This article is from the source 'guardian' and was first published or seen on . It last changed over 40 days ago and won't be checked again for changes.

You can find the current article at its original source at http://www.theguardian.com/politics/2016/jan/14/election-polling-errors-blamed-failure-speak-tory-voters

The article has changed 2 times. There is an RSS feed of changes available.

Version 0 Version 1
Election polling errors blamed on failure to speak to Tory voters Election polling errors blamed on failure to speak to Tory voters
(35 minutes later)
A new report traces the roots of the pollsters’ failure to predict the Conservative majority in last year’s election to their lack of adequate contact with Tory supporters.A new report traces the roots of the pollsters’ failure to predict the Conservative majority in last year’s election to their lack of adequate contact with Tory supporters.
Earlier speculation about what went wrong has focused on poorly designed questionnaires, a late swing to the Conservatives, a failure of “lazy Labour” supporters to turn out, or reticence on the part of “shy Tories” to reveal their leanings.Earlier speculation about what went wrong has focused on poorly designed questionnaires, a late swing to the Conservatives, a failure of “lazy Labour” supporters to turn out, or reticence on the part of “shy Tories” to reveal their leanings.
But, ahead of Prof Patrick Sturgis’s presentation of the first findings of his postmortem for the British Polling Council next week, new analysis by the leading psephologist Prof John Curtice blows these theories out of the water and suggests the problem was pollsters’ failure to reach the right people. But, ahead of Prof Patrick Sturgis’s presentation of the first findings of his postmortem for the British Polling Council next week, analysis by the leading psephologist Prof John Curtice blows these theories out of the water and suggests the problem was pollsters’ failure to reach the right people.
Related: Polling puzzle: how response times may explain election surpriseRelated: Polling puzzle: how response times may explain election surprise
Crunching British Social Attitudes (BSA) data – a large face-to-face survey that has charted deep tides of opinion over a third of a century – about how people voted, Curtice is able to replicate the Conservative advantage to within half a percentage point – projecting a Tory lead of 6.1 percentage points, against the actual figure of 6.6 points.Crunching British Social Attitudes (BSA) data – a large face-to-face survey that has charted deep tides of opinion over a third of a century – about how people voted, Curtice is able to replicate the Conservative advantage to within half a percentage point – projecting a Tory lead of 6.1 percentage points, against the actual figure of 6.6 points.
This is in keeping with the recent face-to-face British Election Study (BES) that also gauged the result accurately at around 7 points. But it represents a sharp contrast with all the published pre-election polls, which had typically pointed to a political dead heat, and also with the pollsters’ “reconnect” post-election studies, which went back to their final samples post-election and asked the respondents about what they had done in the ballot box. When the pollsters tried this, they barely shifted their numbers, which discounts the late swing theory. This is in keeping with the recent face-to-face British Election Study (BES) that also gauged the result accurately at around seven points. But it represents a sharp contrast with all the published pre-election polls, which had typically pointed to a political dead heat, and also with the pollsters’ “reconnect” post-election studies, which went back to their final samples post-election and asked the respondents about what they had done in the ballot box. When the pollsters tried this, they barely shifted their numbers, which discounts the late swing theory.
Curtice believes the big difference with the BSA data is that – in keeping with the similarly accurate BES survey – is that it is collated on “a genuinely random basis, where everybody has a particular chance of being asked to take part”.Curtice believes the big difference with the BSA data is that – in keeping with the similarly accurate BES survey – is that it is collated on “a genuinely random basis, where everybody has a particular chance of being asked to take part”.
Addresses are picked out by chance, and then researchers go out and make determined efforts to speak to a named individual who is, again, picked out by chance from within each selected household. The interviewers can call round as many as nine times. Such efforts to draw respondents randomly from the population as a whole is very different to online polls, which are drawn from panels of people who have put themselves forward to take part in surveys, and indeed to phone polls, which still depend on a weakening inclination to pick up a landline and talk to a stranger.Addresses are picked out by chance, and then researchers go out and make determined efforts to speak to a named individual who is, again, picked out by chance from within each selected household. The interviewers can call round as many as nine times. Such efforts to draw respondents randomly from the population as a whole is very different to online polls, which are drawn from panels of people who have put themselves forward to take part in surveys, and indeed to phone polls, which still depend on a weakening inclination to pick up a landline and talk to a stranger.
Work by pollsters themselves, including ICM and YouGov, is increasingly also pointing to the awkward explanation that they had the “wrong people” in their samples. But Curtice’s study sheds valuable new light on how they got the mix wrong.Work by pollsters themselves, including ICM and YouGov, is increasingly also pointing to the awkward explanation that they had the “wrong people” in their samples. But Curtice’s study sheds valuable new light on how they got the mix wrong.
He uncovers a new link – independent of social class and age – between the ease with which voters can be got hold of, and their political leanings. Among those respondents whom the BSA researchers succeeded in talking to on their first visit, Labour was 6 points ahead. But among those who required between three and six knocks at the door, the Tories enjoyed an 11-point lead. He uncovers a link – independent of social class and age – between the ease with which voters can be got hold of, and their political leanings. Among those respondents whom the BSA researchers succeeded in talking to on their first visit, Labour was six points ahead. But among those who required between three and six knocks at the door, the Tories enjoyed an 11-point lead.
The Tories are not so much shy as busy with other things, Curtice told the Guardian. The lesson is to “stop doing so many surveys in such a hurry, and instead take a little bit longer to do them better”. If telephone and online pollsters would run their surveys over a week rather than two or three days, he said, this could allow them to send reminder emails and make repeat calls and therefore get at those sort of hard-to-reach people who held the key to election 2015. The Tories are not so much shy as busy with other things, Curtice told the Guardian. The lesson is to “stop doing so many surveys in such a hurry, and instead take a little bit longer to do them better”. If telephone and online pollsters would run their surveys over a week rather than two or three days, he said, this could allow them to send reminder emails and make repeat calls and therefore get at those sort of hard-to-reach people who held the key to the 2015 election.
By looking at turnout, he also reaffirms the pollsters’ own hunch that they may have overstated Labour strength by failing to talk to enough apolitical people. Most of the pre-election polls suggested more voters would turnout than the 66% turnout registered on the day, but using the BSA data Curtice gets close, with a figure of 70%. By looking at turnout, he also reaffirms the pollsters’ own hunch that they may have overstated Labour strength by failing to talk to enough apolitical people. Most of the pre-election polls suggested more voters would turn out than the 66% registered on the day, but using the BSA data Curtice gets close, with a figure of 70%.
Related: The Guardian view on opinion polling: quality before quantity | EditorialRelated: The Guardian view on opinion polling: quality before quantity | Editorial
More particularly, the BSA underlines the pollsters’ failure to sample enough apolitical young people. This led them to underestimate the strength of the link between age and inclination to vote. And, because 2015 was an election where the traditional tendency for voters to move to the right with age was heightened, this in turn led to overstatement of Labour strength vis-à-vis the Tories. More particularly, the BSA underlines the pollsters’ failure to sample enough apolitical young people. This led them to underestimate the strength of the link between age and inclination to vote. And, because 2015 was an election where the traditional tendency for voters to move to the right with age was heightened, this in turn led to overstatement of Labour strength.
Days before he unveils his own preliminary findings of his inquest for the British Polling Council into the debacle, Sturgis would not be drawn on the detail. He did, however, tell the Guardian that Curtice had produced an important piece of evidence that his inquiry has considered. He also pointed to the continuing analytical difficulties of cleanly identifying any one cause of the polling miss. Days before he unveils his own preliminary findings of his inquest for the British Polling Council into the debacle, Sturgis would not be drawn on the detail. He did, however, tell the Guardian that Curtice had produced an important piece of evidence that his own inquiry had considered. He pointed to the continuing analytical difficulties of cleanly identifying any one cause of the polling miss.
The 2015 BSA survey consisted of 4,328 interviews – undertaken by NatCen Social Research – with a representative, random sample of adults, carried out between 4 July and 2 November.The 2015 BSA survey consisted of 4,328 interviews – undertaken by NatCen Social Research – with a representative, random sample of adults, carried out between 4 July and 2 November.