The limits of demographic predictability: forecasting fertility, mortality and migration under uncertainty

Entry

2027 / 2026/2027 applications

Course

DPhil in Population Health

Supervisors

Jakub Bijak, Charles Rahal

Research unit

Demographic Science Unit
Oxford Population Health, University of Oxford

Which aspects of population change are predictable, over what horizons, and under which assumptions? This project will compare fertility, mortality, and migration, tracing how component-level uncertainty shapes projections of population size and structure.

Background

Population projections inform health services, infrastructure, and labour-market planning. Demographic inertia can make some changes predictable across decades, yet pandemics, policy changes, and shifting social norms can disrupt established patterns. Sophisticated models do not remove this tension: they may introduce additional uncertainty about model choice and behaviour under distribution shift.

This project will distinguish epistemic uncertainty, associated with limited knowledge or imperfect models, from aleatory uncertainty, associated with irreducible variation. The aim is a comparative account of demographic predictability, not simply a competition to produce the lowest forecast error.

Computational Approach

1. Characterise processes and shocks. Develop a typology for studying how demographic dynamics and disruptions affect the predictability of fertility, mortality, and migration.
2. Establish comparative baselines. Analyse public time series using established forecasting and entropy-based methods, examining differences between demographic components and prediction horizons.
3. Trace uncertainty through projections. Investigate how uncertainty in component forecasts propagates into population size and age structure, and what this implies for the interpretation of projections in policy and practice.
4. Investigate extensions where justified. Depending on the candidate’s interests and initial findings, extend the framework towards probabilistic forecasting, machine learning, or individual-level approaches. More complex models will be considered in relation to the substantive limits of prediction, rather than assumed to resolve them.

Data will draw on the ONS, Eurostat, the Home Office, the Human Mortality Database, the Human Fertility Database, and the experimental Human Migration Database.

Training and Research Environment

The student will join the Demographic Science Unit within Oxford Population Health and the interdisciplinary environment of the Leverhulme Centre for Demographic Science. Training will cover demographic forecasting, uncertainty quantification, time-series analysis, statistical and machine-learning methods, and reproducible research, with access to advanced courses in statistics, computer science, and demographic methods. The project uses existing statistical and research datasets; no fieldwork or industry placement is envisaged.

Prospective Student

Applicants should be computationally strong, with a background in statistics, computer science, machine learning, economics, demography, or data science. Programming experience in at least one statistical or general-purpose language is expected, together with a strong interest in population processes and the ability to reason critically about forecasting and uncertainty.

Enquiries and Applications

For an informal discussion, contact Charles Rahal, quoting DSU004 and outlining your research interests and relevant computational experience.

This is a project for 2027 entry to the DPhil in Population Health. Please consult the official Oxford project advert and course page for current application requirements, deadlines, and funding information.