Doctoral Dissertation Research in Economics: Individual, Family and Community Impacts of Criminal Sentencing

在过去的30年里，美国的监禁人口增长了近300%。司法统计局报告称，2009年，美国有720万成年人（即每31名成年人中就有1人）处于“惩教监督”之下，这种状态除了监禁之外还包括缓刑和假释。2011财政年度，惩教活动的广泛影响相当于大约517亿美元，占州预算的7.4%。尽管美国的这一社会计划规模庞大，但几乎没有可靠的证据可以量化监禁的成本和收益，并证明其广泛使用是合理的。本项目旨在通过使用美国的新数据估计刑事判决对个人、家庭和社区的影响，帮助解决这一问题。研究人员的原始数据收集工作创建了一个新的经验数据集，包括德克萨斯州哈里斯县30年的刑事法庭案件。它包括270多万个不同的案件和110万个独特的被告。法庭记录数据可以随时间推移跟踪被告，使研究能够考虑对累犯和犯罪行为升级的影响。此外，与当地县机构的独特数据共享协议将允许刑事法庭数据与有关被告的劳动力市场活动，婚姻行为，家庭组建，子女教育程度和邻里逮捕记录的行政记录相关联。将刑事判决与各种生命周期行为和代际结果联系起来将产生一个前所未有的数据集。为了克服遗漏变量偏差的挑战，该研究将利用一个可能的外生变异来源：将刑事法庭案件随机分配给法官。在对被告提出刑事指控时，案件被随机分配到审判室，以保持法官之间案件量的平衡。哈里斯县刑事法院系统的法官在审判室的日常运作以及对被判有罪的被告的确切判决方面都有很大的自由裁量权。由于随机化和自由裁量权的结合，对被告的量刑结果的一些变化是由于随机的机会。本研究将寻求隔离的程度，量刑的变化，以使因果关系statements关于刑事sentencing.To推进这一文献的方法，研究人员将进一步完善他们的新的估计技术：异质反应工具变量（HRIV）的影响。HRIV模型，它非参数估计法官的决策规则过程，可以减少偏见，提高效率，在实证application.At本研究的结论，刑事法庭的数据将存档长期保存与大学间政治和社会研究联盟（ICPSR）。还将努力为今后的工作提供链接数据的匿名版本，也是通过ICPSR。

Over the past 30 years, the incarcerated population in the United States has grown by close to 300 percent. The Bureau of Justice Statistics reports that in 2009, 7.2 million adults in the United States, or 1 out of every 31 adults, were under "correctional supervision," a status that includes probation and parole in addition to incarceration. The wide reach of corrections activity translates into roughly $51.7 billion dollars or 7.4 percent of state budgets in fiscal year 2011. Despite the size of this social program in the United States, little reliable evidence exists to quantify the costs and benefits of incarceration and justify its widespread use.This project seeks to help address this problem through estimating the impacts of criminal sentencing at the individual, family and community level using new data from the United States. Original data collection efforts by the researchers have created a new empirical dataset comprising 30 years of criminal court cases from Harris County, Texas. It includes over 2.7 million distinct cases and 1.1 million unique defendants. Defendants can be followed over time in the court record data, allowing the research to consider affects on recidivism and escalation in criminal behavior.In addition, unique data sharing agreements with local county agencies will allow the criminal court data to be linked with administrative records concerning defendants' labor market activity, marital behavior, family formation, children's educational attainment, and neighborhood arrest records. Linking criminal sentencing to a variety of lifecycle behavior and intergenerational outcomes will generate an unprecedented dataset.To surmount the challenge of omitted variables bias, the research will leverage a source of plausibly exogenous variation: random assignment of criminal court cases to judges. As criminal charges are filed against defendants, cases are randomly assigned to courtrooms in an effort to maintain balanced caseloads between the judges. The judges in the Harris County criminal court system then hold significant discretion in both the day-to-day operations of their courtrooms as well as the precise sentences delivered to defendants found guilty of a crime. Due to the combination of randomization and discretion, some variation in the sentencing outcome for defendants is due to random chance. This research will seek to isolate that degree of sentencing variation in order to make causal statements regarding the impact of criminal sentencing.To advance this literature's methodology, the researchers will further refine their novel estimation technique: heterogeneous response instrumental variables (HRIV). The HRIV model, which non-parametrically estimates the judge's decision rule process, can reduce bias and improve efficiency in empirical applications.At the conclusion of this research, the criminal court data will be archived for long-term preservation with the Inter-university Consortium for Political and Social Research (ICPSR). Efforts will also be made to make an anonymized version of the linked data available for future work also through ICPSR.