The Power of Policy Ideas: Tracing Language in Policymaking

拟议项目的目的是启动对政策理念的系统研究。法案往往包含大量的政策条款，并在审议过程中不断演变。这个项目解决了成功的法案在多大程度上纳入了最初由立法者而不是法案发起人提出的政策想法。当以政策想法的进展而不是法案来评估政策影响时，政策的影响是不是要广泛得多？哪些个人和制度因素促进或抑制了政策思想的纳入？外部重点活动或内部时间表(如重新授权)如何塑造推进政策想法的机会？该项目的智力优势与其全面的数据收集和分析、跨学科组成部分以及对关键政策问题的关注有关。为了研究这些问题，该项目使用“大数据”方法来审查1993-2011年期间提出的大约8万项参众两院法案的实质内容。在使用为研究计算机科学中的“文本重用”而开发的算法之前，每个法案都按部分(约800,000个部分)进行划分或分析，然后评估每个部分之间的相似性。一个潜在的假设是，不同法案中类似的部分构成了相同的政策理念。这使得研究人员能够跟踪一个想法的进展，而不是特定法案的进展。验证将是该项目的中心重点。该项目将聘请计算机科学家测试评估截面相似性的替代方法，以更好地了解它们对第一类和第二类错误的影响。预计对政策思想的研究将导致对与主流立法研究整体相关的三个主题的重要见解：政策影响、立法议程设置和比较委员会程序。该项目的更广泛影响与它将产生的数据直接相关。该语料库将成为未来许多领域的研究人员和学生的宝贵资源。通过将复杂且往往庞大的法律和法案划分为一种可研究的格式，学者们将能够系统地窥探法案的“幕后”，从而获得关于政策制定过程的有价值的新观点。例如，该数据集将为对立法网络感兴趣的学者提供宝贵的资源，如政策子系统、政策代表、摸索、游说影响和许多其他主题。此外，计算机和信息科学家将把该语料库作为自然语言处理研究的试验台。最后，任何公民或记者都将能够利用我们的可搜索数据库来追踪特定立法者提出的政策想法的实质和进展。这样的搜索将产生对优先事项和政策影响力的更准确评估。

The purpose of the proposed project is to initiate the systematic study of policy ideas. Bills often contain a multitude of policy provisions and evolve during their consideration. This project addresses the extent to which successful bills incorporate policy ideas originally proposed by lawmakers other than the bill's sponsor. Is policy influence substantially more widespread when assessed in terms of the progress of policy ideas rather than bills? What personal and institutional factors promote or inhibit the incorporation of policy ideas? How do external focusing events or internal schedules such as reauthorizations shape opportunities to advance policy ideas?The project's intellectual merit is associated with its comprehensive data collection and analysis, interdisciplinary component, and attention to critical policy questions. To study these questions, the project uses a "big data" approach to examine the substance of the approximately 80,000 House and Senate bills introduced during the 1993-2011 time period. Each bill is divided, or parsed, by section (approximately 800,000 sections), before assessing the similarity of each section to each other section using algorithms developed to study "text reuse" in computer science. An underlying assumption is that similar sections found in different bills constitute the same policy idea. This allows the researcher to follow the progress of an idea as distinct from the progress of a particular bill. Validation will be a central focus of the project. The project will engage computer scientists to test alternative approaches to assessing section similarity to better understand their implications for Type I and Type II errors. It is expected that a study of policy ideas will lead to important insights into three topics integrally related to mainstream legislative research: policy influence, legislative agenda setting, and comparative committee processes.The broader impacts of the project are directly related to the data it will produce. The corpus will be a valued resource for future researchers and students in many fields. By partitioning complex and often large laws and bills into a researchable format, scholars will be able to systematically peer "under the hoods" of bills and therefore gain valuable new perspectives on the policymaking process. For example, the dataset will provide a valuable resource for scholars interested in legislative networks such as policy subsystems, policy representation, logrolling, lobbying influence, and many other topics. In addition, computer and information scientists will value the corpus as a test bed for natural language processing research. Finally, any citizen or reporter will be able to utilize our searchable database to trace the substance and progress of policy ideas proposed by a given lawmaker. Such a search will yield a more accurate assessment of priorities and policy influence.