LSE Research Proposal Example: Benefits adviser to social security reform (Score 93)

The United Kingdom's social security system has undergone sustained structural reform since 2012, most consequentially through the rollout of Universal Credit (UC). Policy documents and ministerial statements consistently frame UC as a simplification that improves work incentives and reduces administrative error. Yet frontline evidence — gathered through advice services, tribunal records, and administrative data releases — points to a persistent and poorly explained divergence between the stated objectives of reform and the outcomes experienced by claimants, particularly those with health conditions or caring responsibilities. This divergence is not merely anecdotal. It raises a tractable research question about the mechanisms through which implementation translates — or fails to translate — policy design into the outcomes that reform was intended to produce. This proposal asks: To what extent does street-level discretion in Universal Credit conditionality decisions explain variation in claimant outcomes across jobcentre districts in England and Wales? Two subsidiary questions follow. First, which claimant characteristics — including health status, household composition, and prior benefit history — are most strongly associated with adverse conditionality decisions at the district level? Second, does variation in local labour market conditions moderate the relationship between conditionality intensity and employment outcomes, or does it operate independently? The rationale for focusing on conditionality decisions specifically, rather than UC as a whole, is that conditionality is the mechanism through which adviser discretion is most directly exercised. Work coaches set claimant commitments, apply sanctions, and grant deferrals within a framework that grants substantial local latitude. This makes conditionality decisions a tractable unit of analysis: they are partially observable through administrative data, they vary across districts, and they have measurable downstream consequences for claimant income and employment status. Two bodies of scholarship are directly relevant. The first is the street-level bureaucracy literature, originating with Lipsky's account of how frontline workers exercise discretion under resource constraint. Subsequent work — including studies of welfare-to-work programmes in the United States, Australia, and the Netherlands — has extended this framework to activation regimes, showing that worker discretion shapes outcomes in ways that aggregate policy evaluation often obscures. Within the UK context, qualitative studies of jobcentre practice have documented how work coaches navigate competing institutional pressures, but these studies are typically small-scale and cannot establish whether discretion patterns are systematic or idiosyncratic. The second literature concerns the employment and income effects of benefit conditionality. Quantitative evaluations — including DWP-commissioned analyses and independent work using administrative data — have examined sanction rates and employment transitions, but most treat conditionality as a uniform treatment rather than a variable one. The identification strategies used in this literature frequently rely on national-level variation in policy timing, which limits the ability to isolate the discretionary component of conditionality from its rule-based component. The gap between these two literatures is methodological as much as substantive. Qualitative work captures the texture of discretion but cannot scale; quantitative work achieves scale but collapses the variation that discretion produces. A district-level analysis that uses administrative data to operationalise conditionality intensity — measured through sanction rates, deferral rates, and commitment revision frequency — while controlling for local labour market conditions, could bridge this gap. To my knowledge, no published study has used district-level administrative variation in UC conditionality decisions as the primary source of identification for estimating claimant outcome effects. This is the gap the proposed research addresses. The study uses a quantitative cross-sectional design with district as the unit of analysis, supplemented by a small number of semi-structured interviews to interpret anomalous district-level patterns. The primary data source is DWP's Stat-Xplore platform, which publishes UC caseload statistics, sanction rates, and claimant characteristic breakdowns at jobcentre district level. These are publicly available and require no special access agreement for aggregate analysis. Local labour market data — vacancy rates, employment rates by occupation, and earnings distributions — will be drawn from the Office for National Statistics Annual Population Survey and the NOMIS labour market statistics portal, both of which are freely accessible. The dependent variables are district-level employment transition rates and income adequacy indicators derived from UC payment data. The primary independent variable is a conditionality intensity index constructed from sanction rates, deferral rates, and the share of claimants subject to the most demanding commitment regimes, standardised within district and year. Control variables include claimant health status composition, household type distribution, local unemployment rate, and vacancy-to-claimant ratio. The identification strategy exploits cross-district variation in conditionality intensity after controlling for observable claimant composition and labour market conditions. This is not a causal design in the experimental sense; the study does not claim to identify the causal effect of conditionality on employment. Rather, it aims to establish whether district-level variation in conditionality intensity is systematically associated with outcome variation in ways that cannot be explained by claimant composition or local economic conditions alone. If such residual variation exists, it constitutes evidence that implementation discretion matters independently of policy design — a finding with direct implications for reform evaluation. The interview component — approximately eight to twelve work coaches and welfare rights advisers, recruited through Citizens Advice and Jobcentre Plus public contact channels — serves an interpretive rather than inferential function. It will be used to assess whether the conditionality intensity index captures what practitioners recognise as meaningful variation in local practice, and to identify any systematic data limitations that the quantitative analysis cannot detect. The data sources described above are publicly available at aggregate level, which means the primary analysis does not require ethics approval for data access. The interview component will require institutional ethics review, which I plan to initiate in the first term of the programme. Participants will be recruited through public-facing organisational channels; no claimant data will be collected, and no individual-level administrative records will be requested. The main data risk is that Stat-Xplore district-level breakdowns may suppress cells with small counts, which could limit analysis in rural or low-caseload districts. A contingency is to aggregate to a coarser geography — for example, DWP contract package areas — if cell suppression is too extensive at district level. The timeline is structured across three phases. In the first term, I will complete the literature review, construct the conditionality intensity index from available Stat-Xplore data, and submit the ethics application for the interview component. In the second term, I will run the regression analysis, conduct interviews, and produce a first draft of findings. In the third term, I will revise the analysis in response to feedback, complete the write-up, and prepare the dissertation for submission. This is a realistic scope for a one-year MSc: the quantitative component uses pre-existing public data, the interview component is modest in scale, and the research question is bounded to a single policy mechanism in a single national context. LSE's Department of Social Policy and the School of Public Policy together host research on welfare state reform, activation policy, and public administration that is directly relevant to this project. The quantitative methods training offered within the MSc Public Policy programme — including courses in applied regression and policy evaluation — provides the technical foundation the analysis requires. The School's access to NOMIS, the UK Data Service, and DWP research publications through the LSE library means that the data infrastructure for this project is available without requiring external institutional agreements. The research sits at the intersection of welfare state analysis and implementation studies, two areas in which LSE has sustained research activity. I am particularly interested in engaging with work on conditionality and activation that has been produced within the Department of Social Policy, and in using the methods training available in the programme to develop the quantitative skills the district-level analysis demands. The project is designed to be completable within the MSc timeframe, to use data that is already publicly accessible, and to produce findings that are directly relevant to ongoing policy debates about UC reform — without overstating what a single dissertation study can establish.