by Greg Valliere, AGF Management Ltd.
LETâS LOOK THIS MORNING at a question weâve gotten repeatedly in the past two weeks â why were the polls so wrong, again, in the past election?
JUST BEFORE THE ELECTION, most polls showed Joe Biden leading by 7 or 8 points; he won by just under 4%. The polls were even more inaccurate in some states â in Wisconsin Biden had a double digit lead in October but won by slightly under 1%. In Maine, GOP Sen. Susan Collins trailed throughout the campaign but won by about nine points. There are many more examples.
SO WHAT WENT WRONG? Weâve plowed through lots of autopsies. First of all, we reject the view of some apologists like Nate Silver, who argue that the polls werenât that bad. But they were â in many cases worse than in 2016.
OUR SENSE IS THAT THE POLLING INDUSTRY FACES THREE PROBLEMS:
1. âShyâ voters: This theory is conceded by polltakers, who know that some
respondents donât want to tell a complete stranger that they love Donald Trump. They may say one thing to a polltaker and do something different in the voting booth. The fancy term here is âsocial trust.â Many Trump voters simply donât have social trust to talk candidly to polltakers.
2. Bad respondent samples: Polltakers claim they have accurate samples â just the right percentage of rich/poor, young/old, male/female, etc. But do they? A suspicion persists that the polltakers still donât include enough white males who didnât go to college â the core of Trumpâs base. These people are suspicious of talking to polltakers.
3. Turnout: Polltakers have always had difficulty with this one. They often canât
accurately predict voter enthusiasm on a rainy, windy election day. Trump supporters will turn out in a hurricane, but many Democratic candidates may have lost because their supporters werenât enthusiastic enough to turn out.
WEâRE NOT INCLINED TO INCLUDE ON THIS LIST a diabolical âvoter suppressionâ plot, as Trump and other Republicans allege; polls showing Democrats with leads actually may have motivated Republicans to vote. Nevertheless, this issue will be debated endlessly in the coming months, and itâs easy to understand Trumpâs frustration with polls that showed him hopelessly behind in Wisconsin.
OUR BOTTOM LINE: Throughout the fall, we always chopped Bidenâs poll lead in half, and that unscientific approach seemed to work; he won by about half of the polling lead he had just before the election. Until the polling industry gets its act together and improves its methodology, weâll stick with that rule of thumb â and weâll never fully trust the polls again.