How accurate were the polls for the 2016 election?

During recent elections to both the Scottish Parliament and the UK Parliament, opinion polls have been subject to criticism for not accurately predicting the final result.  In Scotland in 2011, polls underestimated the size of the SNP regional list vote, and in the UK in 2015, polls predicted a hung parliament, rather than the resulting Conservative majority.  This section of the briefing discusses opinion polls for the Scottish Parliament election in 2016.

Since the last Scottish election in 2011, the SNP have consistently been in the lead in opinion polls for both the constituency and regional vote, including during 2016 and in the campaign.  However, during 2016, and especially during the campaign itself, the Conservatives began to draw ahead of Labour in the race for second place in some polls.

The figure below shows averages of the last 10 polls and the last 3 polls, compared to the final results.

Opinion polls compared to final results – difference in percentages for each party

Polls

Source: ScotCen Social Research (2016)

Note: As the Green Party only stood in three constituencies, their final constituency vote is not comparable to a national share, so have not been included in the constituency figures here.

Overall, the polls underestimated the strength of the Conservative vote in both the constituencies and the regions, and overestimated the vote of the SNP.  To a lesser extent, the other parties were underestimated on the constituency vote and overestimated on the regional vote.

Polls conducted closer to the election result were more accurate.  In terms of the average of the final three polls, except for the Conservatives, predictions for all parties’ share of the vote was within the standard +/-3% margin for error.  Indeed, the final poll, from YouGov, with a larger sample size than normal, was the most accurate, but still underestimated the strength of the Conservative vote (although it was within the margin for error).  YouGov discussed this final poll and its new approach to sampling after the election.

Rather than the percentages expressed in each poll, much of the press coverage surrounding opinion polls is concerned with what they mean when extrapolated into predictions of the breakdown of seats in the Parliament.  Generally this is done by applying the national swings implied by the poll to each individual constituency and regional contest.

Using this method, both the average of the last 10 polls, and the last 3 polls predicted an SNP majority, with the Conservatives in second place and the Greens beating the Liberal Democrats into fourth.  However, national opinion polls cannot take into account local issues, and the profile of candidates standing in each seat.  And, when the national election results themselves are applied on a uniform national swing basis to each seat, the SNP still receives a hypothetical majority, with the Conservative seats underestimated.  Without constituency-specific polling, or much larger sample sizes, this will always be an issue for opinion polls.