FAIR USE NOTICE. This document contains copyrighted material whose use has not been specifically authorized by the copyright owner. The CHANCE project is making this material available as part of our mission to promote critical thinking about statistical issues. We believe that this constitutes a `fair use' of the copyrighted material as provided for in section 107 of the US Copyright Law. If you wish to use this copyrighted material for purposes of your own that go beyond `fair use', you must obtain permission from the copyright owner.

Copyright 1996 The New York Times Company
The New York Times

April 17, 1996, Wednesday, Late Edition - Final

SECTION: Section D; Page 1; Column 5; Business/Financial Desk

LENGTH: 940 words

HEADLINE: In a Recount, Cyber Census Still Confounds


Millions of Internet users have vanished without a trace since last summer. Millions more have reappeared since December. And a couple of million disappeared again just yesterday, when Nielsen Media Research revised its estimates of Internet use among adults in the United States.

More than half a year after the most ambitious attempt to count the number of people using the Internet, Nielsen and its academic advisers -- looking at the same set of data -- appear to be hopelessly split on how to interpret the numbers. Depending on whom and when one asks, there were 22 million adult Americans on the Internet last August. Or 19.4 million. Or 16.4 million.

Meanwhile, no one knows for sure what the population of cyberspace is in April 1996.

And after months of a surprisingly rancorous dispute over the normally dry science of statistical weighting, the companies that commissioned the survey have come up with their own conclusion: The results of last summer's head count simply do not matter any more. The companies intend to keep on doing what they have been, pumping billions of dollars into the Internet and its World Wide Web on the assumption that if they build it, the multitudes -- whatever the number -- will come.

"At this point, we don't want to be involved with the bickering over which statistical technique is accurate," said Asim Abdullah, acting executive director of Commercenet, a consortium of more than 130 high-tech companies interested in Internet commerce, which commissioned Nielsen's original report. "Our members have other things to worry about."

"Yeah, we're in a hurricane, and they are arguing about whether the wind is blowing 150 miles an hour or 120 miles an hour," said Mark Resch, general manager of advanced technology business services for Xerox Business Services in Palo Alto, Calif. "The argument is intellectually interesting, and it totally misses the point. Activity on our Web site is up 10 percent a month, steadily."

Based on an ambitious telephone survey last August that was designed by Nielsen Media Research and Profs. Donna L. Hoffman and Thomas P. Novak of Vanderbilt University in Nashville, the researchers initially reported that there were 22 million adult Internet users in the United States.

The numbers, which more than doubled previous estimates of Internet users, were quickly embraced by companies hoping to strike it rich in electronic commerce.

But in a departure from the normally staid realm of academic debate, Professor Hoffman in December publicly disavowed the Nielsen report, saying that its conclusions were invalid. She said her own analysis of the survey data indicated that that there were fewer than 10 million Internet users in the United States and Canada combined.

Professor Hoffman said she went public only after Nielsen officials did not respond to her concerns.

But Jack Loftus, a Nielsen executive, said the company was surprised and stunned in December by what he described as Professor Hoffman's "brutal, bitter and unprofessional attack on us."

"That really hurt us," Mr. Loftus said. "It caused us all kinds of trouble. Sales came to a screeching halt." Nielsen had been selling the report for $5,000 a copy.

Last Friday night, after a more detailed analysis of Nielsen's raw numbers and methodology, the Vanderbilt professors, joined by William D. Kalsbeek, a researcher in biostatistics at the University of North Carolina, posted a report on the World Wide Web, calling Nielsen's analysis "seriously flawed" and "grossly inflated."

A more reasonable estimate of Internet users, the professors concluded, was 16.4 million. While significantly higher than the 10 million figure Professor Hoffman proffered in December, it still represented 5.6 million fewer users than Nielsen's count. The discrepancy, the professors said, was due to fundamental mistakes Nielsen made in interpreting the raw data.

Further, she and the other professors contended, Nielsen's estimate of 16.9 million American users of the World Wide Web should have been 11.5 million. And Nielsen's conclusion that 1.97 million people had purchased something on the Web should have been 1.51 million, the professors said.

Nielsen, a unit of Dun & Bradstreet that is best known for providing advertisers and broadcast networks with estimates of television viewership, responded with its own Web posting over the weekend, accusing Professor Hoffman of deliberately misrepresenting the company's research and "jimmying the numbers."

And yet, shortly after the professors' posting, Nielsen officials sharpened their pencils and took another look at their own numbers. Yesterday, explaining that it had analyzed the data using a more complex weighting system -- placing more emphasis on age, for instance, and somewhat less on income -- Nielsen released a new cyber census.

While insisting that the original count was also accurate, Nielsen officials said that this new interpretation of the August numbers would yield the following results: 19.4 million Internet users, 14.6 million Web users, and 1.9 million people who had bought something on line.

Sunil Gupta, a business professor at the University of Michigan who also had access to the Nielsen raw data, said it appeared that both sides had made mistakes in their first attempts to measure the size of the Internet.

"Nielsen did a great job in the way they conducted the survey, but they got really sloppy at the end in how they analyzed the data," Professor Gupta said. "Still, whether you agree with Donna Hoffman or with Nielsen, it is probably the best survey among all that have been done."

GRAPHIC: Chart: "Splitting Hairs Over a Head Count"

Using the same survey data, Nielsen Media Research and its principal academic adviser have arrived at different estimates of Internet use in the United States last year. A heated debate began in December, when the adviser -- Donna L. Hoffman, a business professor at Vanderbilt University -- said the Commercenet/Nielsen Internet Demographic Survey may have "significantly inflated" its findings. On Friday, Professor Hoffman and others released their latest conclusions. Nielsen lowered its estimates yesterday. Chart compares findings of Professor Hoffman with those of Nielsen in currently and formerly. (Sources: Nielsen Media Research; Professor Donna L. Hoffman) (pg. D5)