LSE Research Proposal Example: Criminology student to crime and justice policy (Score 93)

England and Wales imprison a higher proportion of their population than almost any other Western European state, yet reconviction rates within two years of release have remained stubbornly above twenty-five per cent for the past decade. This pattern raises a tractable policy question that sits at the intersection of criminological evidence and public expenditure: under what conditions does short-term custodial sentencing reduce reoffending compared with community-based diversion alternatives, and what features of programme design explain variation in outcomes across local criminal justice areas? Three more specific research questions organise this proposal. First, does the available administrative evidence support a causal claim that short custodial sentences of twelve months or fewer reduce reoffending relative to community orders of equivalent tariff severity? Second, to what extent do observable differences in diversion programme design — specifically intensity of supervision, employment support, and housing provision — account for variation in two-year reconviction rates across probation regions? Third, what does the policy process literature tell us about why evidence-based diversion programmes have been adopted unevenly across local areas despite a shared national framework? The practical stakes are not trivial. The Ministry of Justice spends approximately forty-seven thousand pounds per prison place per year. If community alternatives of equivalent or lower cost produce comparable or better reoffending outcomes for the short-sentence cohort, the case for reallocation is empirically grounded rather than ideological. My undergraduate independent research work on crime and justice policy identified this cost-effectiveness gap as under-examined in the UK context, and the present proposal develops that observation into a structured research design. Two bodies of scholarship bear directly on this question, and they have not been adequately integrated. The criminological deterrence and desistance literature — associated with life-course criminology in the tradition of Sampson and Laub, and more recently with systematic reviews of custodial versus non-custodial sentencing effects — has produced mixed findings. Several quasi-experimental studies exploiting sentencing threshold discontinuities in Scandinavian and US jurisdictions find small or null effects of short custody on reoffending, and some find marginally worse outcomes relative to community supervision. The UK-specific evidence is thinner. The Ministry of Justice publishes annual proven reoffending statistics disaggregated by disposal type, but these are observational: offenders receiving custodial sentences differ systematically from those receiving community orders on prior record, offence seriousness, and social vulnerability, making raw comparisons unreliable. Regression discontinuity designs exploiting the twelve-month sentencing threshold have been attempted in the literature, but existing analyses do not control adequately for local area fixed effects. The public policy implementation literature offers a complementary but largely separate account. Scholars working on street-level bureaucracy and policy translation — from Lipsky's foundational framework through to more recent work on probation reform in England and Wales following the Transforming Rehabilitation restructuring — document how national policy frameworks produce divergent local outcomes depending on organisational capacity, resource allocation, and practitioner discretion. This literature rarely engages with reoffending outcome data directly; it explains process variation without connecting it to measurable effectiveness. The gap this proposal addresses is the absence of a study that combines a credible quasi-experimental identification strategy on the sentencing threshold with a structured comparative analysis of programme design features across probation regions. Connecting these two traditions is both analytically feasible with available administrative data and directly relevant to current policy debates about the future of the Probation Service following its 2021 reunification. The proposal proceeds in two analytical phases, each matched to one of the first two research questions; the third question is addressed through a targeted document analysis that does not require primary data collection. Phase one uses a regression discontinuity design applied to Ministry of Justice proven reoffending data, exploiting the twelve-month custody threshold as a quasi-random assignment mechanism. Offenders sentenced just below and just above twelve months are, on average, more similar to one another than offenders at the extremes of the distribution, which provides a plausible identification assumption. The outcome variable is two-year proven reoffending and the running variable is sentence length in months. I will estimate local linear regressions on either side of the threshold with a bandwidth selected by the Imbens-Kalyanaraman procedure, and test robustness to alternative bandwidths and polynomial specifications. The Ministry of Justice publishes individual-level reoffending datasets under a data-sharing agreement available to academic researchers; I will apply for access through the LSE's existing data governance arrangements. Phase two uses a structured comparative case analysis across six to eight probation regions, selected to maximise variation on two-year reconviction rates while holding approximate offender composition constant using propensity-score matching on observable characteristics. For each region I will code programme design features — supervision intensity, employment support provision, and housing referral pathways — from publicly available inspection reports published by His Majesty's Inspectorate of Probation, supplemented where possible by freedom of information requests for programme delivery data. The analysis will use qualitative comparative analysis to identify which combinations of design features are consistently associated with lower reconviction rates, rather than assuming linear additive effects. Phase three conducts a structured document analysis of national policy documents, spending review submissions, and parliamentary committee evidence from 2010 to 2024 to map the stated rationale for diversion policy and identify where the evidence base was invoked, contested, or absent. The Ministry of Justice proven reoffending dataset is publicly documented and has been used in peer-reviewed research; access requires a data-sharing application rather than primary data collection, which is the main administrative risk. If access is delayed or restricted, the regression discontinuity analysis can be conducted on published aggregate tables disaggregated by sentence band, with reduced statistical precision — a contingency I have planned for explicitly. The qualitative comparative analysis phase relies on HMIP inspection reports, which are publicly available, and freedom of information requests, which carry a response-time risk; I have identified twelve months as the outer boundary for data assembly. The research does not involve human participants, interviews, or sensitive personal data beyond what is held in administrative records under existing legal gateways. I will register the project with the LSE Research Ethics Committee and follow its guidance on secondary data use and data storage. The provisional timeline runs to eighteen months: months one to four cover literature consolidation and data access application; months five to nine cover phase one analysis and drafting; months ten to fourteen cover phase two data coding and qualitative comparative analysis; months fifteen to eighteen cover phase three document analysis, integration, and write-up. This is calibrated to an MSc dissertation schedule with the understanding that the regression discontinuity phase is the most time-sensitive component. The LSE Department of Social Policy and the School of Public Policy both host research on criminal justice policy, welfare state design, and evidence-based governance. The methods training available through the LSE's quantitative social science pathway — specifically regression discontinuity and programme evaluation modules — maps directly onto phase one of this design, and the qualitative comparative analysis component is supported by the comparative politics and policy methods offerings within the School. My undergraduate crime and justice analysis work and the applied public policy project I completed as part of my degree have given me working familiarity with Ministry of Justice statistical publications and HMIP report structures. I am not claiming expertise in causal inference beyond what I have described; the MSc year is intended in part to consolidate the econometric methods required for phase one, and I have identified this as a training need rather than a completed competence. I would welcome the opportunity to discuss supervisory alignment with faculty whose published work addresses sentencing policy, probation reform, or quasi-experimental evaluation of criminal justice interventions.