Referendum night is going to represent something of a departure from usual. There will not be the drama of an exit poll announcement to stir excitement – and possibly shock – at 10pm. Meanwhile, when the actual results do start to be announced, except in Northern Ireland they will not be declared by the parliamentary constituencies with which we have all become familiar. Rather they will be unveiled local authority by local authority. As a result, we will get just one declaration for the whole of Birmingham, while, at the other end of the spectrum, the Isles of Scilly will get their moment in the sun.
But perhaps the biggest departure from the routine of election night will be that there will be no ‘last time’ against which to compare the results as they are declared. So when Sunderland or Swindon announce their result we will not be able to say whether it represents a ‘swing’ to Remain or Leave – and thus for which side, if either, it represents a good result.
To overcome this problem we have, on behalf of the BBC, been beavering away at establishing which local authorities appear to be more likely to record a relatively strong vote for Remain, which are the ones where Leave can be expected to do relatively well, and which are the council areas where the two sides could be expected to be equally matched. Our evidence has come primarily from a dataset of over 61,000 interviews about people’s attitudes towards the EU. These interviews were conducted with people in Great Britain by YouGov between March of last year and March of this year and we are deeply grateful to the company for making these data available.
We have analysed these data as follows. First, because age and education are the biggest demographic correlates of referendum vote choice, we have weighted the YouGov respondents in each local authority so that their age and education profile matches that for that council as measured by the 2011 Census. Second, we have assumed that within each local authority the variation in turnout between different demographic groups will be the same as that which pertained in last year’s general election across the country as a whole (as measured by the British Election Study, a major academic high quality survey of how people voted in last year’s general election). Between them these two steps give us an estimate of the share of the vote for Remain and Leave in each local authority.
We do not, however, take these data at face value. Even with 61,000 interviews the estimates for each local authority are potentially subject to quite a large sampling error. Instead we model the data (using regression analysis) in order to ascertain the systematic patterns in the data. This analysis confirms what we might have anticipated from the evidence of the polls – local authorities appear to contain more Leave voters if there was a large vote for UKIP there in the 2014 European elections, if there was a small vote for parties of the ‘left’ (Labour, Liberal Democrats, Scottish and Welsh Nationalists and Greens) on the same occasion, and in places with relatively low proportions of graduates, young people, and people from an ethnic minority background. Our estimate of the likely success of Remain and Leave in each local authority are the values predicted for that authority by this statistical model. Estimates for Northern Ireland and Gibraltar,however, are informed by other polling evidence from those places.
We have then made a couple of further tweaks to our data. First we have assumed that turnout will overall be 60%, but the variation in turnout between local authorities will mirror that in the 2014 European elections. Then, second, given that assumption, we have adjusted the estimates from our statistical model so that if all of our assumptions were to be filled Remain and Leave would be tied at 50% each across the UK as a whole. Thus, if one side starts consistently to do better than anticipated by us across a range of local authorities, this will be a strong signal that it is likely to prevail once all the results have been declared.
So where are the places where Remain and Leave are expected to do well? And is there such a thing as a bellwether local authority where who wins is likely to be the winner overall? Below is a list of the top ten local authorities where Remain are expected to do best, the ten where Leave are expected to do best, and the ten where the outcome would be expected to be on knife-edge if the country were to divide 50:50.
The Top Ten for Leave
The Top Ten for Remain
City of London
The Ten in the Middle
Rhondda Cynon Taff
Two of the earliest results expected are Sunderland and Wandsworth. How should we read those? Well, if Leave does not have a substantial lead in Sunderland, Remain are more likely to win overall. In contrast, if Remain does not have a very big lead in Wandsworth then Leave are likely to win overall.
But a word of warning about the last of our three lists. There are in truth a lot of places where the outcome locally is expected to be close if that proves to be the outcome across the country as a whole. Our estimate puts Remain and Leave within four points or less of each other in no less than 58 council areas. There are in truth no key bellwethers, but rather a large swathe of the country that is likely to tell us the story of the night. We now just have to wait and see what that story proves to be.
This post originally appeared on the blog Elections Etc.