Using online e-petition data to estimate EU referendum results for UK Parliamentary Constituencies

British Society for Population Studies, Winchester, 10 to 12 September 2018.

The United Kingdom’s 2016 referendum on membership of the European Union (EU) is perhaps one of the most important recent electoral events in the UK. The political sentiment of the electorate, with a narrow vote to leave the EU, confounded pollsters, media commentators and academics alike, and has challenged elected Members of the Westminster Parliament. Unfortunately, for many areas of the UK this referendum outcome is not known for Westminster Parliamentary Constituencies, rather it is known for the coarser geography of counting areas. This has implications because it is Parliamentary Constituencies which return an MP. This paper uses novel data and machine learning algorithms to estimate the Leave vote percentage for these constituencies where it was not reported. We utilise information on political sentiment captured by signatories to UK government e-petitions as input data. A range of machine learning algorithms are tested on these data which provides methodological advances beyond traditional regression or ad-hoc approaches. We highlight that such methods are just one of a range of modelling approaches and there is scope for political scientists to apply machine learning algorithms to gain confirmatory or alternative insight in a range of different applications. Our results are found to correlate well with other estimates which use different methods and data to those outlined in this paper.