« Greg Clark MP: How and why we’re devolving powers to cities | Main | Henry Hill: How the arch-devolutionists plan to side-step referendums »

Opinion Pollster Stephan Shakespeare

Stephan Shakespeare: Politicians like to have gurus. They need geeks.

Follow Stephan on Twitter.

Politics in Britain today is still about gurus and wannabe gurus; it ought to be about geeks. The edge in today's game is to be had in campaigning by numbers - it's data that will let you maximise your vote, it's data incorporated into your mobilisation machine that could take Cameron or Miliband from mere coalition to an overall majority. Anyone who doubts it should read Sasha Issenberg's excellent account of Obama's big win in bad circumstances, 'Victory Lab: The Secret Science of Winning Campaigns'. Unfortunately, for a few more years, it's likely to stay a secret - at least for practical purposes - in our old-fashioned, creaky political machines.

Obama's campaign was able to secure victory not by some mysterious quality of inspiration or leadership or great political will, but by geeks who understood how to win extra votes with extra information at the micro level. It's not about big ideas, what matters is tiny real-time experiments which gain a few extra votes in ten thousand nooks. I should declare an interest (and also a proud boast): YouGov/Polimetrix, the US arm of my polling firm, played a key role in creating the data that transformed the Democrats' campaigning organisation from 2006, a process described in the must-read sixth chapter of Issenberg's account, 'Geeks versus the Gurus'. Our huge-scale polling combined with DNC datasets helped to create a likely-to-support predictor and likely-to-vote predictor person-by-person across the country. It replaced the absurdly unusable stereotypes of conventional political strategy (like, in the UK, the Worcester-women and the Ford-Focus-Essex-man) with individual case-level media and message tactics.

Most people in politics prefer gurus, not data. Gurus are people with big ideas about what to do - expressed with lots of confidence and almost no genuine evidence. They rely on the kind of clever talk that can never be proved right or wrong. The good ones throw in a few facts and canny observations, the bad ones talk about hunches and what they feel in their gut, but none of them will help you win. In contrast, the geeks are scientists, and most particularly computer scientists armed with sophisticated statistical analysis programmes and mountains of raw information. Instead of having a few big ideas about swing voters and spreading them pointlessly across the airwaves, they ask - and answer - thousands of tiny specific questions: how will this voter in this house watching this obscure gardening programme on this local cable channel react to this or that particular message? Now let's do it, feed the response back into the database and crunch it again for the next prediction. And again, and again.

Gurus make big sweeping assertions; geeks prefer lots of tiny statistical tests. Gurus guess; geeks know. Gurus are plentiful; true geeks are rare. If you really want to win, invest in the science.

The shift that is taking place in politics - already started in America, not yet glimpsed in Britain - is like the shift that took place in finance over the past thirty years, from the sharp-suited MBA-toting 'masters of the universe' to the phd mathematicians in jeans who created the algorithm-only quant funds. The money went to the narrow mispricing models - and the votes went to Obama via the micro-targeting contact programmes.


The big growth area in scientific understanding over the past decade has been in explaining the human mind as a system, a machine, an amazing, complex but ultimately comprehensible sequence of processes. Across Kahneman's work on behavioural economics (demonstrating the tiny-to-huge ratio of analysis-to-associative memory activity) to the neuro-level exploration of our grey matter; from grasping the misleading role of 'narrative' to adapting ever more powerful Bayesian techniques of statistical inference; the most exciting recent developments of science have been about the role of numbers in human behaviour. There isn't space to demonstrate its efficacy here, but there's ample practical, real-world evidence from the worlds of commerce, sport, medicine, and now American politics. The British political party that makes the first serious attempt to grab the data advantage will have the best chance of running the country. That party needs to set up, urgently, a proper analytics capacity and start generating a lot more usable data about voter behaviour.

I'm not disparaging traditional campaigners; they too are in short supply and are hugely valuable. When it comes to turning data-advantage into vote-advantage, you need the strong, purposeful players who actually make things happen on the ground. These are the ones that can usefully play the guru, as their job is to convince and lead teams; but don't mistake the ability to get things done with the ability to first work out what needs to be done.

We certainly need experienced campaigners who can manage operations; and we do need the story-tellers who can craft convincing messages; but what's missing most in our political machines are the data scientists, the only ones who can find out where the missing votes will be found.


You must be logged in using Intense Debate, Wordpress, Twitter or Facebook to comment.