We all know about opinion polls, especially during major elections. The statistical sampling method of most polling places is called quota sampling, a technique in which interviewers are given a quota of specific types of subjects to poll. For example, the sampling may involve asking 10 adult men, 10 adult women (both groups over 20 years of age) and 10 teenage (18- to 19-year-old) voters for their opinion on the presidential election. But often these types of polls are not as accurate as they should be, mainly because the sample is not random. (For more about how political polling works, see “Recreational Math.”)

How do television ratings work? Most people associate such ratings with Neilsen Media Research, a group that measures the number of people watching television shows, making that data available to television and cable companies, marketing people, and the media. Overall, like political polling, they use statistical sampling to rate a show. To determine who is watching what shows, they have around 5,000 households that have agreed to be part of a representative sample— a group that is thought to represent the close to 100 million viewers in the United States. The data are collected from installed meters in the households; from there, each member of the household turns a button on and off to show when he or she begins and ends viewing. The data are collected each night and statistically analyzed to determine just what “everyone in America” is watching—or not watching—on television. This one statistical rating determines not only the placement of commercials, but can make or break a program. Such a small sampling of people is also why favorite viewer programs sometimes get cancelled.