Administrators, Legislators, and Presidents: An Analysis of Roll Call Votes.

The spatial theory of political control posits that bureaucratic drift results from asymmetric information gathered by agencies, veto opportunities for political branches, agency policy preferences, and transactions costs (e.g., Weingast and Moran 1983; Calvert, McCubbins, and Weingast 1990; McCubbins, Noll, Weingast 1987, 1989; Banks and Weingast 1992). The canonical model formally represents interbranch bargaining among the president and houses of Congress. To this model, we add a quasi-autonomous administrative agent—which shares policy setting authority with an interest group. The agent and group must bargain for a policy choice in the final round of the political control game, rather than simply setting policy as in the canonical model. This model uncovers the incentive for the president to strategically choose an agent to enhance his welfare from policy outcomes that occurs quite apart from any informational or transactions cost sources. We compare preexisting models and our spatial bargaining model to generate competing empirical implications. Testing these spatial theories requires ideal point estimates for the House, Senate, and president, which have been available (Poole and Rosenthal 1997; Clinton, Jackman, and Rivers 2003), and administrative agencies, which have not. To empirically test these implications, we estimate ideal points and the distances between the relevant actors (the president, administrative agency heads and regulatory commission members, and members of Congress) in the competing models. Using Markov Chain Monte Carlo (MCMC) ideal point estimation (Clinton, Jackman, and Rivers 2003), we are able to compare the distances between the ideal points of these actors on a comparable scale. The ideal point estimates presented in this paper are improvements over past work in the area for three reasons: (1) ideal points for administrative agents are estimated; (2) the MCMC method allows us to estimate ideal points over time for presidents, which are only available as NOMINATE scores for a president’s entire tenure in office; and (3) directly comparable ideal point estimates are available for members of Congress, presidents, and administrative agents. Thus, not only are these estimates ideal for testing political control models, but are likely to be useful for scholars testing other theories. [ABSTRACT FROM AUTHOR]