Research into radio audiences has been a part of radio broadcasting from its beginning. In the earliest days of radio, stations were concerned with audiences at a distance and the distant places where the signal of the station could be heard. Stations relied upon motivated listeners to send postcards and letters reporting which programs and stations they had heard.

Newspapers carried stories about the distances at which local stations had reportedly been heard; they also reported which far-distant stations had been received by readers. Consider this report in the New York Times (18 March 1924):

Pope Hears Opera on His Radio and Picks Up a London Station

ROME, March 17 (Associated Press)-The radio receiving set at the Vatican has been installed and Pope Pius already has been "listening in." Last night the Pontiff heard the opera "Boris Godunov" played at the Cos­tanzi Theatre in Rome, and later picked up a London station which was broadcasting.

Pope Pius expressed great pleasure at the clearness with which the sound waves were received, notwithstanding the fact that there was some static interference.

The set at the Vatican is said to be powerful enough to pick up some of the stations in America, and an effort is to be made to hear KDKA (Pittsburgh). Up to the present, however, there has been no attempt made to listen in on other than Continental stations.

This interest in the reach of radio signals continues today among shortwave broadcast audiences. Many stations encourage listeners to write or e-mail them about the shortwave programs and personalities they have heard. When listeners correspond with a station, they are rewarded by receiving colorful photo cards (QSL or "distant listening" cards) featuring favorite performers.

By the late 1920s advertising on radio had grown to the point that advertisers desired to know at a quantitative level the reach of their radio commercials. Broadcasters also needed to learn whether they were charging enough for the advertising opportunities they sold. The result was systematic radio audience research.

The principal questions addressed by audience research were ( 1) who is likely to be in a given station's or program's audience? (2) what is the popularity of a program or station? (3) what is the success of a program or commercial announcement? and (4) what is the probable success of a program or commercial announcement that had not yet been broadcast?

The term ratings is often used by media professionals to refer to all measures of audience listening and sometimes to describe the commercial companies that conduct syndicated audience research. Of greatest weight in the view of broadcasters is the fact that audience research is the principal tool used to persuade advertisers that significant audiences will be delivered for their advertising. The evidence of the future value of a station or network advertising opportunity is measured by the size and composition of the audience provided by particular programming in the past, and these are reported in audience research.

The first national radio survey conducted on a systematic basis took place in 1927 when Frank Giellerup of Frank Seaman Advertising asked Archibald Crossley to study audiences for the Davis Baking Powder Company. In March and April 1928, NBC commissioned Daniel Starch, a Harvard professor and pioneer market research consultant, to conduct an extensive survey east of the Rockies. By the 1929-30 radio season a regular program rating service, Comparative Analysis of Broadcasting (CAB) , had been established, providing routine reports on the audiences of network programs. Significantly, the CAB governing board assured that the Association of National Advertisers controlled the infant rating service. Later, broadcasters rather than advertisers became the prime force in establishing standards and practices for audience research.

Methods of Data Collection

  Audience research must always ask what listeners have heard (which programs, which stations). The methods of collecting this data have evolved over the history of radio ratings. They include techniques known as telephone recall, telephone coincidental, roster interviews, diaries, meters, and recorders.

In a telephone recall survey, a radio listener was called by an interviewer from the research company at some time after a program has been presented. The interviewer asked whether the respondent listened. In March 1930 the CAB interviewed regularly in 50 cities using the telephone recall method. Telephone recall interviewers asked respondents about their radio listening in the previous 24 hours, noting the time of listening, who was listening, and the programs and stations heard.

Telephone interviewers using the telephone coincidental technique asked respondents about radio listening taking place at the time of the call. Their questions revealed whether anyone was currently listening to a radio in the household, which family members were listening, and what programs and stations were being heard. In the early years of radio, this method was associated with the research firm of Clark­ Hooper formed when Montgomery Clark and C.E. Hooper left the Daniel Starch organization in 1934. Later this technique became known as the Hooper ratings. As only one time of listening was researched per telephone call, this style of research was labor-intensive for the interviewer and somewhat expensive. However, it is thought by audience research authorities to be the best measure of audience activity when conducted correctly.

Roster interviews were those in which the respondents were interviewed face-to-face. At designated points in a roster interview, respondents were shown a list of programs and stations and asked to identify those that they remembered hearing within a specified period of time. This method of data collection for audience research was also known as aided recall measurement. The roster method was used by The Pulse during the several decades of its history and is associated with Sydney Roslow, a psychologist who formed the organization in 1941.

The diaries used in audience research are special questionnaires in booklet form in which listeners record their times of listening and the stations or programs heard. This method has the advantage of collecting many times of listening over a given period of time (typically a week). The development of this method is often associated with James Seiler, founder in 1949 of the American Research Bureau. Diaries may be kept by an individual for his/her own listening (individual diary) or by one household member for all of the household (household diary). Contemporary radio audience research in the United States and Canada asks that all individuals in selected households maintain individual diaries, a pattern called flooding the household.

Listening meters were devices that automatically recorded times and tuner settings for radio receivers. The first of these was devised by Robert Elder and Louis Woodruff of the Massachusetts Institute of Technology and first used in 1935 for CBS. As radio meters measure the potential listening in a household by monitoring which receivers are on or off at what times and to which signal each is tuned, they produce household ratings. In 1936 market researcher A.C. Nielsen attended a luncheon at which Elder spoke about the Elder-Woodruff Audimeter. He was impressed and bought out the inventors, later establishing a radio meter rating service known as the Nielsen Radio Index.

Sound recorders are potential audience research tools, and a variety of recording methods have been employed to capture listener behavior. One form of recorder, carried by the listener, tunes to a special inaudible identifying signal in the transmissions of local radio stations. The recorder logs signals received and heard by the listener at stipulated intervals throughout the day. A computer later assemblies this data as a record of listener exposure to stations throughout the day. Experiments with this sort of recorder have continued for decades without producing a syndicated research service built upon this technology. The Arbitron and Nielsen companies have announced their plans to launch a research service using this method to measure audiences of both radio and TV.

Another sort of recorder captured samples of sounds heard by a listener who carried the device throughout a listening period. On a carefully controlled schedule, the recorder sampled sounds in the environment of the radio listener carrying the device. Later a computer compared the sounds recorded by the listener for their match to signals being broadcast in the market at that time. This method requires a relatively unsophisticated recorder but a complex analytical system and was not used in an established system of audience research reports.

Some researchers have proposed that recorders be carried by individual listeners participating in audience research surveys. Others have proposed that stationary recorders record sounds from portable receivers operating in their vicinity, with the data then analyzed to produce audience research reports.

Samples of Listeners

  In the early years of radio broadcasting, samples were typically drawn from phone listings in cities selected because they were served by radio stations affiliated with the networks sponsoring the study. During the 1940s and 1950s there were considerable efforts to produce samples that would be perceived as excellent by researchers in the broadcast and advertising industries.

Different samples are used for audience research with national and local audiences. Because the motive for much research has been to substantiate the value of advertising opportunities, the areas where surveys are conducted are called markets.

Networks, advertisers, and others interested in nationwide entertainment and advertising support national market audience surveys. National surveys must give weight not only to every local market but also to listening in rural areas where national signals may reach. RADAR (Radio's All Dimension Audience Report) conducts the only regular national surveys of the radio audience.

In local market surveys the sample is drawn from three survey areas. The smallest of these is the metropolitan (metro) area, usually a core urban area as defined by the U.S. Office of Management and Budget. The largest area surveyed in a local market survey is called the total survey area (TSA), which typically is an aggregate of units of county size, including the relevant metro area or areas. The TSAs of adjacent markets may include the same counties; for example, the total survey area for City A may include Brown County because an important part of the listening in the county is to City A radio stations. At the same time, the TSA for City B may also include Brown county for the same reason. An "exclusive area" is sometimes included, in which each market consists only of those counties where the plurality of listening is to the market being surveyed. In this system of exclusive areas, any particular county can belong to only one market.

Random Sampling

  In sampling for audience research, random sampling is preferred, because theoretically a random sample maximizes the probability that a sample will be very similar to the population from which it is drawn. A random sample is one in which each member of the target population is equally likely to be chosen for the survey (participants are chosen at random); each choice is also entirely independent of the others. The first criterion­ equal likelihood for selection-requires that the researcher name the "sample frame" from which the sample is to be drawn (i.e., to list all members of the target population). In the early days of radio audience research, telephone directory listings were the sample frames. Listed telephones serve best as sample frames when nearly all residential telephones are listed and when nearly all residences are equipped with telephones. In the early years of audience research the first condition was typically met, but telephone penetration had yet to reach its peak.

    In the past several decades residential telephone listings have become progressively poorer telephone sample frames. Although nearly all residences now have telephones, fewer and fewer have listed telephone numbers. One of the methods audience researchers have adopted to cope with this problem is random digit dialing, a method in which all possible telephone numbers within a target area are listed and the sample frame is drawn from that list. A number of variations on this procedure are used as contemporary sample telephone frames. There are sample frames for households rather than telephones; an example is a city directory. Enumerations of households within census tracts (designated by the U.S. Census Bureau for control of their surveys) also serve as household sampling frames. A quite different approach to sample frames is a frame of clusters, which are sampling units that each consist of two or more interviewing units such as residences. A city block, for example, may become a cluster in a sample frame, as it is a cluster of households or interviewing units.

     In most contemporary radio audience research, sampling procedures are mixed. Thus the initial sampling frame could be a residential telephone listing, later supplemented by a second frame of telephone numbers computer-generated at random.

Producing Ratings Survey Reports

  Whenever a radio audience report lists a rating (percentage of potential audience) or a listening estimate (numbers of listeners), a degree of error is also implied. This is called sampling error, a scientifically determined estimate of the difference between research results if the entire target population were surveyed and those obtained using a sample. The probability that such an error will occur is given by the confidence interval listed for any professional research study. The confidence interval standard in audience research is 95 percent, meaning that if the same study were conducted many times, the same results would be obtained at least 95 percent of the time. Commercially produced audience survey reports include descriptions of the methods used for estimating sampling error and confidence intervals.

     Stations, networks, and program suppliers who use audience data in their sales and planning prefer to receive audience data at a modest cost, so the research suppliers must not spend more on audience research than their customers are willing to pay. The statistic used to compare the cost basis of competing radio audience research reports is price per "listen­ing mention."

     A listening mention is the smallest unit of reported radio listening. It consists of at least five minutes of continuous listening to a certain station or program. One listening mention, then, means that one member of the sample reported listening to a particular station/program for at least five minutes during a 15-minute interval of the time period being surveyed. The telephone coincidental method for collecting listening mentions from audiences is the most expensive method of producing routine audience reports, as any respondent can provide only one listening mention-the one in which the individual was involved at the time of the interviewer's call. The lowest cost per listening mention is nearly always a research method employing a meter or recorder, as analysis of one instrument's data can provide a train of listening mentions over months or even years.

     The number of hours per week over which ratings or audi­ence estimates are provided is another cost factor. If every hour of the week were surveyed for listening, a maximum of 672 listening mentions could be recorded. To reduce costs, audience research firms have typically limited the number of hours reported in their surveys. Thus over the years they have reported listening during prime time only, or they have excluded the hours of lowest listening (such as those between midnight and dawn).

     The break-even cost of an audience survey is reached when the number of clients who will pay for the survey at a designated price meets the cost of collecting, tabulating, and printing the data they are willing to buy. When the number willing to buy increases above the break-even point, then the research company becomes profitable (at times very profitable). This explains why a number of new companies over time have entered into the radio audience research business, though relatively few have survived.

Special Research Studies

      A number of research methods are used to study the desirability of using particular songs or groups of songs within the established format of a radio station or network. Each of the following music audience research methods has its advocates and detractors, but all have persisted in one form or another over the past several decades.

     Telephone call-out and call-in is a method that focuses on recent musical releases and older (but still fairly recent) songs that have remained popular and are still frequently played. Each music selection being studied is prepared as a recorded hook-that is, a representative excerpt of a recording.

     In telephone call-out research, the researcher will have previously identified a pool of qualified study participants (listeners to the station in question or to the categories of music being studied). The interviewer plays over the phone one hook at a time and asks the study participant to respond with phrases such as (1) "I've never heard of it," (2) "I dislike it strongly," (3) "I dislike it moderately," (4) "I don't care," (5) "I'm tired of it," (6) "I like it," or (7) "That's my favorite record."

     Stations and networks that make use of call-out research conduct their studies weekly or semiweekly. Satellite programming services making music available to stations throughout the country conduct this research on a continuing basis in many markets.

     In one form of telephone call-in, study participants receive a letter in the mail asking them to participate in the study. The letter identifies a telephone number for participants that connects them to a recording of the hooks for the study. The participant listens to the hooks when convenient, then returns the questionnaire by mail or telephones a researcher who writes down the responses read by the participant from the completed questionnaire.

     Auditorium studies are surveys often used to study audience response to "oldies" (songs that were quite popular during the more distant past and that are still popular with at least part of the radio audience) and "standards" (new versions of oldies). Researchers recruit a sample of radio listeners, who assemble in an auditorium or rented meeting room. Hooks of the music are played over high quality sound systems. Participants then mark their responses to each hook in questionnaire booklets or on digital responders (keypads that summarize responses into a convenient computer file). Auditorium studies commonly include hooks for large numbers of recordings, greatly reducing the cost per hook for respondent data. Because of the effort required to assemble hooks, arrange for facilities, and recruit respondents, auditorium studies are conducted less frequently than call-out or call-in studies.

     The relatively low costs of mail and internet surveys make these surveys appealing. However, mail surveys require that study participants recognize songs from written descriptions that may include the name of a song or performer or some words from a song's lyrics. This limitation often leaves researchers wondering whether the music has been correctly identified by study participants. Internet surveys permit playing hooks over a respondent's computer speakers, reducing the possibility that study participants will not recognize the music being studied. Survey questionnaires are then presented for completion on respondents' computer screens. When all answers have been provided, the data is immediately returned to the researcher's computer. As not every radio listener has access to a computer connected to the internet, study participants must be identified in a pre-survey as (1) listeners accessible by internet, and (2) listeners to the radio music being studied.

Radio Program Format Research

     Radio program format research concerns the mixture of music, news, and talk programming that is best for a given station or network. Although a number of methods are used in format research, focus groups have received the greatest attention in the literature about radio.

     A focus group is a group of research participants who are selected for their relevance to the matter being studied. Their viewpoints and opinions are collected with a guided conversation about the research topic. In the case of format studies, one strategy is to recruit a variety of groups-those who listen only to the station in question, those who listen to the station some­ times, and those who never listen to the station but by their media habits show that under some circumstances they could become listeners.

     The results of station format focus group studies identify the "position" (reputation) of the station in its market. In addition, the specific language used by study participants during the focus sessions may suggest useful slogans or themes for station promotional campaigns.

     Because many radio stations today are owned by corpora­tions that own large groups of stations, station format studies may be a matter of researching format issues in several markets simultaneously. The results may lead to the choice of a station format that will function competitively in all of the group's markets.