The History of Polling

Before Science: Straw Polls

Attempts to measure public opinion before elections are as old as democracy itself. The earliest known straw poll appeared in 1824, when the Harrisburg Pennsylvanian and the Raleigh Star canvassed citizens ahead of the Adams-Jackson presidential race. Through the 1800s, newspapers conducted informal 'straw polls,' asking people on the street or at public gatherings how they intended to vote. These had no scientific basis but created a durable tradition of pre-election prediction.

The most famous straw poll was conducted by The Literary Digest, a popular magazine that mailed millions of postcards to automobile and telephone owners before each presidential election. From 1920 to 1932, the Digest correctly predicted every winner. Then came 1936. The Digest mailed 10 million postcards, received about 2.4 million responses, and predicted a landslide for Republican Alf Landon over Franklin Roosevelt, forecasting Landon at roughly 57%. Roosevelt won 46 of 48 states with about 61% of the vote. The Digest's sample was biased: automobile and telephone owners in 1936 were disproportionately wealthy and Republican during the Great Depression. A huge sample size did not compensate for the bias, and the magazine folded within two years.

The Gallup Revolution

George Gallup changed everything. Using a sample of just a few thousand people selected to be representative of the US population, Gallup correctly predicted Roosevelt's 1936 victory and, in a memorable stunt, even forecast that the Literary Digest itself would get the election wrong. The key insight was that a small representative sample is more accurate than a massive biased one. Gallup founded the American Institute of Public Opinion in 1935 and established the principles of scientific polling: quota and later random sampling, standardized questions, and transparent methodology.

Gallup's triumph established polling as a legitimate profession. By the 1950s, campaigns, media organizations, and governments relied on polls for decision-making. The industry grew from a handful of firms to hundreds, and polling became embedded in democratic culture. Today, thousands of polls are conducted during each election cycle, creating both valuable information and considerable noise.

Modern Reminders: 2016 and 2020

Scientific polling did not end the era of surprises. In 2016, national polls actually came fairly close to the popular vote (Clinton led by about 3 points in final averages and won it by about 2), but many state-level polls in the upper Midwest underestimated Donald Trump, in part because they failed to weight by education. In 2020, the error was larger: national averages showed Joe Biden ahead by roughly 8 points, and he won the popular vote by about 4.5, one of the least accurate cycles in decades even though the winner was called correctly.

These misses echo the old lessons. The problem is rarely arithmetic; it is who ends up in the sample and who is missed. Each cycle, pollsters patch the last failure, and each cycle a new source of bias can emerge.

A Timeline of Landmark Moments

Read this table as a single story: every generation of pollsters learns, forgets, and relearns that a poll is only as good as the representativeness of the people it actually reaches.