The final yes/no figure is based on everybody who gave a specific voting intention (either yes or no) and who answered “10 – I will definitely vote” when asked how likely they were to vote in the referendum on a scale of 0 to 10.
We used a number of factors to make the sample match the Scottish electorate both demographically and, unlike our current Westminster voting polls, politically.
For UK general election polls we make our samples match published ONS figures for gender, age group, region, working status and socio-economic grade. Our weighting factors for Scotland were a variation of this with gender, age group, working status and socio-economic grade being the same while region was calculated by asking respondents which county they lived in and combining these into larger regions to be used in weighting.
To match Scotland’s political balance, we used the results of the 2011 Scottish parliament elections and the 2010 UK general election.
Because of the way online polling works, the sort of people who sign up to take part in surveys tend to be more politically enthusiastic than normal and reported turnout in general elections is always far above the true figure. Turnout in the 2011 Scottish election was around 50% yet 78% of our sample say they voted and similar problems occur in all online samples.
Weighting the sample to match the true turnout figure would risk counting the 22% of the electorate who are non-voters as more than double their actual weight so, as a compromise, we leave the proportion of past voters in the sample as it is but weight their proportions to match the results of the election.
Another issue often encountered is how many people say that they voted for the SNP in the 2010 UK general election, usually similar to the proportion who voted Labour. The problem is that Labour had more than twice the share of the vote of the SNP in 2010 so weighting our sample by 2011 vote alone would risk under representing the sizable number of Scots who switched from red to yellow whose views on independence would probably differ from SNP loyalists. With 2010 vote we have done the same as 2011, keeping the share of the sample who say they voted unchanged but weighting those voters to match the election result.