Constitution Daily

Smart conversation from the National Constitution Center

Pollsters’ Trump mistake takes its place in history

November 9, 2016 by NCC Staff


Donald Trump’s convincing victory in Tuesday’s election is the latest black eye for pollsters who undertake the perilous job of forecasting presidential winners.

trumptwitter456In 2012, pollsters overestimated the vote total they projected for Mitt Romney. Four years later, few pollsters gave Trump a strong chance in his general election matchup against Hillary Clinton.

This year, prognosticator Nate Silver gave Clinton a 71 percent chance of winning the election; other forecasters gave Clinton a better-than-80-percent chance of a win. Only two pollsters, Investors Business Daily/TIPP and the Los Angeles Times/USC, saw Trump as a favorite on election day. The other major pollsters, TV networks and newspapers saw Clinton as a clear favorite.

Instead, Trump used support in Pennsylvania, the Midwest and the South to confound the pollsters, and take the general election against Clinton.

Analysts will be debating the methods used by pollsters in the 2016 election for years to come. But in historical terms, if some political polls were truly accurate, Alf Landon would have been America’s president during World War II, instead of FDR, and Thomas Dewey would have defeated Harry Truman in 1948.

Political polls are mostly a 20th century invention. Straw polls were actually started in 1824 in Pennsylvania, when a Harrisburg newspaper forecast that Andrew Jackson would win the popular vote in the general election by a wide margin. (Jackson did, but lost the presidency in the House, since he didn’t have a majority of electoral votes.)

Straw polls were far from scientific, but they were widely used in newspapers and magazine until 1936, when the straw poll went up in flames.

Here are five notable political poll blunders in U.S. history, before the 2016 election.

Alf Landon beats FDR in a landslide

The mother of all botched political polls was a 1936 Literary Digest straw poll survey that said GOP challenger Alf Landon would win in a landslide over the incumbent, Franklin Delano Roosevelt, with 57 percent of the vote.

The Literary Digest used national straw polls in 1920, 1924, 1928 and 1932, and it guessed the winner of each presidential election. In 1936, a young rival pollster, George Gallup, made his own prediction before the magazine issued its poll: He said Literary Digest would get it all wrong, despite the Digest’s decent track record in previous polls.

So was right? The Literary Digest disaster helped establish Gallup as the nation’s pre-eminent pollster. The Digest polled about 2 million people, most of who were magazine readers, car owners or telephone customers—and had money during the Depression. It was not a representative sample. Gallup used a random poll sample of 50,000 people.

President Roosevelt won the 1936 election easily, with 63 percent of the vote,  and the Literary Digest was out of business the following year. If he had won, Landon could have been our wartime president.

Thomas Dewey beats Harry S. Truman

In a replay of 1936, newspaper polls printed headlines about an easy win for GOP challenger Thomas Dewey over the incumbent President Harry S. Truman.

The Roper poll had stopped doing surveys in September about the race. The Gallup poll had the race at 45 percent for Dewey and 41 percent for Truman.

In the end, Truman won by a 50-45 percent vote, and he got to hold up a copy of the Chicago Tribune and gloat a lot.

Gallup and other pollsters survived a wave of negative opinion by adjusting their sampling methods and apologizing to customers.

In a 1998 AP article, George Gallup Jr. explained one valuable lesson his father’s company learned.

“We stopped polling a few weeks too soon,” said Gallup Jr. “We had been lulled into thinking that nothing much changes in the last few weeks of the campaign.”

Bob Dole’s third-place finish in the 1996 Arizona primaries

Three major TV networks said that based on exit polling, Kansas Senator Bob Dole would trail Steve Forbes and Pat Buchanan in the Arizona presidential primary.

But later in the evening, it was apparent that Dole, who later won the presidential nomination, was clearly in second place.

So what happened?

CNN later reported that Buchanan followers targeted the exit pollsters and tried to influence their work. CNN, CBS and ABC used the same polling company, and that magnified the error.

All three networks apologized.

Who’s the president: Gore or Bush or Kerry?

In one of the great controversies in modern politics (and TV news coverage), the TV networks called the presidential race for Al Gore, then George Bush, and then for no candidate after exit polls indicated Gore had won Florida—and the 2000 presidential election.

The same service that did the 1996 Arizona primary, the Voter News Service, called Florida for Gore before polls had closed in the Sunshine State – and it was all downhill from there. The U.S. Supreme Court finally decided the 2000 election with its Bush v. Gore decision.

In 2004, a new service replaced the Voter News Service. Early exit poll results from the National Election Pool showed John Kerry had won Ohio and New Mexico. President Bush actually won those states, which gave him a victory margin.

A Clinton wins in New Hampshire

In the 2008 New Hampshire presidential primary, Barack Obama was projected by pollsters to win easily over Hillary Clinton.

How off were the polls in that primary?

The consensus of seven polls taken just before the primary had Obama winning by 8.2 percent in the voting. Instead, Clinton beat Obama by a 39 to 36 percent margin. So the polls were off by a staggering 11 percent.

And in a twist, exit pollsters said Hillary Clinton would have finished second if another Democrat had been allowed into the primary: Bill Clinton was the choice of 56 percent of the voters, if he had been eligible to run for a third term as president.


Sign up for our email newsletter