by Robert Turner
Senior Fellow, McCormack Graduate School

Beware the polls of November–especially the ones picking “likely” winners.

Whom to believe? Is Clinton ahead by 7 percentage points (51-44), as the latest NBC/Survey Monkey poll indicates?  Or perhaps by 3 or 2 or 1, as other polls show?

Or, is Trump ahead by 5 points, as the Los Angeles Times/University of Southern California tracking poll suggests? Or is it a dead heat, 46-46, as the new Washington Post/ABC poll indicates?

Variance in the polls is nothing new, but it is exacerbated this year by the challenge pollsters face in figuring out who will actually vote. Will the new voters who supported Trump in the primaries stick with him next week? What about the millions of voters who have an unprecedentedly low opinion of both candidates?

And then there’s the Electoral College. National surveys can be misleading if enough key states cluster behind one candidate to create an electoral vote majority, even if the popular vote goes elsewhere. Just ask Al Gore.

So you would think the watchword would be caution.

Instead, many of the polling operations are raising their risk exponentially, and damaging our democracy in the process, by claiming to know who is “likely” to win. The process is simple: candidates who are ahead usually win, so, to make up an example, a candidate who is four points ahead at a certain date ends up winning four times out of five. So all of a sudden a candidate with a tenuous four-point lead among the voters is touting as having an 80 percent chance, or likelihood, of winning.

This is grossly misleading. It makes it sound like that candidate is almost sure to win, so attention can focus on other races, or the makeup of the next administration. Yet a change of only one or two points in the voting poll – within the margin of error – can change the “likelihood” of winning by 10 or 20 points, or more.

Yet the pollsters continue to play this game, producing results that are incomprehensible. This week, the Princeton Election Consortium still says Clinton has a 99 percent chance of winning, while polling guru Nate Silver has lowered her chances to 71 percent, based on polling that gives her a 3.9 percent lead nationally.

“Likelihoods” that vary by 28 percent, and that can change by double digits overnight, are, to put it mildly, not very “likely.”

Better to pay attention to the polls of actual voters, and, even then, beware.

Robert Turner (MPA, Harvard University) joined the McCormack Graduate School community as the Boston Globe Fellow in 2007, the same year he helped launch Commonwealth Compact. He spent the majority of his career at the Boston Globe serving as State House bureau chief and assistant city editor, columnist, chief editorial writer, and deputy editor of the editorial page.