Pollsters called correct results in 2017

Following the polling industry’s failure to accurately predict the outcome of the US presidential election, the UK’s general election in 2015 and the vote to leave the EU in 2016, the study by Kantar’s Lightspeed division analysed the polling conducted and published around 15 major international legislative and presidential elections, finding that the winning outcome was correctly forecast in all 15.

The average polling error for the 15 elections was +/-1.9%, 24% lower than the international norm of +/-2.5%.

The most accurate polling came during the first round of the French presidential election. Compared to the industry average polling error for first round votes of +/-6%, the French polls predicted the 2017 result with an accuracy of +/-0.2% leading to the most accurate results in the country’s history, according to the research.

Germany’s polling industry, meanwhile, performed most consistently, forecasting the result in under +/- 1.5% accuracy for the fourth election in a row. The average absolute error for this year’s federal election was +/- 1.2%.

In the UK, the polling industry predicted the result of this year’s general election with an average absolute error of +/- 1.5%, lower than the +/- 1.9% international average for 2017.

“Polling companies around the world are judged by how accurately their polling projections predict the outcome of elections” said Jon Puleston, vice-president of innovation at Kantar’s Lightspeed. “So an exceptionally consistent year like this is worthy of some recognition.”

Dr Michelle Harrison, chief executive, Kantar Public, said: “Polling is an enormously valuable insight in to the pulse of public opinion. We shine a light on the issues and concerns that make up the story of every unique election and support transparent democracy.”