One of the unknowns heading into Scotland’s independence referendum was how different parts of the nation would vote. But now that we know how the regions voted, it turns out that we had a decent guide all along: how the regions voted 17 years ago on a similar referendum that decided whether to create a Scottish parliament with some self-governing powers.
Pollsters didn’t have big enough samples to accurately portray the 32 local constituencies that would be counting and reporting their residents’ votes. Prior votes — for the U.K. and Scottish Parliaments, or for devolution of powers from London to Edinburgh — weren’t perfect analogues for the first-of-its-kind referendum. And the eight major electoral regions of Scotland simply didn’t vary much in the most recent vote for Scottish Parliament or in the polls, making it tough to separate the signal from the noise.
Now that the vote is in, we can assess which factors best predicted how various parts of the nation felt about independence. The most reliable forecasts among the ones I’ve seen used data that combined a large sample size with a direct indication of what people thought about Scotland’s self-governance. One used the 1997 vote for devolution, and the other used polling data from this year’s British Election Study, with thousands of Scottish respondents stating their attitudes about independence — supplemented by other data to overcome the small sample size in some localities.
I checked six forecasts: three from a blogger who goes by Number Cruncher; one from Chris Hanretty, a political scientist at the University of East Anglia; one from Carl Cullinane, who is a research assistant at the Democratic Audit, a London School of Economics-based research organization; and mine, based on 2011 Scottish National Party vote share in the Scottish parliamentary elections and published as part of my rough guide to watching the vote results come in.
For each one, I calculated the predicted “yes”-minus-“no” margin for each council, relative to the rest of the country — by how much was each council expected to lean toward independence? Then I compared that to the actual “yes”-minus-“no” margin relative to the rest of the nation, based on results counted and reported Friday morning in Scotland.
All of the forecasts showed some skill in predicting how the 32 councils would vote, with correlation coefficients ranging from 0.54 for mine to 0.81 for Hanretty’s and 0.82 for the best of Number Cruncher’s — based entirely on the 1997 devolution vote. Hanretty’s had the least average error, followed by the same Number Cruncher forecast.
Some areas confounded pollsters and forecasters alike. “Places in the northeast of Scotland that have returned SNP [Scottish National Party] members of Parliament for almost 30 years have let Alex Salmond down,” said James Crouch, research executive at pollster Opinium, referring to the first minister of Scotland and the leading advocate for independence. Salmond announced Friday he’d step down after losing the vote. “In fact, some of ‘yes’s’ best results came from some of the ‘old Labour’ heartlands in and around Glasgow.”
Hanretty wasn’t surprised that the 1997 vote was so predictive. “Certainly the patterns in the Borders and Islands are long-standing,” he said in an email. But the failure of 2011 voting to better predict the results was “weird,” he said. “It goes to show that you can support independence without supporting the SNP (which was always a claim made by the Scottish Greens and other radicals), and that you can support the SNP without supporting independence.”
Glasgow was one of the big wins for the two leading forecasts: Each anticipated that Scotland’s most-populous city would lean heavily toward independence, while the other forecasts didn’t. They also did better than most other forecasts at anticipating the big “no” wins in the Orkney Islands, the Shetland Islands, and Perth and Kinross.
Stephen Fisher analyzed the results and found another factor that predicted local voting patterns: areas with economic hardship leaned toward “yes.” “It seems that economic hardship led people in the hardest hit areas to conclude that economic management from Westminster was not working for them,” Fisher wrote on Elections Etc.
The referendum’s singular nature was always going to make it tough to predict voting in every region. It’s also easier to identify differences between regions when the regions are very different. As has been the case in other recent national votes in Scotland, voting was relatively uniform. The vote share for “yes” ranged from 32.8 percent in the Orkneys to 57.3 percent in Dundee. And nearly 3 in 4 localities (23) ended up in an 11 percentage-point range for “yes” support: 36.3 percent to 47.2 percent.
To get a sense for how uniform those vote numbers were, I compared the Scotland vote with the 2012 U.S. presidential election. “No” won 55 percent of votes across Scotland; President Obama won 52 percent of votes cast either for him or for Republican challenger Mitt Romney. I then compared the standard deviation of “no” votes with the standard deviation of Obama’s two-party vote share across the 50 states and Washington, D.C. The standard deviation for “no” in Scotland across the 32 localities was 6 percentage points; the standard deviation of Obama’s vote share across 50 states and D.C. was 12 points.