In 1997, on average every 40 to 45 hours criminals either shot or killed a victim in the City of Richmond, Virginia. This resulted in 122 firearm homicides in that year alone. This gun-related violent crime epidemic so terrorized law-abiding citizens that many of them became hostages in their own homes. In response to this horrific social problem, Project Exile was developed in late 1997. Project Exile is a multi-level (federal, state, and local) law enforcement effort aimed at the amelioration of Richmond's high per-capita rate of gun violence and gun homicide. Through the Richmond U.S. Attorney's Office, Project Exile takes advantage of stiffer bond rules and sentencing guidelines in federal court, where all cases involving felons with guns, guns and drugs, and guns and domestic violence are prosecuted. Although Project Exile has received extensive television and print media coverage, it has not caught the attention of the academic world. This dissertation begins to fill this research gap by combining Kingdon's (1995) Multiple Streams model with Sabatier's (1999) Advocacy Coalition Framework to develop a "Specific Collins Classification and Elaboration Model" and a "Generic Collins Classification and Elaboration Model" that will be used to analyze the formation and implementation of Project Exile. The three purposes of this research will be:

1. To elaborate and analyze Kingdon's and Sabatier's frameworks as a means for understanding Project Exile
2. To draw on these two frameworks to create both Specific and Generic new "Collins Models: to assist in furthering a deeper understanding of this case study as well as similar policy subsystems.
3. To explain the genesis and development of Project Exile.

The most salient result of this research is that it shows the disparate ways in which variables, taken from the work of Kingdon, Sabatier, and the Project Exile case, fit in Schroeder's (2001) operationalization of the Political Economic framework. In addition, this research shows how both Kingdon and Sabatier compensate for the respective limitations of the other when the two of them are combined into one model.