Who Walks the Watchdog? Bureaucratic Oversight and the Government Accountability Office

Abstract

Created in 1921 and strengthened after Watergate, the Government Accountability Office (GAO) is Congress's primary watchdog agency over the administrative state. The GAO investigates, culls, and synthesizes stances on a multitude of policy programs and expenditures on its own initiative, by legislative mandate, and at the request of congressional committees and individual members of Congress. This paper examines how members of Congress use the GAO to advance their own policy preferences and how the GAO makes strategic auditing decisions on its own. From a simple model supplemented by interviews of GAO officials, I generate several testable empirical propositions. First, in periods of divided government, members and committee chairs should request more GAO investigations; in periods of united government, members of the minority party should request more GAO studies. Second, members of Congress may use the GAO to increase their chances of being reelected. In particular, Senators facing reelection sooner should request more studies than Senators facing reelection in a later cycle. Third, if the GAO is concerned about how it is perceived by its sponsors in Congress, it should make fewer recommendations when its work is not congressionally requested. I test these propositions using information from the GAO Documents Database, which I constructed from details on all published GAO reports and testimony to Congress, for 1986-1997. I find some support for the institutional theory of how Congress uses the GAO. House committee chairpersons are more likely to request a GAO investigation when there is divided government. A change from united to divided government corresponds to a 5.68 percent increase in the probability that a House committee chairperson will join a request. House committee ranking minority members are less likely to turn to the GAO when there is divided government. A shift from united to divided government leads to a 2.16 percent decrease in the probability that a House committee ranking minority member will sign onto a request. It also appears that Senators facing the voters in the next election request more studies than other Senators. Finally, the GAO seems to issue recommendations carefully. It makes more recommendations to Congress when a study is requested but makes more recommendations to agencies when a study is self-initiated, perhaps anticipating that Congress will call for agency changes when an investigation is requested.