In 2017, Ipsos Mori found that 43% of people didn’t trust pollsters – an all-time high.[nice quirk to start with] ...For its phone polls, Ipsos Mori calls a mixture of landlines and mobiles that have been selected at random. What surprises me, sitting in the call centre, is just how hard it is to convince people to participate...some interviewers resort to a stiff, robotic delivery. Others make an effort to keep the interviews lively: one young man with jaunty diction, emphasises different words as he sets out the options on whether Britain will leave the EU without a deal: “Would you say it’s very unlikely, fairly unlikely…” It reminds me of Who Wants To Be A Millionaire?....
It can take four or five days of polling to fill the final quota: 18- to 24-year-old men...It is no surprise, then, that almost all voting-intention polls now take place online. It is cheaper, easier and quicker [but you don't know who actually fills in the form]. Each agency has its own online panel: a group of people – more than 800,000, in the case of YouGov’s UK panel – who have agreed to take part in surveys, on politics or anything else, in exchange for a small payment....pollsters adjust their results through a process known as weighting: if they needed to poll 120 people aged 25 to 34, but could find only 100, for example, they will make each response in that age group count for 1.2 people....[OK, but if ther are only 50 people to extrapolate?]
The things that determine how people vote are always changing: to get a representative sample, pollsters must now pay particular attention to age, education, and, of course, views on Brexit.
Polling firms weight their results for turnout, too: they take into account what someone says about how likely they are to vote or whether they voted in the previous election.... Most polling firms weight their data according to how people voted in the previous general election – but it is impossible to weight by a vote for the Brexit party, because it only launched in April....“In my professional life, I’ve never known a time when voters have been so promiscuous with their votes,”
“We came to the conclusion that the underlying methodological problems of voting-intention polls are so fundamental that we can’t solve them,” ...{another commentatror said] “In any case, if I predicted election results with a pin, I’d get it right half the time,” he says. Polling does have its uses, he concedes: it is effective at broadly gauging views on a wide range of subjects. But “the worst thing to use it for is predicting the results of elections.”
Guardian journalists need to be told. There are promises of further (pretty crackpot) dark arts:
Delta uses something it calls “emotional resonance scoring” (ERS). A traditional poll might ask someone how far they agree with a statement; ERS tries to determine the intensity of their feeling, by measuring how quickly – and therefore how emphatically – they respond to it....Other firms are also experimenting with cutting-edge techniques to complement traditional methods. They show people speeches by politicians, then measure their heart rate or monitor their facial expressions, which are then recorded using a webcam and analysed by an algorithm to see how intensely they react to the different messages. But when I speak to people in the industry, the development they appear to be most excited about is a kind of modelling called multilevel regression and post-stratification (MRP), which YouGov used in the 2017 general election....In 2017, YouGov polled about 50,000 voters a week, across the country. With those results, it created different voter types...By estimating how many people in each voter type lived in a constituency, it was able to forecast a result for every seat.
Politicians seem especially gullible, although we only have the pollsters' word for that
In 2014, 12 days before the Scottish referendum, the Sunday Times ran a YouGov poll showing a two-point lead for independence. It was only one poll – an outlier – but it rattled Britain’s political establishment and compelled David Cameron to launch a final intervention: he promised to give the Scottish parliament more powers if Scotland voted to stay in the UK. And it was encouraging polls in the early months of 2017 that helped convince Theresa May to call a general election, with disastrous consequences.
No comments:
Post a Comment