Imperial Research Proposal Example: Pension fund analyst to retirement policy (Score 93)

The United Kingdom's shift from defined benefit (DB) to defined contribution (DC) pension provision has transferred longevity and investment risk to individual savers at a scale that public policy has not yet fully absorbed. Auto-enrolment, introduced under the Pensions Act 2008 and extended incrementally since 2012, has substantially increased participation rates among private-sector workers. Yet participation is not the same as adequacy. Minimum contribution rates — currently set at 8 per cent of qualifying earnings — were calibrated against replacement-rate targets that predate the low-yield environment of the 2010s and the post-pandemic inflationary period. Whether those rates, applied to the earnings bands and contribution schedules currently in force, are sufficient to deliver the retirement income that policy documents implicitly promise is an open empirical question. This proposal asks: to what extent do current auto-enrolment contribution parameters generate adequate retirement income for median and below-median earners in the UK private sector, and which policy levers — contribution floor, qualifying earnings band, or default fund design — produce the largest adequacy gains per unit of fiscal cost? Two subsidiary questions follow. First, how sensitive are projected replacement rates to plausible variation in real asset returns and annuity pricing over a thirty-year accumulation horizon? Second, do adequacy shortfalls differ systematically by gender and employment type in ways that existing policy instruments do not address? The practical stakes are concrete. The Department for Work and Pensions' 2017 automatic enrolment review deferred a decision on contribution escalation; the 2023 Mansion House reforms signalled renewed interest in DC fund design without resolving the adequacy question. A research-grade quantitative analysis of the gap between current parameters and defensible replacement-rate benchmarks can inform that unresolved policy debate. Two bodies of scholarship bear directly on this question. The first is the actuarial and financial economics literature on retirement income modelling. Booth and colleagues' work on stochastic asset-liability modelling for pension funds, and the broader tradition of lifecycle consumption-smoothing models following Modigliani and Brumberg, provide the theoretical architecture for projecting accumulation outcomes under uncertainty. More recent empirical contributions — including work by Blundell, Emmerson, and Wakefield on UK household saving behaviour — have documented the gap between stated retirement saving intentions and observed contribution rates, but have not translated that gap into a policy-parameter sensitivity analysis calibrated to the post-2017 auto-enrolment schedule. The second literature is the public policy and pension governance strand. The OECD's Pensions at a Glance series and the Pensions Policy Institute's annual DC adequacy reports provide descriptive benchmarks, but they typically present point estimates rather than confidence intervals across return scenarios, and they do not disaggregate adequacy shortfalls by earnings decile and employment contract type in a way that maps onto specific policy instruments. Work by Cribb and Emmerson at the Institute for Fiscal Studies has examined auto-enrolment's effect on participation, but the adequacy dimension — what the accumulated pot actually buys in retirement — remains underexplored in the academic literature relative to the participation dimension. The gap this proposal addresses is therefore methodological as much as substantive: there is no published study that combines stochastic projection modelling with a systematic policy-instrument sensitivity analysis using the current UK auto-enrolment parameter set, disaggregated by earnings level, gender, and contract type. Filling that gap requires both the financial modelling toolkit and access to earnings distribution data — a combination that the proposed research design is structured to provide. The study will proceed in three phases. Phase one constructs a stochastic projection model of DC accumulation under current auto-enrolment parameters. The model will simulate thirty-year accumulation paths for synthetic worker cohorts defined by earnings decile, gender, and employment type (full-time permanent, part-time, and self-employed where auto-enrolment applies). Asset return assumptions will be drawn from a calibrated block-bootstrap of UK equity and bond returns over 1990–2023, generating a distribution of projected pot sizes rather than a single deterministic estimate. Annuity pricing will be modelled using current gilt yield curves with a sensitivity range of plus or minus 150 basis points to capture plausible medium-term variation. Phase two introduces policy counterfactuals. Four parameter variants will be tested against the baseline: (a) raising the contribution floor to 12 per cent; (b) removing the lower qualifying earnings threshold; (c) extending auto-enrolment to workers aged 18–21; and (d) mandating a default fund with a higher equity allocation in the accumulation phase. For each variant, the model will compute the change in median projected replacement rate and the 10th-percentile replacement rate, alongside a simplified fiscal cost estimate based on tax relief implications. Phase three conducts a heterogeneity analysis. Using earnings distribution data from the Annual Survey of Hours and Earnings (ASHE), the model will be re-run for female part-time workers and workers in the bottom two earnings deciles — groups for whom the qualifying earnings band is most binding — to test whether the adequacy shortfall is concentrated in identifiable demographic subgroups. The primary data source is ASHE, a publicly available Office for National Statistics dataset that does not require individual-level access permissions. The projection model will be built in R, with code deposited in a public repository on completion. No primary data collection from human participants is required. The research design is deliberately bounded to what is achievable within a one-year taught master's dissertation, with a quantitative component that extends naturally into a research-degree project if the programme is structured accordingly. Data access presents no material obstacle. ASHE microdata are available through the UK Data Service under standard academic registration; the application process typically takes two to four weeks. Asset return data are available from the Bank of England's statistical database and from publicly accessible academic repositories. No proprietary fund-level data are required for the core analysis, though a sensitivity check using industry DC fund return data published by the Pensions Regulator's annual DC trust survey is planned as an extension. Ethical risk is low. The study uses aggregate and anonymised administrative data; no individual participants are recruited, interviewed, or tracked. Standard data management protocols — secure storage, no re-identification attempts — will be followed in line with Imperial's research data policy. Provisional timeline: months one and two, literature review and model architecture; months three and four, baseline model construction and validation against published PPI benchmarks; months five and six, policy counterfactual runs and heterogeneity analysis; months seven and eight, write-up and sensitivity checks. The main contingency risk is a delay in ASHE data registration; this is mitigated by initiating the application at programme start and using publicly available ASHE summary tables for preliminary model calibration. The scope is intentionally limited to the UK private-sector DC context. International comparison and public-sector pension reform are excluded from this proposal; they represent natural extensions rather than core commitments. Imperial College Business School's research in finance and public economics provides the institutional environment most suited to this project. The School's strength in quantitative finance methods — including asset pricing, risk modelling, and empirical corporate finance — maps directly onto the stochastic projection methodology proposed here. The MSc Management and Finance programme's emphasis on rigorous quantitative analysis and its access to Bloomberg terminals and financial databases supports the asset return calibration component of the model. The project sits at the intersection of financial economics and public policy, a combination that aligns with research activity in the School's finance group and with the broader Imperial commitment to policy-relevant applied research. The proposal does not name a specific supervisor, as supervisor alignment should be confirmed directly with the department before application; the research design is, however, structured to be legible to faculty working in empirical finance, pension economics, or public economics. The primary computational resource required is standard statistical software (R), which is available through Imperial's research computing environment. No laboratory access, fieldwork budget, or proprietary data purchase is anticipated. The project is therefore feasible within the resource envelope of a standard master's research allocation, with a clear path to extension if pursued at doctoral level.