Atomik Research

The Creative Market Research People

Read All About It

Archive

Failure to Predict Trump’s Strength: What Went So Wrong in the Polls?

polls got it wrong

Just last week we were pretty confident that Hillary Clinton would win the show, so how is it that today we’re left feeling as deflated as we did the morning of Brexit? The Donald has won – he is in power. We’re still in shock, waiting for the news sink in.

In the lead up to Election Day, Clinton was ahead in the polls by about 2 – 4 percentage points depending on where you looked. However, as November 9th approached, the margins tightened after yet another scandal rocked Clinton’s campaign – the private email server she used while secretary of state. It turns out that this ‘scandal’ wasn’t in fact a scandal at all; the FBI concluded that there was no need to press criminal charges. If you have a spare five minutes, watch Sarah Bealson’s dramatic reading of the unmalicious, frankly dull emails that have caused so much controversy.

Unfortunately it seems that the damage was already done. The negative headlines so close to Election Day gave Trump the momentum he needed at just the right time. Clinton’s slogans “I’m with her” and “stronger together” failed to resonate like Trump’s ‘Make America Great Again’ or Sanders’ “Feel the Bern”. Indeed, Trump out-performed in New Hampshire and Pennsylvania, two states that had minimal early voting, and took the key swing states of Florida, North Carolina and Ohio.

As we explored in our previous polling piece, polling isn’t designed to be completely accurate. Due to the heavy media coverage of the electoral race, the polls – and their accuracy – are over-reported and, as Peter Woolley, a professor of comparative politics at Fairleigh Dickinson University in New Jersey said, “We tend to forget very quickly that it’s an estimate within a range.”

Ann Selzer is an Iowa pollster who conducts surveys for Bloomberg Politics. “There was a lot of experimenting this year with the types of questions they were using, the types of methodologies they were using,” she said in an interview at Bloomberg News headquarters in New York. “There’s the continuing barrier of the lack of landlines, the erosion of landlines. In the old days, if we knew your landline phone number we knew where you lived and that was fantastic for pollsters. Now it’s very difficult.”

In addition to technological changes, the polls simply couldn’t keep up with an electoral race that featured so many twists and turns. As unexpected changes took place day by day, public opinion yo-yoed and the polls broke.

An internal review carried out by Gallup Inc. found that four areas contributed to the mistaken predictions during the 2012 pre-election presidential polling: likely-voter estimates, regional representation, weighting of race and ethnicity, and outreach to landline-phone users.

Getting a sample that accurately reflects the electorate seems to be one of the biggest problems of polling predictions. Many surveys had under-sampled non-college-educated whites, a group that Trump appealed to. Geoff Garin, a Democratic pollster who worked for the pro-Clinton super PAC Priorities USA, argued that there was too much emphasis on the idea that the country’s demographic diversity would lead Clinton to the win.

Not only have the pollsters dealt with a decline in participation for decades – from the household participation down to 14% in 2012 from 43% in 1997 – there was no way of the pollsters knowing exactly how many people were going to secretly vote for Trump. Many who did participate in surveys would have been too embarrassed to admit what they were really going to do in the privacy of the voting booth.

Even when the polls did hint at Trump’s strength, they didn’t seem to be taken very seriously. Perhaps this was a way of dealing with the reality of an alternate outcome – an outcome we all hoped wouldn’t come true.

By Bella Foxwell


  1. No comments yet