Always Watchful: Political Context and Congressional Oversight through the Government Accountability Office

Abstract

Oversight of the executive branch is one of Congress’s most critical functions. It is meant to ensure accountability, consistency, and fairness in the implementation of programs by the federal bureaucracy. Much previous scholarship has argued that oversight is more common when Congress and the White House are controlled by different parties. However, much of this research has focused principally on committee hearings, and other tools of oversight have received less attention. One such tool is the Government Accountability Office (GAO). I test competing theories of GAO usage: one which argues that GAO reports are simply another tool of oversight, like hearings, and tend to follow similar patterns and thus are more common in conditions of divided government. The other theory argues that the GAO, because of its commitment to providing useful, nonpartisan information, is more likely to be used when legislative and presidential priorities are aligned. Leveraging reports from the GAO between 1995 and 2022, I find support for the latter proposition, at least in the modern Congress. There is evidence that Congress changes its methods of oversight in response to political incentives, but oversight continues regardless.

Acknowledgements

The author wishes to acknowledge Adam Cayton, Brett Curry, Sara Hagedorn, Jason MacDonald, Karen McCurdy, Robert McGrath, Josh Ryan, Jennifer Selin, Laine Shay, and the anonymous reviewers for their gracious assistance and feedback.

Notes

1 According to the GAO (2004, 5) itself, work is prioritized according to the source from which the request originates. In descending order, these sources are, “Congressional mandates,” “Requests from senior congressional leaders and committee/subcommittee Chairs and/or Ranking Minority Members of committees of jurisdiction,” and “Requests from individual Members, with additional consideration given to requests from Members who are on committees of jurisdiction.”

2 “Scant” is not “none,” of course. Frederick C. Mosher (1979) offers a comprehensive review of the GAO’s early history and functions, Aberbach (1990) addresses the GAO and other kinds of support agencies, and more recently, Bridget C. E. Dooling (2020) calls for more attention to the GAO, and other legislative agencies, to gain a better grasp of the inner workings of the administrative state.

3 The GAO can and does initiate some investigations on its own, reporting the results to Congress after the fact.

4 These data were initially collected in the spring of 2022 and re-collected in the spring of 2023 to remove written GAO testimony to ensure no overlap with committee hearings.

5 These agencies are: the Departments of the Army, Commerce, Homeland Security, Defense, Energy, Education, Justice, Labor, Transportation, the Treasury, Veterans’ Affairs, Health and Human Services, Housing and Urban Development, the Interior, the Navy, State, the Air Force, and Agriculture; and the Environmental Protection Agency, General Services Administration, Nuclear Regulatory Commission, Social Security Administration, U.S. Agency for International Development, Export-Import Bank of the United States, Federal Communications Commission, Federal Deposit Insurance Corporation, Federal Emergency Management Agency/Administration, Federal Trade Commission, Office of Management and Budget, Securities and Exchange Commission, and Small Business Administration.

6 The Department of Homeland Security is not included for the years of 1995–2002.

7 An agency is coded as “Liberal” (–1) if its ideology estimate is statistically distinguishable from 0 in a negative direction, as “Moderate” (0) if its ideology estimate is statistically indistinguishable from 0, and as “Conservative (+1) if its ideology estimate is statistically distinguishable from 0 in a positive direction.

8 See Cameron, Gelbach, and Miller (2011).

9 Importantly, this variable is not a perfect recreation of the variable used in Model 3 because it includes both reports and written testimonies, not reports only. This is a limitation of the GAO website’s search function. However, for illustrative purposes, between 1995 and 2022 measuring at the agency-level, a measure that includes both written testimonies and reports and a measure that includes only reports are correlated at 0.9934.

10 This includes the 31 agencies in the principal analysis plus the National Aeronautics and Space Administration, for which I have data on GAO reports but not agency ideology (hence its exclusion from preceding models).