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/commentisfree/2016/jan/19/polls-guide-political-polling
The article has changed 4 times. There is an RSS feed of changes available.
| Version 2 | Version 3 |
|---|---|
| Flawed though they are, the polls are still the best guide we've got | Flawed though they are, the polls are still the best guide we've got |
| (7 months later) | |
| The long awaited initial findings of the British Polling Council into the failing of the 2015 general election polls are finally out. | The long awaited initial findings of the British Polling Council into the failing of the 2015 general election polls are finally out. |
| The main culprit, according to the council, is unrepresentativeness in the composition of the election polls samples. The polls systematically over-represented Labour voters and under-represented Conservative voters. And any weighting and statistical adjustments applied by polling organisations were not effective in mitigating these errors. | The main culprit, according to the council, is unrepresentativeness in the composition of the election polls samples. The polls systematically over-represented Labour voters and under-represented Conservative voters. And any weighting and statistical adjustments applied by polling organisations were not effective in mitigating these errors. |
| Other factors, such as overestimating turnout, were either a consequence of the original sampling sin or secondary (any signs of a late swing to the Tories were not major) or unsubstantiated (the order and type of questions, for example: it is unclear how answers to the preferred prime minister can be more reliable than a straightforward voting intention question when it is the same people answering the two queries). | Other factors, such as overestimating turnout, were either a consequence of the original sampling sin or secondary (any signs of a late swing to the Tories were not major) or unsubstantiated (the order and type of questions, for example: it is unclear how answers to the preferred prime minister can be more reliable than a straightforward voting intention question when it is the same people answering the two queries). |
| The findings are not particularly surprising, and were heavily trailed by Number Cruncher Politics analysis, and research by the British Election Study, the British Social Attitudes survey and YouGov. | The findings are not particularly surprising, and were heavily trailed by Number Cruncher Politics analysis, and research by the British Election Study, the British Social Attitudes survey and YouGov. |
| Yet some of these issues still persist. We need only to look at the raw data of national election polls to realise that samples are still too Labour heavy, or at EU referendum polls to notice the difference between phone polls, which show the remain vote ahead by 20 points, and online ones that point to a tight contest. The difference between phone and online EU polling is so great that they cannot both be right. | Yet some of these issues still persist. We need only to look at the raw data of national election polls to realise that samples are still too Labour heavy, or at EU referendum polls to notice the difference between phone polls, which show the remain vote ahead by 20 points, and online ones that point to a tight contest. The difference between phone and online EU polling is so great that they cannot both be right. |
| The council will not be releasing detailed recommendations until its final report (due in March). In the meantime, in order to make samples more representative, there are wo main areas of possible intervention: better raw samples and/or better adjustments weighting. | The council will not be releasing detailed recommendations until its final report (due in March). In the meantime, in order to make samples more representative, there are wo main areas of possible intervention: better raw samples and/or better adjustments weighting. |
| Fixing sampling issues is not simple, especially when pollsters are asked to produce a poll in 24 hours. Bringing together a decent random sample in a day is improbable and impractical. | Fixing sampling issues is not simple, especially when pollsters are asked to produce a poll in 24 hours. Bringing together a decent random sample in a day is improbable and impractical. |
| It should also be noted that budgets for election polling are minuscule compared to the majority of what the market research industry spends, while solutions such as the more accurate exit poll – which cost £200,000 – are beyond the budget of most UK newsrooms. | It should also be noted that budgets for election polling are minuscule compared to the majority of what the market research industry spends, while solutions such as the more accurate exit poll – which cost £200,000 – are beyond the budget of most UK newsrooms. |
| Although more can be done to reach out to under-represented groups, most of the focus is likely to be on adjustments and weighting to make the sample more closely match the wider population. And this will undoubtedly involve risks if the composition of the electorate changes dramatically by the next general election in 2020. | Although more can be done to reach out to under-represented groups, most of the focus is likely to be on adjustments and weighting to make the sample more closely match the wider population. And this will undoubtedly involve risks if the composition of the electorate changes dramatically by the next general election in 2020. |
| One of the reasons that methods which worked in the past failed last May is that Britain went from a two-and-a-half party system to a multiparty one between 2010 and 2015. Polls were unable to capture how votes flowed to Ukip, where Lib Dem voters went, and that a loud, but, electorally speaking, small, subset of young voters are not representative of a broader age group that tends to vote in smaller numbers compared with the electorate as a whole. Should the UK’s political system once again undergo significant changes over the next four years then adjustments, weighting and assumptions based on the past could once again prove ineffective. | One of the reasons that methods which worked in the past failed last May is that Britain went from a two-and-a-half party system to a multiparty one between 2010 and 2015. Polls were unable to capture how votes flowed to Ukip, where Lib Dem voters went, and that a loud, but, electorally speaking, small, subset of young voters are not representative of a broader age group that tends to vote in smaller numbers compared with the electorate as a whole. Should the UK’s political system once again undergo significant changes over the next four years then adjustments, weighting and assumptions based on the past could once again prove ineffective. |
| The truth is that there is no magic bullet solution. It is important to remember that polls – even the most accurate ones – always carry uncertainty and a margin for error. Moving forward, becoming better at communicating uncertainty should be key when reporting and analysing polling – and not just in the weeks that follow a polling debacle. | The truth is that there is no magic bullet solution. It is important to remember that polls – even the most accurate ones – always carry uncertainty and a margin for error. Moving forward, becoming better at communicating uncertainty should be key when reporting and analysing polling – and not just in the weeks that follow a polling debacle. |
| Despite these health warnings, surveys remain the best way of “predicting” an election (they get it right more often than not), and, more important, as a tool for understanding public opinion and views on often crucial issues. | Despite these health warnings, surveys remain the best way of “predicting” an election (they get it right more often than not), and, more important, as a tool for understanding public opinion and views on often crucial issues. |
| In light of the May polls, coverage of the recent Oldham byelection notably paraded different alternatives to polling – from commentators to betting markets and even word on the street– but they were all wrong. They had one important difference from polling: polls are transparent and can be scrutinised. The same cannot be said of gut instinct. | In light of the May polls, coverage of the recent Oldham byelection notably paraded different alternatives to polling – from commentators to betting markets and even word on the street– but they were all wrong. They had one important difference from polling: polls are transparent and can be scrutinised. The same cannot be said of gut instinct. |