Rob Calvert Jump and Jo Michell

We have a “featured graphic” article on Brexit voting patterns and education forthcoming in Environment and Planning A. The electronic version is here, and an ungated pre-print is here.

It is by now well established that education is one of the most statistically important demographic factors in “explaining” the Brexit vote. It also has substantial predictive power. What has not been explored, to our knowledge, is the extent to which educational attainment and geography interact: scatter plots and regressions tell us how variables move together, but they don’t (usually) include information about geographical patterns.

The figure below demonstrates why such information might be important (hi-res version here). The maps show colour-coded Leave vote percentages by local authority. Black outlines denote authorities for which the proportion of the population with less than five GCSEs is below the national average of around 36%. The match between low GCSE attainment and Brexit votes is striking: only 4 of the 85 local authorities that voted Remain had lower than average high school educational attainment: Liverpool, Sefton, the Wirral, and Leicester (researchers have recently speculated that Liverpool’s Remain vote was influenced by the city’s boycott of The Sun) Conversely, there is only a single local authority, North Kesteven in the East Midlands, with better than average educational attainment and a Leave vote share of 60% or higher.

Choropleth map and cartogram hex map of Leave vote share and educational attainment in England and Wales by local authority.

Areas of below-average educational attainment and strong Leave voting show a clear geographical structure: clusters extend from the east coast into Essex and Kent in the south, and into the West Midlands and the north west of England in the north. The correlation does not hold so well in Wales, Cumbria, Northumberland and County Durham, where several local authorities with below-average GCSE attainment returned Leave votes between 50% and 60%.

Simple bivariate correlations like this are of course likely to be confounded by other factors that covary with aggregate educational attainment: age is the most obvious. To test the extent to which this may be driving the result, we construct a measure of age-adjusted educational attainment, which adjusts for the age and sex structure of the local authority – the details are in the paper. The figure below shows how this measure correlates with Leave voting (hi-res version).

Choropleth map and cartogram hex map of Leave vote share and age-adjusted educational attainment in England and Wales by local authority

Once age and sex are adjusted for, the extent to which low GSCE attainment is clustered strongly among parts of the east coast, the West Midlands and the North West is even more apparent. Again the match with strong Leave votes is striking. Adjusting for age and sex removes many of the local authorities returning Leave votes between 50% and 60% from the “below average” educational attainment category, although parts of Wales are still apparent “outliers”. The area of below average (adjusted) educational attainment also extends into Remain-voting East London and Manchester.

How should these patterns be interpreted in light of the various narratives seeking explain the Brexit vote? These can broadly be divided into those emphasising cultural divergence between socially liberal Remain voters and socially conservative Leave voters, and those emphasising economic drivers such as inequality, austerity, and the effects of globalisation. Since educational attainment is closely correlated with both social attitudes and economic success, our results could be invoked in favour of either set of narratives. We therefore caution against drawing any firm conclusions on the basis of these results. The distribution of strong Brexit votes and below average educational attainment does raise potential problems, however, for narratives of the vote based on an assumed North-South divide. More on this soon.