YouGov on why their UK polls were wrong

YouGov have published why they think their polls were out in the UK election. Their conclusions:

The younger age range within the samples over-represent those who are more engaged in politics and are therefore more likely to vote. As younger people, they disproportionately supported Labour, so having too many young voters in our likely voter sample skewed the overall result towards Labour. We believe we had the party voting proportions for this age group correct but that fewer of them actually voted than our sample suggested. This can be corrected in the future in two ways: a) interviewing the correct proportion of people who are less interested in politics, and b) weighting the sample to the expected turnout for different demographic groups. The problem with both of these is that, unlike in the US where detailed exit poll data is publicly available, in the UK no detailed information is available by which we can know the correct target proportions for each age group. However, we can make better estimates of them.

Youth turnout is low almost everywhere. And if you only get to poll the politically motivated youth, then you will over-estimate their likely turnout.

In NZ the Electoral Commission has released turnout by age, so pollsters should be able to take this into account when weighting.

The oldest demographic group, the over-seventies, were under-represented in our samples. They voted disproportionately for the Conservatives, and having too few of them in our samples skewed it slightly against the Conservatives. This can be corrected in the future in two ways: a) interviewing the correct number of over-seventies, and b) weighting the over-seventies in our samples to the correct target weights.

Elderly people vote far more than younger voters.

In NZ only 62% of under 30s enrolled, voted. For over 65s it is over 85%.

One cannot discount misreporting (“shy Tories”), but we can find no direct evidence for it. In this election, polling showed dissonance between the outcome which people (in aggregate) said they wanted, and their underlying party preference. There was a strong overall preference for a Cameron-led government over an SNP-influenced government led by Labour leader Ed Miliband, although stated voting preferences would not have delivered that. It is possible that this led to some respondent misreporting, if people wanted to express their party preference and not their actual tactical vote, but it is impossible to establish this objectively as we can never know how individual respondents really voted.

This is that basically people changed their mind at the last minute as some Labour voters didn’t want Labour propped up by the SNP so chose Conservative as the lesser evil. Same in NZ where some left voters hated the idea of a Labour Government propped up by Kim Dotcom, so voted National.