Perry, Paul K.

Paul Keely Perry was born on February 22, 1910, in Camden, New Jersey, and grew up in Philadelphia, Pennsylvania and Rochester, New York . He graduated from Tufts University in 1933.

Perry joined the American Institute of Public Opinion in 1935 not long after it was founded by George Gallup . The next year it made history not only by predicting correctly that Franklin Roosevelt would defeat Alf Landon but also by estimating the magnitude of the error the Literary Digest would make in its final release before the 1936 election.

In 1942 Perry moved to Audience Research, Inc., a Gallup affiliate, where he pioneered a multivariate approach for calculating the expected gross revenue of a motion picture. It was based on surveys regarding the public’s awareness of and interest in the film, its marquee appeal, and audience enjoyment as measured with the then new Hopkins Televote meters used by viewers to register their impressions of what they saw on the screen.

A turning point in Perry’s career came in the wake of the failure of the polls to anticipate the last-minute surge of support for Harry Truman in 1948. Even though he was still at Audience Research, Perry was asked by George Gallup to set up a system for the upcoming Congressional elections in 1950 that would pick up any late shifts in voter sentiment. Perry had always regarded the preelection poll as one of the most rigorous tests of the survey method and rose to the challenge.

To capture late trends and identify likely voters, Perry developed a design that included a pair of matching surveys within selected precincts. The first survey was conducted in the first week of October; field work for the second wrapped up the Friday of the week before the election. Results from the second survey were telegraphed to the Princeton offices of the Gallup Poll . The studies were comparable in terms of selected precincts but were not panel studies. Perry also devised a turnout scale based on several questions relating to voter participation.

Gallup was impressed with the rigor of Perry’s design and analysis and approved release to subscribing newspapers of a projection that the national vote for Congress would divide 51 percent Democratic and 49 percent Republican. Gallup also approved Perry’s recommendation that the release include the projection of a 28-seat gain for the Republicans. Perry based this projection on his analysis of regional shifts in voting intentions. Once the votes had been counted, Perry’s estimate of the division of the national vote was off by less than a percentage point, and his 28-seat figure was exactly on target.

Rather than rest on his laurels after the election, Perry turned immediately to a postmortem to validate the relative predictive value of the turnout questions that had been used by checking precinct records to determine which respondents had actually voted.

After serving as vice president of Audience Research and chief statistician of the Gallup Poll, Perry became president of the Gallup Organization in 1958. His new responsibilities did not diminish his efforts to improve preelection polling methods. He continued to explore one issue after another. He tested alternative ways of asking the candidate preference question. He experimented with split samples to shed light on whether respondents who marked “secret” ballots were less likely to be undecided than those who told interviewers which candidate, if any, they supported. He then used the difference to determine how to allocate undecided respondents in the final preelection poll. He designed procedures to integrate data from the last preelection survey and results from the more comprehensive early October survey. This made it possible to check for changing voter preferences while retaining a basis for robust estimation of likely turnout and an election’s probable outcome. And as noted, he continued postelection validation studies to gauge the efficacy of different questions and approaches in scaling respondents’ likelihood of voting.

What is noteworthy is that as he blazed a trail, he left markers for those who have followed. Perry was very forthcoming about the procedures he was developing. His articles in Public Opinion Quarterly went into considerable detail about sample design, turnout estimation, timing of surveys, allocation of undecided respondents, and use of the secret ballot.

In this same spirit, Perry was the featured speaker when George Gallup invited directors of state polls to Princeton the day before the 1951 American Association for Public Opinion Research (AAPOR) meeting. Mervin Field of the California Poll recalls this as the first of several “Local Pollers Conferences” that sprang from the openness of Gallup and Perry about the work they were doing. It imbued in him and others attending a sense of the importance of sharing developments with colleagues that would advance the cause of all polling.

Perry was responsible for all Gallup preelection polls from 1950 through 1976. Over the years Perry also oversaw dozens of non-election studies. One that particularly engaged his considerable imagination was the collaboration with the National Opinion Research Center on what would become Samuel Stouffer’s classic, Communism, Conformity, and Civil Liberties (1955).

Perry is remembered for his generosity and quiet manner. He spoke often of all he learned from others and the confidence with which he could turn projects over to associates at the Gallup Organization, particularly his longtime deputy and close friend, the late Irving Crespi . Those who worked with Perry reciprocated by striving to meet the high standards he set-especially for himself.

Perry’s contributions provide ample evidence that there is nothing so practical as a theory if it is applied with care and that nothing advances a theory so much as putting it to the test in the real world.

Paul Perry passed away April 7, 2005 at the age of 95.