| Fundamental to many accounts of decision-making within political institutions is the interdependence between simultaneous choices. For instance, members in a legislature are hypothesized to take voting cues from party leaders, and the chief justice of the U.S. Supreme Court is thought to vote with the majority on the merits so as to assign opinion authorship. In this thesis I show that none of the conventional methods that have been used by political scientists for testing theories of simultaneous interdependence are statistically sound. I then propose a machine-learning algorithm that finds unmodeled interdependence in discrete-choice data. Next, I develop a novel statistical model that allows the researcher to explain -- in a methodologically appropriate manner -- the probability that an actor makes a particular choice as well as the probability that a collective-decision occurs in a particular form. In the last chapter of my dissertation, I demonstrate that U.S. Supreme Court case outcomes are interdependent and that the U.S. Supreme Court is best characterized as an institution striving to produce an ideologically optimal body of law rather than ideologically optimal independent case outcomes. |
