Voter Transitions in the 2015 British General Election: Combining Online Panels and Aggregate Data at the Constituency Level to Explain the Transition to Brexit

The background of the 2015 British general election was marked by the promise of David Cameron in 2013 to hold a referendum on the EU membership in case of Tory victory. What are the underlying voter shifts that led to the electoral outcome that was, in hindsight, the precondition of the Brexit decision? Given the electoral system, nation-wide comparisons are challenging. Estimations of voter transitions based on (online) surveys and recall questions are notoriously plagued by biases, whereas aggregate data analyses can suffer from ecological fallacy. We propose the usage of a systematic statistical combination – a recently developed Bayesian approach – of individual data and aggregate data at the constituency level to assess electoral swings between the 2010 and 2015 elections. The individual data stems from the British election study (BESIP) and allow us to locate respondents in their constituencies. We are able to show that our data integration procedure leads to relevant deviances from both pure ecological inference and pure online panel based estimates of transitions. Deviations in contrast to the online panel are mainly due to our inclusion of ecological information on nonvoters. These are clear hints for the superiority of our hybrid model. Substantially, our results demonstrate that e.g. in England, the most dramatic party shifts occurred for the Liberal Democrats. The party has the smallest loyalty rate of only 26%: the major transitions from the Liberal Democrats are to the Conservatives (17.1%) and Labour (22.7%). By contrast, 11.9% of former Conservative voters, 8.8% of Liberal Democrats voters and 6.6% of Labour voters moved to UKIP.