Pollsters ‘herding’ to blame for vote mistakes
POLLSTERS “herding” together could be one reason that they failed to accurately predict the outcome of last year’s General Election, according to an official report.
Investigators say that they found a surprising lack of variability across the polls.
Herding describes a phenomenon where firms produce results that closely match one another.
The independent inquiry, set up by the British Polling Council and the Market Research Society, found that the main problem was that companies polled the wrong people.
This meant that they systematically over-represented Labour voters and under-represented Tory supporters.
The weighting that all polling organisations do – to ensure a sample of 1,000 people more accurately reflects the voting public – did not effectively deal with the original errors, preliminary findings show.
But the report’s authors were also unable to rule out “herding”.
They added, however, that it does not necessarily imply malpractice on the part of polling companies.
Labour peer Lord Foulkes said that the findings “vindicated” his call for the sector to be regulated.
The report found that in the run up to last May’s vote the size of the Conservative lead was significantly under-estimated, with all the final polls suggesting a dead heat.
The findings will be presented at the Royal Statistical Society in London.
Other factors, such as question wording and order, made at most a “modest contribution”, according to the report.