Why do House members delegate authority to party leaders? This dissertation provides a new theory to address this question. It explains why it is individually and collectively rational for congressional members to delegate authority to party leaders, when we ought to expect that delegation will occur, as well as what form of delegation provides maximum benefit with minimal risk to members. This dissertation provides a new principal-agent theory of variable leadership involvement, progressive in nature across the stages of legislation as the risks of defection diminish. Members minimize the risks of delegation though screening and selection mechanisms, as well as through institutional checks throughout the legislative process. This work empirically tests and finds support for the predictions derived from the agency theory at four distinct stages of the legislative process. Leadership selection, bill introduction and referral, party scheduling with the Rules Committee, and conference committees are all found to conform to the hypotheses generated by this framework. The powers of the leadership are cumulative in nature as one passes through the stages of legislation. By severely curtailing the powers of leaders at earlier stages, the risk of adverse results is greatly minimized, allowing the members to safely delegate more authority to save the exponentially increasing transaction costs borne by leaders in the later stages.