| Project Supervisor | Yonatan Berman |
| Institution & Department | King’s College London – Department of Political Economy |
| Research Area | RA 2: Business Analytics, Management, and Applied Economics |
| Project Start Date | 28 September 2026 (felxible) |
| Project Duration | 3 months |
| Application Deadline | 31/07/2026 |
| Working Pattern | Part-time (2.5 days per week over 6 months) |
| Working Arrangements | Hybrid |
| Ideally in person meetings are preferred, but there is a place to allow flexibility with remote work and meetings occasionally. | |
| How to Apply | View Guidance Here |
Project Description
Click to View More
Which political groups actually turn out to vote, and how much does that shape who governs? These are old questions, but they have been surprisingly hard to answer. Official election records tell us how many votes each party received, but not how many of their potential supporters stayed home. Surveys try to fill this gap, but they are unreliable: people systematically over-report whether they voted, and the degree of misreporting varies by party and social background, which means survey-based estimates of turnout gaps are not just noisy but can be biased in direction.
This project is built around a new method that sidesteps these problems entirely. The core idea is simple: in any constituency, the total number of latent supporters across all parties must equal the registered electorate. That arithmetic constraint, combined with official vote totals, is enough to recover party-specific turnout rates through constrained optimization, using only administrative data and no surveys required. Unlike existing ecological inference approaches, the method requires no strong distributional assumptions, scales easily across parties and elections, and can in principle be applied wherever official returns are available.
The project will develop and test this method, apply it to an initial set of cases, and use the results to ask a substantive question that has been difficult to study systematically: how large are the turnout gaps between political parties, and what consequences do they have for who governs? Under a majoritarian electoral system like the UK’s, where small differences in vote shares translate into large swings in seats, even modest and persistent differences in mobilization between parties could be enough to determine which party forms a government.
The longer-term ambition is to extend the method across countries and over time. Because the approach requires only constituency-level vote totals and registered electorates, it can in principle be applied across most democracies going back well into the nineteenth century, long before survey data existed. That would make it possible to trace how partisan turnout gaps evolved across different electoral systems, party families, and historical periods, opening up a new line of research on participation, representation, and democratic legitimacy that has not previously been possible at this scale.
The internship will focus on the first stage of that wider project: building and testing the methodology, applying it to a pilot set of cases, and creating a reproducible workflow that can be expanded later.
Internship Details
The internship will suit someone who finds it genuinely interesting to sit at the boundary between methods and substance, where a technical decision about how to handle a small party in a historical dataset also has implications for what you can say about democratic representation.
The work will involve independent problem-solving, but never in isolation. Some of the data will be messy, especially for older elections or smaller democracies, and part of the job is developing good judgement about how to handle that rather than looking for clean answers.
The project is ambitious in scope, but the internship itself is well-defined: the goal is not to build an entire comparative dataset in three months, but to establish the method and create a solid foundation that the wider project can build on.
Internship Structure
The student will work directly on building and testing the method. This will involve reviewing the relevant literature on turnout, ecological inference, and comparative electoral data, identifying and gathering election data from official sources, and helping to implement and test the constrained optimisation approach on a pilot set of cases.
Further tasks will include harmonizing party names and constituency boundaries across elections and countries, diagnosing issues that arise when applying the method to new contexts (for instance, near-collinear party structures or inconsistent historical records), and producing summary tables and figures showing estimated turnout rates by party.
By the end of the internship, the student should have contributed to a focused literature review, a cleaned pilot dataset covering an initial set of countries or elections, reproducible code for running the estimation, and a short methods memo summarizing findings, data decisions, and recommended next steps for the wider project.
Anticipated Benefits for the Student
The student will develop practical skills in working with large, heterogeneous, and often messy datasets, learning to make and document judgement calls about how to handle missing data, inconsistent coding, and boundary changes that are endemic to historical and comparative electoral research. They will gain experience implementing and adapting constrained optimization code in Python, and in presenting quantitative results clearly and reproducibly.
Beyond the technical side, the internship offers the chance to see how a research idea becomes a usable method, and how a method becomes a research resource that others can build on. The student will contribute to decisions that shape the direction of a live project: which cases to prioritise, how to handle edge cases, where the method works well and where it needs more care. That kind of involvement in early-stage research design is hard to get from coursework or data exercises alone.
More broadly, the student will build experience in managing a structured component of a research project, communicating progress to a supervisor, and producing outputs (cleaned data, code, written notes) that are designed to be used by others.
Skills, Experience and Knowledge Requirements
The student should be comfortable working with data in R, Python, or Stata, and should be able to write clean, documented code. They should be organized and careful, with the patience to work methodically through detailed tasks involving imperfect or inconsistent material.
A genuine interest in elections, political participation, or comparative politics will make the work more rewarding. Some familiarity with electoral data or historical research is helpful but not essential.
The most important things are good quantitative instincts, attention to detail, and the motivation to help build something new from the ground up.