Only one survey during this election has used the sampling approach recommended by the independent inquiry into the 2015 election polling miss. This is from the NatCen Panel and it has some very interesting findings. But before I get on to them, it’s first important we understand why this panel is different, why the results should be treated with caution, and why they shouldn’t be used make final election vote predictions.
What makes this survey different?
Sample accuracy: This is a very brief explanation, if you want more detail please read this by my colleague Kirby Swales. Most poll participants choose to sign up for surveys in response to marketing asking for their views (any of us could sign up to some polls today if we wanted). With NatCen's panel you are specifically picked at random. You only choose whether to take part when asked; you can’t volunteer. That’s because the NatCen panel draws its participants from the British Social Attitudes (BSA) survey which itself is based on a 'random probability sample' (the gold standard of samples). To create that, people are randomly selected via Royal Mail’s post code database and then a NatCen interviewer will knock on their door up to 9 different times to encourage them to take part. It takes ages, but it’s worth it. For example, after 2015’s election some polling panels reported non-voting rates amongst their participants of just 9%. NatCen’s Panel was 28% (the actual non-voting rate was 34%).
Survey fieldwork design: The panel also gives people more time to take part- over a month - and allows completion over web and phone. We've found that surveys done over just a few days or excluding those without internet access can skew the sample, and therefore results.
Deeper understanding of the sample: Finally, by the nature of their recruitment, every participant has completed the BSA survey. This gives us a depth of understanding about participants not typically found elsewhere. For example, alongside the usual demographics, we have already asked people a raft of questions to enable us to put them on long-standing BSA left-right and libertarian-authoritarian scales. More on these in a moment.
Fieldwork dates: Because the fieldwork takes a month, this is not a snap measure of public opinion. This particular survey was conducted from April 27th to May 28th, and around 60% responded within the first week. This means that the majority of respondents would have answered the questionnaire before important moments in the election campaign, like the manifesto launches. Any true changes in party voting intention or likelihood to vote since then will not have been picked up.
Turnout weighting: The second caveat is how we’ve weighted for turnout. Alongside sample accuracy, turnout weighting will be vital in the polling for this election. For this survey we have applied two relatively simple weights. First, a self-reported likelihood to vote where only those answering they are definitely going to vote are counted. The second is a weight generated by modelling against the characteristics of who voted in 2015. (There’s more on this issue here by my colleague Curtis Jessop). All the data below uses weighting to the general population (not accounting for turnout), except where stated otherwise. How different sections of society turn out in the actual vote will obviously be decisive on the size of the effects noted below.
5 important election findings:
1. Either the Conservatives will win or this election has seen one of the most dramatic short campaigns in decades
We found, like most pollsters did around the middle of May, that the Conservatives had a significant lead over Labour: 16 or 15 percentage points (depending respectively on whether turnout is weighted by self-reported likelihood or our modelling). Here are the top line numbers. Looking at the much-discussed differences by age, weighted by population only, Labour were ahead of the Tories amongst younger people (18-34 year olds) 46 vs 31%, but the Tories had a huge lead over Labour amongst over-55s, 58 vs 22%.
Since the bulk of this survey’s fieldwork, many polls have shown much tighter leads for the Conservatives and some almost have the parties level-pegging. That may change again, or may prove to have been, at least in part, illusionary. If the shift in some of the closer polls is real, it will trigger a dramatic re-think in how we view politics. For example, local elections have been a reasonable guide to general election results in the past. The recent locals were won comfortably by the Tories (by 11 percentage points) so that precedent points in completely the opposite direction to some recent polls. Unprecedented things do happen in elections, but they are also probably less frequent than polling misses. We’ll soon know what we have here as far as 2017 is concerned.
2. Labour is now not the party of the working classes
If the trends in the NatCen survey broadly hold in the result, Jeremy Corbyn’s Labour party has little to no claim to be the party of the British working classes. Our survey found that working class support for both parties was surprisingly close, with the Conservatives picking up 38% of working class voting intention to Labour’s 37%. If this is seen in the result, it would seem Jeremy Corbyn and his team have spent a lot of time talking about the working classes, but not to them.
(This uses NS-SEC classification, not the cruder ABC1 / C2DE split. Working class voters are judged to be those working in ‘Semi-routine & routine occupations’ and ‘Lower supervisory & technical occupations’. This does not measure whether someone self-defines as working class, or not).
3. May has united the right and centre, Corbyn hasn't united the left
BSA’s left-right scale reveals some interesting trends missed by basic demographics. People are placed on this scale according to their answers to a series of questions on issues like intervention in the economy and redistribution.
Segmenting voting intention by this scale reveals that Theresa May seems to be doing a much better of job than Jeremy Corbyn of winning over those on her natural ideological side and the centre. At the time of NatCen’s survey, 75% of voters identified on the right planned to vote Conservative, and 56% judged to be in the centre. By contrast, only 43% on the left and 23% in the centre were planning to vote Labour.
4. The Conservatives are doing better on the libertarian-authoritarian scale too
Where someone sits on this scale depends on their answers to a range of questions on things like civil liberties and criminal sentencing. Here again the Conservatives have seemingly secured more of what many would consider their natural base and made in-roads across the spectrum. 53% of on the ‘authoritarian’ end planned to vote Conservative, and 41% in the centre. Whereas, Labour could only call on 44% of those on the ‘libertarian’ end of the scale and 34% in the centre.
5. Over half of people feel politically homeless
56% of respondents, whether voting or not, felt no political party represented people like them. It’s perhaps telling that those most likely to feel that way are those who would traditionally be seen as Labour supporters – those on the left, working class people and council and housing association tenants. As I’ve noted in more detail here, it is also more likely to be people on the authoritarian end of the scale. This suggests a new centre-liberal-remain based party, which some have proposed, would not easily sweep up the politically homeless (and that’s before you consider how our voting system treats third parties).
Follow me on Twitter: @Roger_Harding
Click here to download the data tables.