Reflections: The state of American polling
Political polling in the U.S. began as early as 1824. In the two centuries since, pollsters have gotten some elections embarrassingly wrong, but the practice of polling is here to stay.
We have become obsessed with polls and polling. In the current political cycle, it seems like there’s a new poll every day.
It wasn’t always this way.
In the early years of our republic, we seldom had a systematic sense of public opinion.
Even so, the basic principles of polling are not new. As early as 1824, a newspaper, the Harrisburg Pennsylvanian, tried to predict the winner of that year’s presidential race. Later in the 19th century, the Boston Globe sent reporters to selected precincts to try to forecast final returns. And then, in the 20th century, a magazine, the Literary Digest, began conducting postcard polls to predict various political results.
Things became even more sophisticated by the 1930s. Elmo Roper and George Gallup, two of the most important figures in the polling world, made real strides in the field of market research and public opinion polling.
In my senior year in college, I worked for a man named Edward L. Bernays. I mowed his grass, but also helped organize his correspondence files as he prepared to write his memoir entitled “Autobiography of an Idea.” I didn’t know it at the time, but Bernays, whose wife was a relative of the Freud family, was the pioneer of the field of public relations and had an enormous influence on polling as a way to try to find what people wanted.
Polling seemed to be booming, until the Literary Digest made an embarrassing mistake in 1936. The magazine predicted that Republican candidate Alf Landon would defeat President Franklin D. Roosevelt in his bid for reelection, and it proved terribly wrong. It told people that Landon would win 57.08% of the popular vote, and 370 electoral votes. When the final tally was in, Roosevelt won a monumental victory. He carried every state except for Maine and Vermont, and had a lead of just about 25 percentage points over Landon.
What had gone so terribly wrong?
It turned out that the sample was skewed. The magazine had polled mostly higher-income voters who chose Landon, missing the much larger share of lower-income voters. It had polled its own readers, then readers from a comparable list, then automobile owners and telephone owners, all of whom had higher-than-average incomes. The final result was a mess. And for a short while, polling fell into disrepute.
But polling persisted. People still wanted to know what was going on.
Experts reassured them that polling made sense. George Gallup declared that a poll was not magic, but “merely an instrument for gauging public opinion,” especially the views of those often unheard. And Elmo Roper added, the poll is “one of the few ways through which the so-called common man can be articulate.” In fact, polling really can help recover the voices of ordinary people.
Today, polling has become a way of life. Polls appear every week, sometimes every day. And because they often rely on different samples, they frequently have different results.
In the election of 2016, polls seemed to indicate that Hillary Clinton had a substantial lead, but we all remember the final result. Polls didn’t capture fully the support of Donald Trump’s constituency, and so failed to predict his victory.
More recently, as Robert Kennedy, Jr. dropped out of the presidential race, polls were split over which candidate may benefit most.
Yet for all of the quirks and quibbles, polls do make a difference. They do give us some sense of what is going on in our public, and sometimes private, life. They aren’t magical. They aren’t always correct. But they do detect patterns and trends, and shifting shapes of the body politic, and frequently those make a difference. And whatever our reservations, polls are here to stay.
Allan Winkler is a University Distinguished Professor of History Emeritus at Miami University, where he taught for three decades. He serves on the Board of Directors for the Oxford Free Press.
