The release of a new Marquette University Law School Poll showing President Trump losing badly to former Vice President Joe Biden 51%-42% in a hypothetical matchup in Wisconsin generated screaming headlines across the country.
Given that the Badger State was the state that clinched the presidency for Trump in 2016 and is considered key to his re-election effort, this poll suggests that he can't win Wisconsin again in 2020 and as a result, might not be able to win the White House again.
This would be devastating news to the Trump campaign...if the exact same poll didn't show in September of 2015 that Hillary Clinton would trounce Trump 50%-36% in Wisconsin.
Such is the problem of predicting an election 14 months out, especially when it is still unclear who the Democratic nominee will be. All polls, even the best ones, are merely snapshots of a race on the days that pollsters were in the field. There are an infinite number of variables that can alter that snapshot on any of the hundreds of days between the day it is taken and Election Day.
In fact, most polls in recent presidential elections have predicted the exact opposite of what happened 14 months later. For instance, in September of 2015, Hillary Clinton held a commanding lead in RealClearPolitics' average of polls in a hypothetical matchup against Trump.
Ahead of the 2012 election, six of the 13 earliest polls featuring a hypothetical matchup between President Barack Obama and challenger Mitt Romney--from April of 2011 to January of 2012--showed Romney leading. A CNN/Opinion Research survey from October of 2015 showed Romney ahead by five percentage points at 50%-45%. A little more than a year later, Obama won the election with a 51%-47% margin in the popular vote.
Even though Obama first won the presidency in 2008 by a comfortable 53%-46% advantage over John McCain, early head-to-head polls from November 2007 to January 2008 all showed McCain leading--two polls by as many as 12 points. One FOX News poll from December even had him up by 19!
In fact, polls taken during the primary ended up being so inaccurate that they predicted Hillary Clinton to win the Democratic nomination over Obama in a landslide. According to FiveThirtyEight, in the first half of 2007 Clinton had an average lead of a whopping 15% and looked like a shoo-in...until the Iowa Caucuses proved that she wasn't.
The last time a Republican president stood for re-election--2004, when George W. Bush defeated John Kerry by a comfortable but close 51%-48% margin--the earliest polls recorded in the RealClearPolitics average all showed Kerry holding a comfortable lead in a hypothetical matchup.
Like Obama in 2007, early polls in 2003 predicted that Kerry would lose the nomination to Joe Lieberman, who dropped out of the race without winning a single primary.
In crowded Democratic primaries such as this year's, early polls have been markedly unable to pick the winner accurately. FiveThirtyEight noted that in the second half of 1991, eventually primary and general election winner Bill Clinton was running fourth behind Jerry Brown, Mario Cuomo, and Douglas Wilder.
In 1988, Michael Dukakis won the Democratic primary but in the first half of 1987 was trailing Jesse Jackson and Gary Hart, who dropped out of the race in May before re-entering in December and then dropping out for good in early 1988 without winning a primary.
Ahead of the 1976 election, future President Jimmy Carter was in last place with an average of just 0.8% support in polls conducted in the first half of 1975 and dipped to 0.5% in the second half.
Obviously, the science of polling has come a long way in the past 44 years, but early polls are still notoriously unable to predict winners in either primary or general election contests.
There's no reason to believe that early polls will be any different this year.