Prodigy was an early online service that, at its peak in the early 1990’s, had about 1 million subscribing households with about 2 million individual members. From 1992 to 1995, I was involved in an experimental project conducted by the Prodigy newsroom to study the use of email for tracking public opinion over time.
astata2s.pdf is a paper describing the Prodigy experiment and some of its results that was presented at the 1995 conference of the American Association for Public Opinion Research (AAPOR). It was later published in the 1995 Proceedings of the Section on Survey Research Methods of the American Statistical Association and is, to the best of my knowledge, the first paper to describe the use of an online panel to conduct polls among the general population.
During the 1994 general election, Prodigy put up many of what we called “entertainment polls” to attract politically minded subscribers. These polls were entirely self-selected and had no claim to any kind of predictability. During the week leading up to the election, I noticed that some of them included questions that, when combined with the subscriber demographic information I had available, would allow me to apply the same post-stratification we used for the tracking study to the results. Enough responses were received for five high profile state-wide races for me to weight and tabulate them, and Prodigy posted the results on the service the day before the election, duly noting that they had no scientific validity.
We were somewhat surprised at the accuracy of the results but paid little attention to them at the time. More than a year later, I received a compilation of the final major polls published during that election and I realized just how well these online polls did when compared to traditional telephone methodologies.
In 1997, I was asked to participate in a panel on technology and polling at the AAPOR national conference, and I put together an informal talk describing the methods and results of these “unscientific” online polls and comparing them with final published poll results for the same races. aapor97s.pdf contains my speaking notes and the tables I handed out during that talk.
Humphrey Taylor, chairman of Louis Harris & Associates, was in the audience and complained that I had picked one of his polls as a comparison, but, as he told me some years later, he thought about it and then called Gordon Black, chairman of the parent company of LH&A, to tell him that he had just seen the future of survey research. In 1999, LH&A become HarrisInteractive, with online polling as their primary focus. Many others soon followed.
Online polls today have come a long way since then, but they are all based on the fundamental principles established in the Prodigy experiment.