How many people listen to public radio? The answer depends on whom you ask, and some say the differing estimates should be cause for concern.
The issue surfaced recently when NPR reported an estimated cume of 29.8 million listeners to its 857 member stations, based on its analysis of spring 2006 Arbitron data.
The Radio Research Consortium, however, determined that 27 million people listened to CPB-qualified stations during the same period. That’s 8 percent lower than NPR’s estimate, even though it covers more stations—887.
Estimates were even farther apart in fall 2006, when NPR announced a cume of 30.3 million listeners and Arbitron’s Public Radio Today 2006 reported nearly 26 million listeners—a difference of at least 14 percent.
The numbers all come from Arbitron listening data but diverge as they are handled differently. NPR plays down the disparity, as do Public Radio International and American Public Media. All three networks use ACT 1 software to interpret Arbitron’s Nationwide database and arrive at estimates to be quoted to underwriters. The software vendor, ACT 1 Systems Inc., is independent from Arbitron but is licensed to use the ratings company’s data.
Network executives say many commercial networks use ACT 1, making it the standard for quoting cume and average quarter-hour estimates to advertisers or underwriters. But NPR also includes ACT 1 cume estimates in press releases and in analyses labeled “Official National NPR Program Audience Estimates.”
Researchers in public radio say ACT 1 produces cume estimates that are higher than others because the program does not fully account for people turning to more than one public station in a market. NPR does not disclose this in its releases.
Earlier this year John Sutton, a consultant and former head of NPR’s research department, asked the network to reconcile the discrepancies. In an Aug. 23 letter to Jackie Nixon, NPR’s director of audience and corporate research, Sutton said he was “increasingly getting questions from concerned public radio clients and associates, and occasionally, colleagues in commercial broadcasting, about why NPR’s audience estimates are so much higher than those from Arbitron and the RRC.”
Sutton sent copies of the letter to NPR’s top executives, to NPR Board Chair Tim Eby and to Arbitron execs. He also aired his concerns in an Oct. 18 post on his blog, radiosutton.blogspot.com.
“Allowing radio networks to control the conversation about the size of their audiences is unacceptable in this age of shrinking audiences, greater competition for ad revenue, and higher expectations for transparency and accountability,” he wrote.
The differences in cumes stem from the methods used by ACT 1 and by RRC and Arbitron.
The key issue is how the different procedures minimize double-counting of people who listen to more than one public radio station.
To calculate cumes, RRC and Arbitron use what they call respondent-level data—that is, raw responses from diaries filled out by Arbitron’s listener sample. Data from individuals reveal whether they report tuning to specific stations—and whether they would be counted in the cumes for more than one pubradio station, for instance.
Arbitron corrects cume data to account for duplication of listeners and gives the cumes to RRC, which adds them into a systemwide cume.
That cume is “a closer representation of reported behavior” than the numbers that the networks get from ACT 1 calculations, says Joanne Church, RRC’s president.
ACT 1 software starts with market-wide listening data already processed by Arbitron, without information on multistation listening by individuals. The application provides cume and AQH estimates when its user specifies a station, daypart and geographical area, according to audience researcher David Giovannoni.
As a result, listeners to more than one public radio station in a market are likely to be counted in the cumes of several stations. “Any system such as ACT 1 that uses aggregate data faces these limitations,” Giovannoni says.
ACT 1 tries to correct the duplication by applying a formula that lowers cumes. But audience researchers in public radio say the formula does not lower cumes enough. Pubradio stations are more likely than commercial stations to share listeners because they often air the same programs or other programs of similar appeal.
Gary Davidson, president of ACT 1 Systems, acknowledges this but says the remaining duplication is not “that bad.” He declined to quantify the disparity.
Davidson points out that ACT 1 AQH estimates, which are more often quoted to advertisers than cume estimates, are close to those from other sources.
Sutton first questioned NPR about the differences in its cume estimates when the network began using ACT 1, in 2001. He also brought up the matter last year at a research summit hosted by NPR but says he did not get a response until he put his concerns in writing.
Nixon addressed the disparities in an Oct. 16 memo to top execs at member stations. “ACT 1 is licensed by Arbitron and its methodology has been accredited by the Media Ratings Council,” she wrote. “It is the only program available for processing the Nationwide database that aggregates total estimates for and across NPR’s various program and station lineups and can satisfy NPR’s underwriting needs.”
Nixon cites several reasons why NPR uses ACT 1 figures. RRC could not generate an AQH that had currency with underwriters, she says, and Arbitron’s Maximi$er Plus, which the ratings company used to generate its own cume estimate, is unable to handle data for multiple dayparts, programs and stations.
Network execs also cite the need to represent their audience to advertisers. “We’re not selling against PRI and APM,” Nixon says. “We’re selling against other networks, and we’re selling against other media buyers. We want to use the same standards and, in this case, that’s ACT 1.”
Giovannoni acknowledges that ACT 1 suits NPR’s purposes in selling underwriting. “But numbers that are good enough for sales need not fulfill the requirements of honest science,” he says.
“NPR acknowledges in this memo that its cume estimates are overstated,” he adds. “And when an organization chooses inflated numbers as its standard when honest numbers are available, the issue turns from the utility of selling time to the veracity and integrity of the organization itself.”
NPR has “full confidence in its usage of ACT 1 estimates,” Nixon said, and will continue to use them. “I’m not sure it would help anyone if you had one number coming from one source of information and another number coming from another source of information,” she said. “It would lead to greater confusion on everybody’s part.”
Meanwhile, the issue could become moot in coming months. Arbitron now plans to give ACT 1 access to respondent-level data, possibly by spring of next year.
Web page
posted Nov. 13, 2006
Copyright 2006 by Current Publishing Committee
