Television audiences are not measured by counting the actual number of viewers tuned to a program. When you consider that there are approximately 275 million television viewers in the U.S., counting all of them all the time would not be feasible. Measuring  the audience from a sample is not only more efficient, but also allows researchers to collect a lot more detailed information from the people in the sample, and then project that knowledge to a much larger population.

Any system of television measurement depends upon an audience sample that is representative of viewers as a whole. A properly  selected, random  group of households should mirror the behavior and characteristics of the population. By using random probability samples, Nielsen Media Research can project the viewing in its samples to the entire population.

Nielsen Media Research's national sample consists of 5,000 households, including approximately 11,000 persons. We use the U.S. Census Bureau's decennial (updated annually) census counts of all housing units in the nation. Using these data, Nielsen Media Research  randomly selects more than 6,000 small geographic  areas (blocks in urban areas and their equivalent in rural areas) and dispatches surveyors to each area to enumerate  and list housing units. 

Housing units are randomly selected within each sample area. Each occupied housing unit is a household. The sample is designed to give each household in the population a known chance to be selected for the National People Meter Sample. Volunteers, though plentiful, cannot be included in the sample. For more information about Sampling, please visit our web site or What TV Ratings Really Mean.

Again, while all ethnic groups are represented in Nielsen Media Research's national sample in proportion to their percent of the population being measured, only the two largest ethnic groups, African-Americans and Hispanic-Americans are reported  in Nielsen Media Research's standard ratings reports (see African-American Audience and Hispanic-American Audience).