I’ve noticed that networks are always proud of how many people were watching their shows, as shown by the Nielsen ratings. Who is Nielsen, and how is he counting the 7.8 million people who were watching the latest reality show? How do they figure out what I’m watching? I have a feeling they ask a few hundred thousand people what they were watching, and just say that therefore, statistically, x number of people must’ve been watching.

–Joe Marcus, El Paso, Texas

You’re one sharp cookie, Joe. Yes, Nielsen ratings are based on a representative sample–they don’t ask everybody in the country. The sample size isn’t a few hundred thousand people, though–for daily ratings, called “overnights,” it’s just 25,000 homes (5,000 for national programming, 20,000 for local stations), which collectively are supposed to reflect the viewing habits of 106 million U.S. households in 210 markets. That leaves a pretty wide margin for error, you might think–and you wouldn’t be the only one. Nielsen Media Research, an offspring of the firm founded in 1923 by market research pioneer Arthur Nielsen, recently began introducing new sampling technology that it hopes will depict local TV viewership with much greater accuracy. Early results have been controversial, mainly because, in highly simplified terms, they consist of the following:

Q: Do people actually watch all this crap?

A: No.

The modern ratings game began in 1942, when Nielsen introduced the Audimeter, the original “black box.” The Audimeter attached to a radio and recorded when it was switched on and what frequency it was tuned to. An updated version of the Audimeter was used to monitor televisions starting in 1950. (Nielsen quit compiling radio ratings in 1964; that business is now dominated by Arbitron.) Black boxes had the advantage of requiring no action by household members but didn’t tell you who, if anyone, was paying attention to the set. To remedy this shortcoming, Nielsen also mailed out diaries to a completely different set of participants, asking them to write down not only what they watched but who watched it. Diarists provided demographic detail black boxes didn’t, but at the expense of reliability–people might err, forget, or (horrors!) decide to jerk Nielsen around.

Numerous improvements to ratings technology have been made over the years, the most important for our purposes being the People Meter, which Nielsen introduced in 1987. The People Meter has buttons that household members are supposed to push when they start or stop watching, providing demographic detail, and is now used by the 5,000 households who are the basis for national ratings. The old boxes and diaries are still used for local ratings in the 55 largest markets (20,000 households), and smaller markets rely on diaries alone.

Despite the changes, Nielsen reports TV viewership as it always has, in terms of rating and share. The rating for a given show is the percentage of households with TVs that tune in; the share is the percentage of households with their TVs switched on that tune in. Data from meters (both old and new) is transmitted by modem each night to a computer center and compiled into a daily report; diaries are kept only during “sweeps months,” which in most markets are February, May, July, and November. Since so much–ad rates, programming decisions–depends on ratings, during sweeps months both networks and local broadcasters flood the airwaves with special shows and reports, some pretty silly, in a transparent bid for viewers. (My favorite line from a promo for a local news special in LA: “Adultery: Does it have merit?”)

But sweeps may be on the way out. Nielsen wants to use the People Meter to generate local ratings and eventually hopes to eliminate diaries in larger metro areas. Broadcasters have resisted the new system, though, because it’s told them things they don’t want to hear. When introduced experimentally in Boston in 2001, Local People Meters (LPMs) showed that cable viewing was higher and broadcast viewing lower than previously thought. More importantly, TV viewing overall was 8 percent lower than reported by the old system. Upon investigation, Nielsen concluded that half to two-thirds of the difference represented TVs that were on but unwatched.

Advertisers found this news interesting, to say the least; broadcast TV executives choked. Network affiliates in Boston have been slow to sign up for LPM-based ratings, but they don’t have much choice–the old meter/diary system has been discontinued there. Other cities will get LPMs soon. More innovations are in the offing, notably Portable People Meters (PPMs), pagerlike devices that record whatever TV, radio, or streaming Internet programs surveyees watch or listen to, wherever they may be. PPM tests to date have shown higher TV-viewing rates than older methods did, not lower. For now, though, I like to imagine the missing 8 percent are taking a nap, playing with the kids, or reading a good book.

Art accompanying story in printed newspaper (not available in this archive): illustration/Slug Signorino.