Predicting Elections: An Introduction to Data Science

Part of Advocata’s “Think” series, delivered in partnership with Echelon magazine on 4 November 2020, this online talk and Q&A stepped outside Advocata’s usual economics-and-policy territory to explore a more technical, methodological question: can data science genuinely predict election outcomes, and how does the underlying methodology actually work? Speaker Nuwan Senaratna, a computer scientist and founder of ColomboLabs, was moderated by Research Manager Aneetha Warusavitarana in a session built around the memorable framing that “data is the new crude oil, and data science is the science of extracting and refining this oil into useful information.” Senaratna walked the audience through the fundamentals of data science — from data collection and cleaning through to modelling — before applying those fundamentals specifically to election forecasting, discussing how polling data, demographic patterns, and historical voting behaviour get combined into predictive models, along with the genuine uncertainty and margin for error involved. The main highlight was making an often opaque, statistically dense field accessible to a general policy audience, positioning data science not as an abstract technical discipline but as a practical tool that could sharpen how Sri Lankan social science research and, eventually, political and policy analysis get done.