There’s been a lot of soul-searching among polling professionals since the 2016 election — why was President Trump’s win such a surprise? Since then, political pundits have been looking for better polling methods.
They may have found one — social circle polling. Basically, instead of asking how someone plans to vote, people are asked how many friends in their social contacts will vote and who these friends would vote for.
Why would this be more accurate? First, people will probably be more honest in their answers if they are asked about others. Second, this widens the data base by getting an indication of voting preferences among those who weren’t contacted. An analysis of 217 surveys beginning in the 1930s that was printed in the January, 2014 edition of the Public Opinion Quarterly found that this indeed was a more accurate method to predict election results. This was also tested in the 2016 US presidential election and the 2017 French presidential election with promising results.
So it will be interesting to see if election predictions start to become more accurate.