Explaining the 85% turnout in the referendum is relatively straightforward.  Explaining the 45-55% Yes-No vote is less so.  Turnouts are generally affected by two factors: perceptions of the importance of the issue under debate and perceptions of how close the vote is likely to be.  Taking these into account we would have expected a high turnout but 85% is at a level never previously witnessed in modern Scottish politics.  More explanation is required.  Part of the explanation must lie in the length of the campaign and the level of engagement that this permitted.  General elections are characterized by a long campaign, often lasting from the moment one election is over, and a short intense campaign lasting a few weeks in the run up to polling day.  This distinction is always blurred but more so in this case.  The level of activity surpassed anything we have previously seen and vastly so compared with the 1997 devolution referendum. The Sun’s Andy Nicoll’s comment captured this well in his idiosyncratic response to the Better Together campaign message, ‘If you don’t know, vote no.  Seriously? If you don’t know after two years and nine months, stay home and do some colouring!’ 

Two years nine months ago, the polls suggested that NO would win comfortably.  Support for independence rarely rose much about a third.  One survey in April 2011 showed support for independence at 35% and this was a high point compared with others taken around that time and was at 31% in June 2013.  Some commentators had suggested that supporters of independence had made a mistake in not pushing for a snap poll to take account of the SNP honeymoon following the 2011 elections.  Such commentary failed to take account of the impact of a long campaign.  There were fluctuations in levels of support for independence over the long campaign but with a very clear trend towards increasing support.  The more people engaged with the campaign, the more likely they were to vote for independence.  

A pattern emerges in analysis of local authority areas most likely to vote Yes/No.  Voters in areas with higher levels of multiple deprivation saw higher votes for independence. 

Data derived from combining population data from Scottish neighbourhood statistics and local authority votes

This conforms with polling data before and since the referendum.  Class played a significant part in the referendum.  It is also clear from polling data that the best predictor of referendum vote according to party support was not SNP support but support for the Conservatives.  Tory peer Lord Ashcroft’s poll shows that the overwhelming majority of Conservative voters voted NO (Table 1). 

The age and gender profiles of voters according to constitutional preference was as expected.  Women were more likely to vote NO than men though the difference was not as stark as some had suggested:  47% of men and 44% of women voted Yes.  A more striking difference was evident in the age profiles of voters: with the exception of 18-24 year olds, who split fairly evenly, voters below the age of 54 were more inclined to vote YES than those over 55. 

It is possible to draw up a picture of the typical voter for Yes and NO - though it must be stressed that there were, of course, individuals with these characteristics who voted differently.  The typical NO voter was a Tory-voting wealthy woman pensioner living in an affluent part of Scotland.  The typical YES voter was an urban male who had voted SNP, with a Labour voting background, who strongly supported the welfare state and lived in an area where there was very little support for the Tories.

The pattern of voting is clear but the explanation for the result less certain at this stage.  There will be temptations to rush to conclusions and suggestions that late interventions by the Prime Minister and Gordon Brown saved the day for NO.  This assumes that these polls were correct and that YES was ahead (even though margins or error left open the possibility that No retained a narrow lead).  The broader ‘poll of polls’ trend shows that Yes gained significant ground over the course of the campaign.  This was a referendum in which supporters of the union had a significant advantage but came close to losing.

What then explains the increase in support for independence from the start of the campaign to the vote?  There are a number of possible explanations but at this stage each remains a hypothesis requiring testing with robust data analysis.

HYPOTHESIS  #1: Positive campaigning brought votes to YES compared with the negativity of NO.

There is little doubt that YES was perceived as more positive though there is evidence that the focus on threats to the NHS from continued union – negative campaigning – helped the YES campaign.  The ‘vow’ - the promise of substantially more powers and a clear timetable provide – is evidence of positive campaigning from NO in the final week.  There is evidence that threats to prevent Scotland using the pound proved ineffective especially in the period between June and referendum day.  Evidence also suggests a similar pattern with regard to negative campaigning on Scotland’s continued EU membership.  While this hypothesis should be explored, the evidence at the moment suggests a more complex explanation.  This explanation needs to be considered alongside the next hypothesis. 

HYPOTHESIS #2: Distrust of Westminster politicians/politics

This is the favoured explanation of many NO supporters including Lord Ashcroft.  There is evidence, as the Tory peer provides, for this explanation.  This is essentially the equivalent of the ‘protest vote’ explanation in elections.  There is indeed evidence that voters who were disillusioned with Westminster were inclined to vote for independence but then all votes for change must start with criticism of the status quo.  There is also evidence suggesting that Westminster politicians, especially George Osborne, lacked credibility with sections of the Scottish electorate.  It may be that it was not so much Better Together’s negative message that was a problem but the messengers.  This may be a start in explaining the vote but falls short of a full explanation and requires some reason why disaffection was translated into positive support for independence.  As the vote was perceived to be close, voters could be in little doubt that voting for independence would result in independence making a simple protest vote explanation unlikely. 

HYPOTHESIS #3: The YES side fought a better campaign

Of this there can be little doubt if we listen to commentary from within each camp.  It is striking how many supporters of the union believe that poor campaigning let their side down.  The implication is that had Better Together been better led it would have done better.  There has been comparatively little criticism of the YES side from within though some voices were critical of YES Scotland early in the campaign and particularly of the perceived poor performance of Alex Salmond in the first STV debate.  But even allowing for partisan bias in people’s judgment, the evidence suggests that people were more impressed by the leader of YES and the campaign they mounted.  However, this was insufficient to provide a majority and more likely to have ensured YES was heard but not necessarily that they converted voters. 

HYPOTHESIS #4: Grassroots activity and social media favoured YES

While there is evidence that YES engaged in more grassroots and social media activity, this needs to be set alongside the enormous advantage that NO had in the conventional media.  No had the best hand – more resources at its disposal (the Electoral Commission rules cover only a fraction of resources, ignoring media support and civil service support).  With the exception of the Sunday Herald (read by 0.47% of the population), the press was overwhelmingly opposed to independence.  The evidence would suggest that the combination of old style grassroots activity with new forms of social media are important but the old media remain important for the moment but have lost much of their impact.  The Sun’s uncharacteristic reticence to offer advice suggests acknowledgement that the days when it could claim ‘It’s the Sun wot won it’ may be coming to an end.

HYPOTHESIS #5: The pivotal position of supporters of more powers

The conventional wisdom is that had the Prime Minister agreed to put ‘more powers’ in some form on the ballot paper it would have won convincingly.  Polling provides evidence of this and evidence from experience suggests that the middle option tends to be favoured.  The absence of the most favoured option in part reflects the absence of an agreed form of ‘more powers’ but also a willingness to gamble that NO would win.  Towards the end of the campaign, the Prime Minister may have regretted that decision and the ‘vow’ essentially means that ‘more powers’ ended up on the ballot paper in the form of voting NO.  There is little doubt that this was one of the main battlegrounds and contributed to some extent to the result.

Conclusion                                                                                                                            

In the final analysis, while there was a significant increase in support for independence, the union retained the support of a majority in a lengthy campaign.  Scottish democracy was rejuvenated.  John Smith, late leader of the Labour Party, is often credited with telling the Scottish Labour conference in March 1994 that devolution was the settled will of the Scottish people.  In the referendum on devolution three years later, three quarters of Scots voted for a devolved Parliament and 64% agreed it should have tax-varying powers.  Twenty years on and Scots have voted in favour of the union by 55-45%.  The independence referendum suggests that there is no settled will.  Attitudes change and the dynamic of that change have tended to be perceptions of UK Governments.  The union has been reprieved but the Scottish Question remains unresolved. 

James Mitchell is Professor of Public Policy at the University of Edinburgh.  He is the author of The Scottish Question, Oxford University Press 2014 and co-directs the ESRC’s Scottish Referendum Study having co-directed election studies in 1992, 2007, 2011 and the 1997 devolution referendum study.