Election results 2019: Opinion poll accuracy holds up
After several elections where the polls as a whole were not a good guide to the result, this time they got it right.
The final figures in the BBC poll tracker were very close to the actual result, as the table below shows,
Poll accuracy (GB polls)
|Party||Predicted result||Actual result|
That's a very good performance - just a small underestimate of the Conservative share and a slight overestimate for the Brexit Party, with the other parties on the nose.
Many of the polling companies had individual polls that were close to the result. But first prize should probably go jointly to Opinium and Ipsos Mori, whose final published polls were almost exactly correct.
There was some evidence of the polls narrowing in the final couple of weeks but the polling companies that showed that most were the least successful at estimating the final result.
Furthermore, the polls were also consistently right to point to Conservative strength with Leave-supporting voters.
The biggest swings came in areas that had voted strongly for Brexit in the 2016 referendum - constituencies like Bassetlaw, Dudley North, Redcar and Great Grimsby. In areas that voted strongly for Remain, the Conservative vote share fell.
Still hard to predict seats
What proved much harder was using polls to forecast how many seats each party would win. That is always difficult because of the unpredictable nature of the first-past-the-post electoral system.
In 2017, YouGov's seat projection, using a technique called MRP (multi-level regression and post-stratification), was very successful. But this time around, its final seat analysis wasn't so close. Even where the share of the vote is known, it's not always possible to estimate accurately what the House of Commons will look like.
That's where the exit poll came into its own - getting very close to an accurate prediction of the size of the Conservative majority.