Nielson Media Research Company

Nielsen Media Research supplies the audience data on which television advertising and programming decisions depend. Its electronic people meters can record who actually watches which programs, while its paper-diary sweeps produce local-market ratings; the contrast between those methods reveals both the strengths and limits of measurement. The book’s central institutional point is that even when a more accurate system is available, coordination failures and conflicting incentives can keep an industry tied to a weaker one.

Audience Measurement and the People Meter

In five thousand American homes, Nielsen installed electronic devices called people meters. Each member of a participating household received a unique code and was supposed to use it whenever sitting down to watch television. The meters tracked both the program on the screen and the person watching it, and Nielsen downloaded their data nightly. The measure was therefore designed to distinguish, for example, parents watching one show from a college-age daughter watching another, rather than treating a household television as a single undifferentiated audience.

That distinction mattered to advertisers. A large total audience might interest a company such as Pepsi, but advertisers also wanted to know how many viewers fell within a commercially important age group. The people-meter data supplied demographic information that a simple count of viewers could not. National television advertising relied on those measurements, so the viewing records of five thousand households influenced which advertisements Americans saw and, indirectly, which programs networks had reason to show.

The small sample raised an obvious representativeness question: could five thousand households speak for 120 million people? Nielsen sought to make its participating families a reasonable demographic match for the country. The meters were not flawless; participants could become less conscientious about logging in over time. Even so, they had an important advantage over retrospective surveys: they recorded viewing as it occurred, rather than asking people to recall or report what they had watched. On that basis, the book regards Nielsen’s numbers as probably more accurate than an average public-opinion poll.

The Local Ratings Gap

The same sample that could offer a relatively accurate picture of national viewing could not reliably describe a particular city. The five thousand people-meter households were spread across the country, leaving local advertisers without the detailed demographic information they needed. A health club in Fort Wayne, Indiana, for instance, might want to know which Tuesday prime-time program local viewers aged eighteen to thirty-four watched. The national people-meter sample could not answer that question for Fort Wayne.

Nielsen’s traditional answer was the sweeps system. Four times each year—in February, May, July, and November—it distributed 2.5 million paper diaries to randomly selected people in almost every television market. Recipients were asked to record a week of viewing. Nielsen also gathered information about the diary respondents, enabling it to construct demographic portraits of local markets. Local network affiliates and advertisers then used those portraits to negotiate advertising rates for the months ahead.

The diaries’ scale did not make their results dependable. Only about 30 percent of distributed diaries were returned, leaving room for cooperator bias: the people who completed the survey might watch different programs from those who did not. Respondents often filled in their diaries at the end of the week rather than while watching, so their records depended on memory. Prominent shows were easier to recall than less conspicuous programming, tending to inflate network ratings and depress those of smaller cable channels. Diaries also handled channel surfing poorly. The system thus converted an extensive paper sample into a local measure burdened by gaps between actual viewing and reported viewing.

Incentives and the Persistence of Sweeps

Sweeps distorted behavior as well as measuring it imperfectly. Because advertisers paid on the basis of sweeps ratings, networks had incentives to schedule special, attention-grabbing programming during the survey months. Advertisers could end up paying for ratings produced under conditions that did not represent an ordinary week. Allen Banks, a media executive at Saatchi and Saatchi, called sweeps “a sham, a subterfuge.” Yet in many local markets the sweeps figures remained the only hard audience data available, so advertisers continued to rely on them.

The calendar also encouraged networks to concentrate their strongest programs in the same periods. During a sweeps month, several shows might compete in one hour, although viewers could watch only one. Staggering those programs could have allowed a larger combined audience; head-to-head scheduling instead divided it. The consequences extended beyond the survey months: the concentration of attractive programming in February, May, July, and November left a relative scarcity of new programs in other months, when viewers could face a sea of reruns.

A better local alternative already existed in Nielsen’s people-meter technology, but extending it across the country would be expensive. The meters cost money to install and more to keep running. Nielsen did not disclose the cost, but wiring thousands of homes in each of the nation’s 210 television markets would likely require at least hundreds of millions of dollars and perhaps a billion. Paper diaries, by contrast, cost far less, including postage, because respondents filled them out without pay.

The expense looked less daunting when set against the industry’s scale. About $25 billion in annual advertising was spent on the basis of sweeps data, much of it potentially misallocated. Networks also spent hundreds of millions of dollars during sweeps on programming competition that could reduce their combined viewership. A more accurate local system might therefore benefit the industry overall. But those aggregate gains did not translate into a clear investment case for any one participant: a local advertiser did not spend enough in a single city to justify funding meters there, and an individual network might gain little directly from improved ratings.

The affiliates’ position made the impasse sharper. Their economic attention was concentrated on the months when sweeps measured local ratings, and they were not merely indifferent to people meters. When Nielsen introduced the service in Boston in 2002, not one affiliate signed up; the stations preferred having no ratings to accepting the people-meter figures. The episode shows how a self-interested faction could shape a collective decision, especially when networks and advertisers had not helped fund a reliable local-measurement system and therefore had little leverage over reform.

A Slow Shift, and a Larger Lesson

In 2003, Nielsen announced that it would roll out people meters in the country’s ten largest television markets. The announcement pointed toward change as cable’s importance made paper diaries look increasingly outdated. It did not erase the larger contradiction: a multibillion-dollar industry had remained for years with an inaccurate, backward technology, not because all its participants preferred it, but because they could not coordinate a transition to a collectively better arrangement.

The ratings case belongs to the book’s broader account of collective judgment and cooperation. In other settings, dispersed judgments can become more accurate when a group has mechanisms for aggregating what individuals know. Here, the obstacle was not simply a shortage of information or a failure to invent a measurement device. Nielsen had a more direct way to track actual viewing, but the local data problem, the cost of deployment, and the affiliates’ incentives blocked its widespread use. The contrast is between measurement capacity and the capacity to act together: a technically superior instrument does not become an industry-wide solution merely because its benefits, taken as a whole, are evident.