Shortly after the EU referendum last June, I noticed a tweet pointing out that Birmingham council had published a ward by ward breakdown for the area’s Leave and Remain voting figures on its website. Several analyses of the referendum results for the official counting areas had already shown a strong correlation between levels of educational qualifications and voting patterns – areas with less qualified populations tended to have higher Leave votes.
This link was replicated overall in the more granular Birmingham ward data. But there were several wards which were exceptions to this rule – they had low levels of education but high Remain votes. And it turned out that these were the city’s wards with the highest ethnic minority populations. So this ward data revealed a significant and intriguing aspect of the referendum vote which was not so discernible in the analyses based on larger areas. I wrote about this on the BBC website.
While writing the piece I discovered that some other authorities such as Lambeth and Greenwich had also on their own initiative released a ward breakdown. And on Twitter people were tweeting about trying to get the same information elsewhere. So at the end of my article I rashly composed the following statement: “The BBC will be trying to get comparable data from as many places as possible”.
That time-consuming and laborious task fell on the shoulders of my colleague George Greenwood. It took a great deal of persistence, but it generated enough information to shed more light on the referendum vote, as revealed in a new article for the BBC website last week. The referendum was counted by 399 counting areas (380 councils in England, Wales and Scotland, with a separate tally in Gibraltar, while in Northern Ireland results were issued for the 18 constituencies). We left Gibraltar out of it, but George tried to get the most detailed geographical voting data he could from all the other places. After a lot of chasing, there were only three councils which did not respond at all.
Some councils could not supply any detailed data because they mixed all ballot boxes prior to counting; some did possess more local figures but simply refused to disclose them to us. In at least one region several councils agreed collectively amongst themselves to adopt that stance. Others did provide data, but the combinations in which ballot boxes were mixed before counting were too complex to neatly fit into ward boundaries. One major problem we faced was the way in which postal votes were mixed in for counting with geographical areas. Some councils had combined the postals with just one ballot box prior to counting; others had mixed them in with a few boxes; others had distributed them across the board. Clearly this would create issues with any geographical analysis, a fact which some councils were concerned about. We needed to reassure them that we would draw attention to this caveat prominently in any reporting (as in due course we did). But some remained unpersuaded and insisted on this as a reason for not disclosing figures.
George and I are both specialists within the BBC in using the Freedom of Information Act to obtain information from public authorities for journalistic purposes. However Electoral Returning Officers are not covered by FOI, and so releasing any data was entirely up to their discretion. Nevertheless the FOI Act does seem to have encouraged a greater sense of transparency in local government, and my feeling is that we may well have got more data than we would have done a few years ago. Certainly some councils treated the request like a routine FOI enquiry.
In the end George got some useful information from nearly half the areas – 178. There was no further data from Northern Ireland, and comparatively little from Scotland and Wales, so the most valuable data mainly came from England. It varied from constituency level data, through data at the level of clusters of wards which had been counted together, down to individual ward level data, and even in a few cases remarkably localised data from polling districts. This enabled us to both explore national voting patterns and also to identify particularly interesting localities.
We were able to compare the Leave and Remain votes for 1,070 individual wards, about one in nine of the UK total, to the demographic data of the 2011 census. The level of education had a higher correlation with the voting pattern than any other major census measure. The age of voters was also important, with older electorates more likely to choose Leave.
Ethnicity was crucial in some places, with ethnic minority areas generally more likely to back Remain. However this varied, and in parts of London some Asian populations were more likely to support Leave. The combination of education, age and ethnicity accounted for the large majority of the variation in votes between different places.
Across the country there were stark social contrasts between localities which most favoured Leave, such as parts of Middlesbrough, and Remain, such as central Cambridge. There was a broad pattern in several urban areas of deprived, predominantly white, housing estates towards the urban periphery voting Leave, while inner cities with high numbers of ethnic minorities and/or students voted Remain.
Around 270 locations were identified where the local outcome was in the opposite direction to the broader official counting area, including parts of Scotland which backed Leave and a Cornwall constituency which voted Remain. And (based on a few places where they could be compared) postal voters appeared narrowly more likely to have backed Remain than those who voted in a polling station.
The 527 voters in the neighbouring polling districts of Kirk Langley and Mackworth in Amber Valley in Derbyshire, whose two ballot boxes were counted together, produced a leave proportion of 51.99%, very close to the national figure of 51.89%. So possibly political scientists who seek a microcosm of the UK should visit the Mundy Arms pub in Mackworth, the location for that district’s polling station.
This all adds to our understanding of how the electorate voted in the referendum. And it again raises the question of whether this localised level of voting data should actually be released as a matter of routine by the authorities, as in some other countries, so that we have a comprehensive and easily accessible set of figures to inform the study of voting behaviour.
Martin Rosenbaum is an executive producer in the BBC’s political programmes unit at Westminster.
