Polling Pitfalls: A 90-Year History of Election Misjudgments
From the Literary Digest's 1936 blunder to modern-day challenges, election polling has a long track record of inaccuracies.
The history of presidential election polling is marked by nearly a century of missteps, from sample bias to flawed turnout models. These recurring errors highlight the difficulty of accurately predicting public opinion and, more importantly, voter behavior.
Ninety years ago, the Literary Digest, which had a consistent record of forecasting presidential election winners, incorrectly predicted Alf Landon's victory over Franklin Delano Roosevelt in 1936. The magazine's poll, surveying 2.4 million people via postcard based on telephone books and car registration lists, suffered from a misleading sample that ultimately led to inaccurate results and the magazine's demise.
Polling science advanced with the establishment of Gallup tracking presidential approval ratings in 1938. However, perfection remained elusive. The 1948 election saw the Chicago Daily Tribune famously declare "Dewey Defeats Truman" based on earlier polls that did not reflect the actual Election Day outcomes.
By the 1970s, pollsters became closely associated with presidential campaigns. John F. Kennedy utilized the Harris organization's polls in his 1960 campaign, marking a new era of reliance on polling firms. Figures like Pat Caddell for Jimmy Carter, Richard Wirthlin for Ronald Reagan, and Stan Greenberg and Dick Morris for Bill Clinton became prominent. Despite their successes, these strategists were not infallible. Caddell's influence on Carter's 1979 "Crisis of Confidence" speech, misjudging the public mood, is often cited as a significant misstep. Dick Morris, despite a tumultuous relationship with Bill Clinton, played a key role in advising the president after the 1994 midterm elections, even employing a secret alias to navigate White House politics.
While presidents like Bill Clinton left office with high approval ratings, subsequent presidencies have seen a trend of lower, more volatile approval numbers. George W. Bush experienced a post-9/11 surge to 86% approval, but this was not sustained. Modern presidents have generally struggled to maintain approval above 50%.
Further polling errors continued into the 21st century. In 2004, exit polls incorrectly predicted John Kerry's victory over George W. Bush. Similarly, Mitt Romney's campaign in 2012 was misled by overly optimistic weighting assumptions in its pollster's analysis, which underestimated Barack Obama's 2008 turnout levels among youth and minority voters.
The 2016 presidential election represented another significant blow to polling credibility, with most pollsters confidently predicting a Hillary Clinton victory over Donald Trump. The justification that polls correctly captured the popular vote but missed the electoral outcome did little to assuage concerns about their predictive accuracy.
More recently, partisan polls in midterm primaries have been criticized for overstating progressive support, leading some candidates to withdraw based on inaccurate data. This pattern reinforces the ongoing lesson that polling frequently falters, though the specific reasons for failure often differ.
Despite technological advancements, the challenge of polling persists. While most people own phones, the widespread use of caller ID and a decrease in response rates from unfamiliar numbers contribute to a growing "silent majority." This environment increases the importance of human judgment in interpreting data and deploying new technologies. Pollsters, in their attempts to avoid past mistakes, can sometimes fall into the trap of "fighting the last war."
Ultimately, polls are often more effective as snapshots of public opinion than as precise predictors of voter behavior. The public is advised to approach poll results with caution, emphasizing that active participation in voting remains the most crucial element of the electoral process.