Machine learning methods for studying the trajectories of young offenders in administrative data

行政数据有可能为更好地理解社会现象和支持循证决策提供新的和宝贵的研究机会。行政数据可以显著增强的一个研究领域是分析社会科学家感兴趣的关键领域的人生轨迹，包括教育、经济活动、卫生和犯罪。行政数据是关于关键社会经济成果和公共服务使用情况的丰富纵向信息来源，这对与政策有关的分析特别有用。然而，在英国，到目前为止，它还相对未被开发。随着研究人员越来越容易获得行政数据和与行政有关的数据，学术界或政府的研究小组必须配备适当的方法学工具包，以充分利用这些数据提供的研究可能性。在这次竞标中，财政研究所(IFS)建议与司法部(MoJ)合作应对这一挑战：我们提出了一系列活动，以推进机器学习技术的使用，以利用司法部管理和链接的丰富行政数据集，作为关于个人犯罪和教育轨迹及其通过司法系统的旅程的信息来源。这些活动包括：a)通过首次在英国行政和行政关联数据中应用序列分析方法，调查罪犯轨迹的性质和驱动因素以及司法系统如何塑造这些轨迹的创新研究。B)向司法部借调一名独立研究人员，以确保研究的共同成果和政策相关性，加强司法部的分析能力，并加强两个机构之间的思想和知识交流。C)由一名序列分析方法国际专家领导的培训讲习班，以在研究机构和政府内部建立利用机器学习技术进行行政数据分析的研究能力。D)在赠款结束时为财政研究所和其他机构的研究人员以及司法部和其他有关部门的分析员和政策制定者举办研究讲习班，介绍和讨论在行政数据中使用机器学习方法进行的当前研究。该项目团队由伦敦大学学院经济学教授、IFS研究总监伊姆兰·拉苏尔教授领导，在基于英国和国际行政数据的学术和政策工作方面有着良好的记录。通过在英国行政相关数据中建立这些方法的可行性，并证明它们在通过公共服务系统更好地了解个人旅程方面的有用性，这项研究将对社会科学领域进行的生命历程研究做出重大贡献。这项研究将产生至少两项学术成果，并附有非技术性的、以政策为导向的简报。该小组与司法部的合作，以及IFS与政策制定者和媒体的良好联系，将确保这些成果在学术界之外得到很好的传播，并用于为决策过程提供信息。

Administrative data has the potential to open new and invaluable research opportunities to better understand societal phenomena and support evidence-based policy-making. One research area administrative data can significantly enhance is the analysis of life-course trajectories across key domains of interest to social scientists, including education, economic activity, health, and crime. Administrative data is a rich source of longitudinal information on key socio-economic outcomes and public services use that can be particularly useful for policy-relevant analysis. Yet, in the UK, it has been relatively untapped so far. As administrative data and administrative linked data becomes increasingly available to researchers, it is essential that research teams, both in academia or in government, are equipped with the appropriate methodological tool kit to take full advantage of the research possibilities these data offer. In this bid, the Institute for Fiscal Studies (IFS) proposes to collaborate with the Ministry of Justice (MoJ) to address this challenge: we propose a package of activities to advance the use of machine learning techniques to exploit the richness of MoJ's administrative and linked administrative datasets as sources of information on individuals' offending and educational trajectories and their journeys through the justice system. These activities include: a) Innovative research investigating the nature and drivers of offenders' trajectories and how these are shaped by the justice system, by applying, for the first time, sequence analysis methods in UK administrative and administrative linked data. b) A secondment of an IFS researcher to the MoJ to ensure the co-production and policy relevance of the research, enhance MoJ's analytical capability, and strengthen exchange of ideas and knowledge between the two institutions. c) A training workshop led by an international expert in sequence analysis methods to build research capability, within research institutions and government, in the use of machine learning techniques for administrative data analysis. d) A research workshop at the end of the grant for researchers from the IFS and other institutions, as well as analysts and policy-makers from MoJ and other relevant departments, to present and discuss current research using machine learning methods in administrative data. The project team, led by Professor Imran Rasul, Professor of Economics at University College London and IFS' Research Director, has a strong track record of academic and policy work based on UK and international administrative data. By establishing the viability of these methods in UK administrative linked data and demonstrating their usefulness to better understand individuals' journeys through public service systems, the research will make significant contributions to life-course studies conducted across the social sciences. The research will produce at least two academic outputs, accompanied by non-technical, policy-oriented briefs. The team's collaboration with MoJ, as well as the IFS' well-established links with policy-makers and the media, will ensure that these outputs are well disseminated beyond academia and used to inform the policy-making process.