notable finding is that decisions initiated by GAO itself are significantly more likely to result in a finding that the agency’s action was Not Consistent with Law (p = 0.037), controlling for presidential administration—and, as referenced earlier, all 11 such decisions in our dataset evaluated agency conduct during the Trump administration.

How should these findings be interpreted? After GAO’s initial conclusion that the Biden administration had not violated the Impoundment Control Act in its pause of wall spending, senior Republicans on the Senate Appropriations Committee issued a press release saying that “GAO’s decision today makes clear that there are two sets of rules when it comes to executing funds appropriated by Congress: one for Democrat administrations and one for Republican administrations.” Our data do not support this conclusion, in keeping with previous research on GAO’s work in other contexts.

In particular, there is no significant variation in outcomes during the Obama and Trump administrations. While the findings of comparatively fewer violations during the Biden administration are significant, these outcomes likely do not reflect the full complement of appropriations law decisions that GAO will ultimately issue about the Biden administration.

In addition, given the appropriations law conflicts during the Trump administration, agencies during the Biden administration might have been generally more risk averse. It may also be relevant that Republicans requested fewer decisions overall than Democrats did during the Trump era. Moreover, given the Trump administration’s stated goal of pushing the boundaries of appropriations law, as well as presiding over the lengthiest government shutdown in history, it is no wonder that GAO relied more frequently on its own caretaking responsibilities over the Impoundment Control Act and Antideficiency Act to investigate.

As for the underlying substance that led to the charge of bias, since the initial decision clearing the Biden administration of illegal impoundment and the decision upholding the Trump administration’s pause on State Department Ukraine funding rested on the same legal theory, the suggestion of disparate treatment is not persuasive. Indeed, on one of the other major appropriations law controversies during the Trump administration—whether the Department of Defense permissibly transferred $1.8 billion to build a wall at the southern border after Congress declined to provide the requested amount—GAO held that the agency had not violated the law despite opposition from Democrats.

Instead of bias, these data suggest that GAO’s analysis follows where the facts go in consistent application of basic appropriations law principles.

The new administration shows no sign of pulling back on its expansive arguments about the presidential spending power. As these plans continue to unfold, as courts evaluate the spending cases brought before them, and as Congress considers its own responses, GAO will be there, filling the role Congress has asked it to play in appropriations law oversight across matters both large and small.

For our part, we are continuing to review and code GAO’s appropriations law decisions through previous administrations as part of the previously mentioned book project on how agencies execute their budgets. We are particularly interested in tracking the extent to which requests for these decisions have varied over time; what might we learn about consistency or change in interbranch relations around appropriations through these analyses? For future work, it would also be valuable to compare GAO’s legal decisions on appropriations law with those evaluating executive branch conduct in other contexts, especially GAO’s determinations under the Congressional Review Act and the Federal Vacancies Reform Act. How does Congress use the GAO Office of General Counsel, which issues all of these decisions, to support its oversight of the executive branch more generally? Finally, we are paying attention to two potential means of judicial involvement that might change the nature of GAO legal decisions in different ways going forward: the possibility that GAO might turn to courts itself as a plaintiff under the Impoundment Control Act, as GAO recently noted the Act contemplates, and the possibility that courts might conclude that GAO’s statutory design is structurally impermissible under the Constitution, as a recent lawsuit alleges. Either one of these possibilities could significantly change how GAO conducts its business, not only with respect to appropriations law but also beyond, and could reorient interbranch relations in uncharted ways.

For now, however, we are keeping a close eye on GAO’s appropriations law website, watching as the first decisions on the new administration begin to be issued. Our analysis in this preliminary dataset will help make sense of those decisions as they appear.