Welcome to Deltapoll
Today sees the launch of a brand new opinion research company, Deltapoll, which brings together names and faces that many people on this website might be familiar with.
Joe Twyman has for years been the primary face of public opinion in the UK in his position as YouGov’s head of political research, and Martin Boon spend 23 years at ICM, delivering polls that have been relentlessly reported in these pages.
Our third founder, Paul Flatters, is also not unknown to the polling world in which we operate, having navigated a path via Head of Political News research at the BBC through various opinion research and forecasting agencies. He will remain Chief Executive of Delatpoll’s sister agency, the trend forecaster Trajectory.
Does the world really need another polling company? It’s not as if pollsters like us have covered ourselves in glory over the last few years. This is one reason we’ve joined forces. Few have more collective experience and knowledge of looking under the polling covers than us, and having experienced the glory of pinpoint accuracy and the devastation of the opposite, we know how much work needs to be done, and more importantly what needs to be done.
Political polling is very much at the cross-roads in its evolution. We know now that stated response is an unreliable measure of human behaviour, and quietly we’ve been testing new methods that measure the ‘emotional certainty’ that people reveal when they mark their vote intention box on a poll questionnaire.
And that’s a key point. Just like many decisions that humans make, the act of voting is a highly personal one, sometimes linked to innate and inviable personal preferences, other times to emotionally charged acts of rejection or anger. Orthodox polling techniques have failed to account for this, particularly when social media has created a release valve for the expression of emotionally-charged political views. This mouse wheel will only spin faster – as the public gets ever more able and willing to emote on public platforms, potentially radically changing the way we ‘do’ politics, polling techniques can’t remain statically accepting of yesterday’s culture.
But there’s more. Emotional measurement is only one aspect of this. Mathematics is the other. The advent of MRP modelling has come at precisely the right time, and not just to (hopefully) improve the quality of prediction. It can achieve what orthodox polls cannot – a seat-by-seat prediction that doesn’t depend on old mantras such as uniform swing in the conversion of vote shares into seats (accepting that the poor reputation of uniform swing modelling reputation is slightly unfair). Really, who cares what national votes are at any given moment in time? We’ve only spent our lives predicting them because polls have been largely incapable of translating them into accurate seat projections. Here’s a chance for pollsters to grasp the holy grail of accurately predicting seats from polling data allied to sophisticated modelling, and the race is on to perfect it.
This race is likely to be long distance rather than a sprint. Accurate MRP techniques still depend on representative samples, and by corollary accurate vote intention data. So yes, we do think the polling world is not yet ready to be ruled by genius data scientists, although we have a few in Deltapoll’s corner, as the coming months will reveal.
And the media world, including the Poliitalbetting site, still needs a range of polling conversations independent of the predictive element, and we hope to be able to oblige. Bookmark for the next data library to hit the online shelves.