Must We Trust Polls?
Polling, and the pundits who reported on it, were not reliable predictors of the outcome of 2016’s election. Has anything changed?
Did the U.S. media, political analysts and pollsters fix the problems that plagued the 2016 pre-election data and we missed it?
There have been floods of ink on the Mueller investigation, the Mueller Report, the Ukraine, Trump’s impeachment and the Trump administration. Rivers of journalistic ink have flowed about various matters linked, however distantly, with the aforementioned themes.
If the reason polls and experts were so wrong in 2016 was a subject of great journalistic interest, why haven’t there been more stories on this subject?
Instead, longtime Democratic political analyst James Carville is confidently predicting that Democrats, led by Joe Biden, will win in a landslide victory in November. He isn’t alone.
To find out if he is right, let’s return in time to the 2016 election: What did James Carville know in 2016 and when did he know it?
In the whole of the election cycle that year, when exactly did James Carville know that Donald Trump would become the 45th President of the United States?
James Carville, according to reporters who were covering the election on Election Day 2016, knew Donald Trump was going to beat Hillary Clinton at 10 p.m. on Election Night.
He wasn’t alone in that either.
“I’ve never been this wrong.” — Hugh Hewitt. Election Night 2016, 11:45 p.m.
“Do you believe in miracles?” — Jim Accosta, quoting a Trump aide who said before the election that it would take a “miracle” for Trump to win. Election Night 2016, 11:57 p.m.
“How did everyone get it so wrong?” — Politico. Election Night 2016, 12:30 a.m.
“It was like turning a battleship around. The narrative needed to be adjusted…You needed to follow these votes to get there, logically.” — David Chalian. Election Night 2016, 12:45 a.m.
“Trump took down the Democratic party and the Republican party tonight.” — Nicole Wallace, MSNBC. Election night 2016, 3:05 a.m.
“They said it couldn’t happen.” — The New York Times. Election night 2016. 3:16 a.m.
Yes, “they” said it couldn’t happen; that Donald Trump couldn’t possibly win. And they are still saying it.
Are “they”still wrong?
One answer to the question of how so many smart, experienced people could be so completely wrong has been a deft bit of straw-manning. The press has repeatedly reminded everyone that Republican-favoring conservative pollsters, Republican Party leaders, and even the Trump campaign had also been wrong about Trump’s chances.
Others have pointed out that the polls were technically right because Clinton did manage to win the popular vote- though not by as much as predicted.
And they’ve left it at that.
Some in the media have dared asked if a misplaced obsession with polling, especially early polling, might again hand the election to Donald Trump.
“This movie gets replayed again and again, where we believe these polls, and then they’re wrong.” — political analyst Chris Kofinis, CEO of Park Street Strategies. October 2019.
In the first few months after November 2016, there was a certain amount of soul-searching by major news networks embarrassed by their complete failure to anticipate Clinton’s loss.
Finding out if there was a pro-Clinton bias in early polling- and why- seemed like a reasonable place to start. And it was. But from there, assumptions were made that have left polling no more reliable today than it was in 2016.
Democratic analysts haven’t wanted to confront the possibility that polling itself may be fundamentally flawed in its current form. So they haven’t.
The idea that “Polls might not be capable of predicting elections,” is too unnerving for political analysts who make their livings advising campaigns based on polling data and banking on its accuracy.
So instead, comforting theories have been floated, though not proven: At the last minute, undecideds broke for Trump; Trump turned out more voters that expected; polls failed to adjust for differences in education levels; nonresponse bias was to blame.
“Another issue is that the polls underestimated Republicans up and down the ticket.” — New York Times, A 2016 Review: “Why Key State Polls Were Wrong About Trump”.
Some media elites excuse 2016 as an understandable failure to predict how the ignorant, uneducated masses would vote because the well-educated and affluent are more likely to respond in polls. But if that isn’t a systemic flaw in polling, nothing is.
It may well be that Trump voters aren’t likely to respond in polls; that isn’t likely to have changed in 2020.
Pollsters can call this a lack of civic engagement if they’d like. But if Trump voters turnout in massive numbers to reelect the President on Election Day, media elites may have to call it something else.
Responding to polls isn’t civic engagement; voting is.
That bias survives to this day. Just ask any Bernie Sanders voter.
None of these challenges to polling and punditry have been dealt with since Donald Trump took everyone by surprise on Election Day in 2016.
There has been no total overhaul of polling and the ways in which it is reported; there hasn’t even been a minor overhaul.
Polls that show former Vice President Joe Biden with a comfortable lead over incumbent President Donald Trump might be comforting to Democratic voters and viewers.
But unless the problems with polling that upended everyone’s predictions in 2016 are addressed, Democrats might want to get ready for another surprise this November.
(contributing writer, Brooke Bell)