Stephen Fisher, 18th September
The results of today’s referendum on Scottish independence are being counted and announced by local authority area. Every vote counts equally. It is not like a British general election or even Scottish Parliament election, at which votes for some parties in some places have more chance of influencing the overall outcome than others.
Given the polls suggest the result could be close this means we might need to wait until all the results are declared before we know the outcome, especially if the two big cities (Edinburgh with 9% of the Scottish electorate and Glasgow with 11%) are among the last to declare. Unless the results strongly favour one side or the other, it will be difficult to interpret the early declarations to say what they imply for the overall outcome.
There will be places people expect to vote Yes and those expected to vote No and some in the middle, but how strong the Yeses need to be and how strong the Noes and which side of 50% the middling councils need to be to tell us the overall result depends on lots of factors.
Opinion polls tell us about what kinds of people are more likely to vote Yes than No. For instance those who voted for the Scottish National Party (SNP) are more likely to vote Yes and so we would expect places where the SNP vote share was relatively high in this May’s European Parliament elections to be more likely to vote Yes, and vice versa.
So on this basis the two council areas with a border with England (Dumfries and Galloway and the Scottish Borders) should be among the councils where the No vote is highest while Dundee and Angus should have higher support for Yes than average.
But other factors are relevant too. Support for Yes seems to be higher among people from working class backgrounds, especially in relatively deprived areas so Inverclyde and Glasgow are expected to be more in favour of independence than most council areas.
Also the proportion of people who were born in Scotland and the proportion born in the rest of the UK is likely to play a role in determining the strength of the Yes vote in a council.
While support for the union is stronger among pensioners than those of traditional working age, council areas differ only somewhat in the proportion of over 65s. Moreover, while polls show a big gender gap in support for independence this will have practically no effect on the geographical distribution of the vote because the proportion of men and women is roughly the same everywhere.
Working out which council results are most likely to be indicative of the overall outcome is not just a matter of working out which councils are the most average on the factors that influence voting. The balance of support for independence in all the different councils plays a part. If independence gains moderate support in most places but rejected by large majorities in a few places the overall result could be No even though the most average councils vote Yes.
Similarly the relative size of the councils matters. So even if most places vote No the overall result might still be Yes if support for independence in large council areas, especially Glasgow, is particularly strong. Of course this scenario requires turnout in Glasgow to be roughly as high as elsewhere even though in previous elections it has been somewhat lower. At the 1997 referendum on the Scottish parliament turnout in Glasgow was nine percentage points lower than the overall figure of sixty. However, the polls suggest a very high level of overall turnout this time which leaves less scope for big differences between councils.
So estimating which councils are most likely to be indicative of the overall result depends on assumptions about the factors affecting the distribution of support for independence across council areas, their relative importance, the pattern of turnout and some other things.
Undaunted, I considered a variety of different plausible statistical models for the distribution of Yes and No votes across councils. For each model I looked to see which councils would be most likely to have a tie between Yes and No in the event that there is a tie for Scotland as a whole.
On this basis the councils most likely to be indicative of the overall result are East Lothian, Fife, Midlothian, Renfrewshire, South Ayrshire and West Lothian. If these councils all have a majority for Yes then the overall result is likely to be Yes.
However this is less likely to work the other way round. If the statistical correlation between social deprivation and support for independence turns out to be very strong, producing massive majorities for Yes in more deprived council areas and especially in Glasgow, this might out weigh even a very large number of council areas voting No, including otherwise indicative ones.
While the six indicative councils are all expected to declare their results before Glasgow and Edinburgh, they are likely to be announced with most others, perhaps between 3 and 4am. So the pattern of the results at this stage will be a more important consideration than the results of the indicative councils. But since East Lothian is expected to be among the first to declare, if most people there vote Yes it could be our first and important indication that Scotland has voted to become an independent country.
Acknowledgements: Thanks to Rob Ford and Chris Hanretty for help with the data.