Restoration of Confidence: Polling’s Comeback from 1948

By Stephen Earl Bennett

Warren Mitofsky’s death late last summer was a blow to students of elections, public opinion, and polling and survey research. Mitofsky pioneered exit polls inside and outside of the United States, and, with the late Joseph Waksberg, he helped improve random digit dialing (RDD), the method most widely used to contact survey respondents today.

Those of us who utilize surveys and polls tend to forget how close this field came to coming to an end after the fiasco of 1948. That was the year pollsters such as George Gallup, Archibald Crossley, and Elmo Roper—pioneers of scientific polling—predicted Thomas Dewey would defeat Harry Truman in the presidential election. A funny thing happened on the way to Mr. Dewey’s inauguration: As Mitofsky himself once reported, Gallup’s and Crossley’s estimates were off by nine percentage points, while Roper missed the election’s outcome by nineteen.

In addition to prompting a loss of public confidence and a flurry of cancelled newspaper subscriptions, the pollsters’ failure to predict the outcome of the 1948 presidential election accurately led to a congressional investigation and at least two scholarly books dedicated to assessing what went wrong (one edited by Norman Meier and Harold Saunders, the other by Frederick Mosteller and colleagues), and proposing improvements.

As Amy Fried has recently reminded us, critics of polling—chief among them Lindsay Rogers—were quick to pounce on the mistakes made in 1948. In his 1949 book, The Pollsters, Rogers—a long-time professor of law at Columbia—not only faulted polling experts for failing to grasp American political history, institutions, and practice, but also for their misconception of the nature of public opinion, and even how to measure it. The fact that Rogers is little remembered today does not discount the damage done to polling and survey research at the time.

In the wake of the disaster, polling organizations made methodological and administrative changes. Quota polls were replaced by probability sampling. Equally important, perhaps, pollsters began polling right up to the day of an election. (In 1948, a 1998 article by Mitofsky tells us, Roper finished polling in September, Crossley’s last poll was October 18, and Gallup stopped polling after October 28. Election day was November 2. Ending polls this early was problematic, for trends indicated Truman was closing the gap on Dewey.)

Nevertheless, the pollsters’ “comeback” took a long time. As Irving Crespi notes, it wasn’t until 1960—when pollsters predicted that, whether Kennedy or Nixon won, the margin would be very small—that public confidence in polling was restored.

Among others, Warren Mitofsky was instrumental in that restoration of confidence. He made many contributions, of course, both to that process and to the polling field; his stints as president of the American Association for Public Opinion Research and the National Council of Public Polls ought not be sloughed off. Still, it is more than sufficient here to consider his contributions to exit polling, for which he is most recognized, and, especially, to the improvement of RDD sampling.  

In 1967, after doing survey work for the U.S. Census Bureau, Mitofsky began conducting exit polls for CBS News and became widely known as “the father” of exit polling. At CBS, Mitofsky developed the projection and analysis systems used to conduct national exit polls. Later, he headed the consortium of news organizations—under the umbrella organization, Voter Research and Surveys—that did exit polls in the 1990 and 1992 contests. In 1993, he founded Mitofsky International, which did exit polling in countries such as Russia, Mexico, the Philippines, and elsewhere. From 2003 until his death, Mitofsky International and Edison Media Research conducted exit polls in America for the Associated Press and the national TV networks.