Pork-barrel legislation has been criticized by some as an excessive and potentially corrupt use of Congressional appropriations. The task of finding the specific parts of legislation that have been “pork-barreled”, however, requires many hours of labor by policy researchers. Using data from government watchdogs and machine learning algorithms, the research explores the idea of creating a model to flag specific line items of appropriations bills for policy researchers to further explore as potential pork-barrel legislation. The model constructed uses data from the Earmark Database by Taxpayers for Common Sense, and the Congressional Pig Book by Citizens Against Government Waste to attempt to identify line items in appropriations bills as either “potential pork” or “not potential pork”. Criteria for the model are based upon Citizens Against Government Waste’s seven criteria for pork-barrel legislation as well as their identification of egregious examples of pork-barrel projects. The research focuses on the fiscal years of 2008-2010 when Congressional rules required the disclosure of earmarks. The first prototype of a machine learning model was trained on fiscal year 2008 data and showed limited effectiveness at differentiating pork and non-pork in the 2008 Consolidated Appropriations Act.
"Identifying Potential Pork-Barrel Legislation Using Machine Learning: A Preliminary Analysis,"
Dartmouth Undergraduate Journal of Politics, Economics and World Affairs: Vol. 1:
4, Article 7.
Available at: https://digitalcommons.dartmouth.edu/dujpew/vol1/iss4/7
Economics Commons, Education Commons, Geography Commons, Law Commons, Legal Studies Commons, Medicine and Health Sciences Commons, Political Science Commons, Public Affairs, Public Policy and Public Administration Commons