Oxford Recommendation Letter Example: Development programme manager to social impact strategy (Score 92)

I am writing in support of the applicant's application to the MSc in Social Data Science at Oxford. I taught the research methods component of our public policy and social impact programme and supervised the applicant's independent research paper over one semester, during which I reviewed weekly methodology memos and met regularly to discuss analytical choices. My comments below are based on what I directly observed in that work. The applicant came to the independent paper with a background in development programme management, and the initial challenge was a familiar one: translating operational experience into a research question that could bear analytical weight. The first two memos were descriptive — essentially programme narratives reframed as research problems. I pushed back on this in our second meeting, asking the applicant to identify the causal mechanism the paper was actually testing and to distinguish it from a monitoring and evaluation summary. What followed over the next fortnight was, frankly, more rigorous than I expected at that stage. The applicant returned with a revised framing that separated the outcome variable from the implementation proxy, drew on a difference-in-differences logic to motivate the comparison group, and flagged, unprompted, that the available administrative data would not support a clean parallel-trends assumption. That last point mattered to me. It is easy to apply a method; it is harder to articulate why the method is limited by the data you actually have. The applicant did the harder thing. A second moment worth describing came mid-semester, when the applicant was working through a section on social impact measurement. The draft used a composite index drawn from a grey-literature source without interrogating its construction. I returned the memo with a note asking where the index weights came from and whether the aggregation assumptions were defensible for the population in question. The applicant's response in the following session was careful: rather than defending the original choice, the applicant walked me through three alternative measurement approaches, explained the trade-offs in terms of data availability and construct validity, and proposed a sensitivity check using a simpler additive score alongside the composite. That kind of methodological self-correction under critique is what I look for in students moving toward quantitative social research, and it is not universal at this level. I want to be honest about one area where the applicant's preparation is still developing. The quantitative work in the paper was competent and, at its best, genuinely careful, but the applicant's engagement with the statistical literature was narrower than the empirical ambition of the project. Citations tended to come from applied policy reports rather than methods papers, and when I suggested engaging more directly with the econometric literature on programme evaluation, the response was willing but took time to gain traction. For a programme as technically demanding as the Oxford MSc in Social Data Science, I would expect the applicant to need to build fluency in the formal methods literature alongside the applied work. That is a gap I think is closeable — the intellectual habits are there — but it is worth naming. What I can say with confidence is that the applicant handles critique constructively, revises with genuine attention to the underlying problem rather than surface compliance, and brings a grounded understanding of what development data actually looks like in practice. The combination of programme management experience and demonstrated willingness to interrogate causal claims is not common. Most students arrive with one or the other. I support this application. The applicant has the analytical discipline and the intellectual honesty that rigorous quantitative social science requires, and the development policy context gives the methods work a specificity that I think will be an asset in the programme. I am happy to discuss my observations further if that would be helpful.