As MD of REaD – a business that has access to demographic and behavioural data on well over 40 million individuals in the UK and the technical nous to analyse it – we have a unique opportunity to generate accurate insights. The world of big data is at the disposal of the marketing world, so why not use it in the political arena to make predictions around the upcoming election? We decided to take the pollsters on at their own game and see whether it really is as close as everyone is saying. Read more >>
To do this, we looked at how demographics influenced party support and then matched these tendencies with the demographics of each constituency to predict the number of seats each party would win. And our comparison pollster of choice? YouGov’s Profiler, which claims to accurately outline the typical supporter of the Conservative, Labour, Liberal Democrat and UKIP parties.
These differences between predictions, based on polling samples of a few thousand and those from a database of over 40 million, should cause anyone with a vested interest in the election to prick up their ears and consider whether the current polling method is fit for purpose, not least because of the final results that they predict.
The predicted results
Given the media focus on a resurgent Labour as the polling date looms, our outcome should give the Conservatives some peace of mind, putting them safely in first place, although still in need of a coalition:
- Conservative: 307
- Labour: 270
- SNP: 35
- Lib Dem: 17
- UKIP: 2
- Others: 19
YouGov’s predictions (as of the 28th April), on the other hand, does reflect this late red resurgence, giving much more credence to the SNP too:
- Conservatives: 270
- Labour: 277
- SNP: 50
- Lib Dem: 27
- UKIP: 3
- Others: 23
And what of the statistical legend that is Nate Silver? Well, his predictions land pretty much in between REaD Group’s and YouGov’s. He gives the Conservatives the upper hand just like ours but, similarly to YouGov, predicts a high SNP turnout:
- Conservatives: 283
- Labour: 270
- SNP: 48
- Lib Dem: 24
- UKIP: 1
- Others: 24
So what accounts for these differences? Let’s take it party by party.
The Typical Tory
YouGov tells a familiar tale when it comes to the supporters of David Cameron. Being over 60 has a major influence on how you vote, with a +12.67 swing towards the party. To put that in context, +0-1 is a weak correlation, +1-2 is a medium correlation, +2-3 is a strong correlation and anything over that is a very strong correlation. Our own analysis does corroborate this, but only slightly. We found that the only age that positively influences Tory support is the over 75s, and even then the correlation is +1.05.
One area where there is parity between YouGov’s data and REaD Group’s analysis is profession. Being in business or at a manager level was the biggest single influencer on Conservative support in our model, with a +9.84 correlation alongside YouGov’s +6.26. So far, no real shocks – Tory support historically is from the older generations and those in business.
Those likely to support Labour are more likely to work in the public sector, with YouGov giving a high +6.13 correlation alongside our more modest +1.24. But this is where our data starts to disagree with that of YouGov’s.
Our data suggests that the typical Labour supporter tends to be younger, with a negative correlation against the 65-74s of -1.31 and a positive correlation with the 35-44s of +1.92, corresponding to perceived wisdom that age brings conservatism. Yet YouGov bucks this trend giving a positive correlation to those aged 60+ of 1.92. While neither of these are strong correlations, it does pose questions as to the veracity of the pollsters and, of course, our understanding of who the typical Labour supporter is.
The ‘yoof’-ful Liberal Democrats
This conflict over the age of typical party supporters occurs again when analysing the classic Lib Dem backer. But this time it is now our data that bucks the agreed young support base of the Lib Dems, giving us a negative correlation of -1.31 for those aged between 35-44.
YouGov sticks with the norm this time around, with those aged 39 and under having a positive correlation of at least +3.21. YouGov also agrees with the typical left-leaning agenda this group is assumed to have, with public sector workers having a +2.37 correlation with the party. However, the largest correlation our data brought up was actually in education, with those working in the sector associated with a +5.16 swing towards the Lib Dems. YouGov corroborated this, but not on the same scale, reporting a +1.33 correlation. So is the age-old Liberal Democrat strategy of targeting the youth vote actually based on reliable data? If I were a strategist in their camp, I’d take another look at where my support lies and where it needs strengthening.
And now we put the cat amongst the pigeons. The rhetoric used by UKIP party members and supporters is that they represent the view of ‘normal Britons’, a silent majority that don’t speak out thanks to the PC brigade. Is this the case? Our results would suggest that yes, it is. Our data indicates no over-riding factors that drive or reduce support for UKIP. In fact, the largest factor was being a manual labourer or factory worker, a group that are less likely to vote UKIP by a minimal -0.41.
YouGov, however, firmly agrees with public opinion on the stereotypical UKIPer, with a huge +19.28 correlation with those aged 60+. So, based on our results, the UKIP supporter really is an identi-kit Brit. The election itself will show whether this really is the case.
So where does this leave us? Hopefully it gives us a reason to pause and think about how we use data to understand and predict human behaviour; depending on how it is measured one can get many different results. Irrespective of the outcome of the election, it would be wise for pollsters turn to our big data model of behavioural analytics to inform their tactics. Even if it does appear that our analysis agrees with the pollsters that this election could really be too close to call.
Scott Logie is managing director of insight at REaD Group