Before going into battle, Roman generals would donate a goat to their favorite god and ask their neighborhood temple priest to interpret a pile of pigeon poop to predict if they would take down the Greeks over on the next island. Americans in the nineteenth century had fortune tellers read their hands and phrenologists check out the bumps on their heads. Statistics came along by the late 1800s, then “scientific polls,” which did something similar. Nancy Reagan turned to star gazers. By 2010, the corporate world was falling in love with Big Data. Today, IBM touts the value of analytics and assigns thousands of experts to the topic, giving them the latest tool in artificial intelligence: a computer named Watson. Whether pigeon poop, private polls, or powerful processors, people have long shared one common wish: to know the future.
But in our Age of Science and our reliance on quantification, the pigeons were not going to cut it. We needed something more accurate, and that was the poll. What does the American experience teach us about polls? Are the statisticians really better at what they do than the temple priests of old?
It did not start out well. Back in the 1820s, with rough-and-tumble frontiersman General Andrew Jackson running for president, the East Coast Establishment media did not know what to make of him. Reporters started going to bars, hotel lobbies, and even on trains to ask people what they thought of him. They then wrote up the results. That was the birth of polling in America. As large corporations emerged after the 1870s, retailers and manufacturing companies applied the new field of statistics to surveying (polling) in order to understand their customers’ views so as to feed their advertising and marketing campaigns. It worked pretty well, and so polling came into being by the end of the century. It was only going to be a matter of seconds before someone figured out that these tactics could be applied to politics.
In 1916, the Literary Digest emerged as one of the most highly respected public opinion surveyors, correctly predicting one presidential election after another. But then its expert political pollsters predicted in the fall of 1936 that Alf Landon, Republican governor from Kansas, would beat incumbent Franklin D. Roosevelt for the presidency. The voters did the unthinkable: 27 million voted for Roosevelt and 16.7 for Landon, who only carried two states. Roosevelt won by a landslide, earning 523 electoral votes versus Landon’s paltry eight.
What happened? The pollsters had sent out surveys to 10 million people, all of them subscribers to the Literary Digest. All drove cars and had home telephones. In hindsight, the problem was obvious: they polled well-off Republicans who, of course, supported Landon. Blue collar workers and poor people did not subscribe to the magazine, own a car, or have a phone. The latter group outnumbered the former, and they voted for Roosevelt since he promised to fix the economy. The magazine apologized with an article entitled “Is Our Face Red!” The Literary Digest went out of business a few years later.
Meanwhile, a young pollster named George H. Gallup did a better job at sampling voters, and so was able to predict the results with greater accuracy. Gallup polls went on to be highly respected across the century. But even he was not perfect. When Harry S. Truman ran for the presidency in 1948, Gallup predicted he would lose spectacularly. He ended up winning.
As recently as 2016, the polls in the United Kingdom consistently predicted that the voters would decide to remain part of the European Union. Brexit turned out to be a disaster for the government. Older voters voted to leave the EU, young voters to stay. Pollsters had failed to poll enough anti-EU voters.
But in each case, pollsters went back to work. Within days they were publishing new statistics which citizens in both countries kept reading, academics and media pundits kept discussing, and politicians kept consulting. It did not matter that there had been spectacular failures. Why?
The history of American polls suggests several reasons. The first is that by the early 1900s polls were more often correct than wrong. Second, as the number of polls increased, they were published in newspapers, later reviewed on radio and television, becoming an ongoing staple of American discourse. Put another way, Americans were increasingly exposed to this type of information during an era in history when statistics were seen as scientific, modern, and accurate. By the 1960s, they could read polling results on all manner of topics: health, sports, eating habits, education, and the odds of living a long life, among others. Most were accurate enough.
But ultimately it was a third reason that made them polling junkies: they wanted to know the future that polls either promised or hinted at. A poll that described your lifestyle and then connected it to statistics on life expectancy told you whether you had a chance to live into your 80s or longer. A poll predicting that “if the election were held today” so-and-so would win the presidency again spoke to the future, as would a series of polls defining a trend.
In the end, Americans were like the Roman general and the Rothschild family: both knew that information conveyed power and opportunity.
Featured image credit: VOTE! by AngelaCrocker. CC BY-SA 2.0 via Flickr.