Why did the forecasts of the opinion pollsters get the result of the British General Election so wrong?
So far, we still don’t know. An official study commissioned by the British Polling Council, being chaired by Professor Patrick Sturgis, is due to report its initial findings in January.
But a survey by the British Election Survey (BES) seems to offer some useful indications. It looks as if the BES approach of random sampling is more accurate than either telephone or online polling used by the pollsters.
Telephone polling has the problems that many households are now increasingly reluctant to pick up their landline and that a growing number of young people do not even have a landline.
Online polling has the weaknesses that it draws on panels of volunteers who may not be statistically representative and that many older people are still not on the Net.
By contrast, BES data is collected from knocking on doors which is obviously more time-consuming and expensive but it appears more statistically representative.
More information here.