My article got
published today at the
New Scientist! One of the biggest science magazines in the world.
See the text here (there is no paywall, you just register and read it for free). It was even on the front page:
The text is about the scientific experiment behind
our prediction survey. It starts by examining why the pollsters are getting it wrong lately and whether or not there is any science at all behind polling. Then it introduces our prediction survey idea and how we’re doing an experiment on US elections to see whether or not science can actually improve polling.
For those who don’t bother to register in order to read it on the New Scientist webpage, I have copied the text here (enjoy!):
As US election looms, time is ripe for a new science of polling
“Growing scepticism about traditional methods for predicting election outcomes is fuelling a search for a more scientific approach to polling, says Vuk Vukovic
As the US prepares to vote for its new president next week, narrowing political polls have suggested that this crucial election may be too close to call.
Why might that be? These days, pollsters find it harder to get responses by calling voters on their home phones. A typical telephone survey now has a response rate of
below 9 per cent, with fewer willing interviewees making the polls less likely to be representative of the wider voter population and, hence, less precise and reliable.
Telephone polls are usually carried out during the day, biasing the results towards stay-at-home parents, retirees and the unemployed. Most people, for some reason, do not respond to cellphone surveys as eagerly as they once did to those by landline.
Online polls have their own weaknesses: they tend to be
biased towards particular voter groups, such as the young, better-educated and urban dwellers.
In both types of survey, pollsters try to
compensate for biases, but the results of doing so can be dubious –
as shownwhen four different pollsters gave four different results for the key swing state of Florida in the current US campaign based on the same data set. Furthermore, a
recent study showed that the actual margin of error of a poll’s finding is about 7 per cent, instead of the typically reported 3 per cent. Not surprisingly,
some critics argue that opinion polls are more art than science.
Turning to science
Putting polling back on a scientific footing will require experiments in the coming years, combining insights from various branches of sociology, economics, mathematics of networks and statistics.
However,
we have added a twist that we hope will improve its power to predict an election outcome, in that we go beyond asking people who they will vote for. We also ask who they think will win and their view on who other people think will win. The idea is to incorporate wider influences, including peer groups, that shape an individual’s choice on voting day.
Why might this work? When people make choices, such as in elections, they usually succumb to their standard ideological preference. However, they also weigh up the chance that their favoured choice has. In other words, they think about
how other people will vote. This is why people sometimes vote strategically and do not always pick their first choice, but can opt for the second or third to prevent their least-preferred option from winning.
It is going to take a fair few experiments to answer the question of whether contemporary polling can be considered scientific.
And though the current US election is widely condemned for its negative atmosphere, it provides a good chance for a new science of polling to begin to take shape.
If you are a US voter, you can help Oraclum test its polling method by participating in its survey and sharing it with your friends.”
Vuk Vukovic is a researcher at Oraclum and a PhD student at the University of Oxford
