Association of Music Personnel in Public Radio 
Summer 2000

      At KSJE, Farmington, New Mexico, we use standard methods of audience measurement:  Arbitron, anecdotal evidence, and NPR studies. We also develop our own surveys to examine unique issues in our programming. This isn’t as hard as it looks, providing you follow a few rules. First, determine what you want to find out, create research questions, and operationally define what you want to measure. These rules will help quantify what you want to know. Second, refine a pool of questions. For new surveys, reuse tried queries. 
        Write closed-ended questions. Give people specific choices to select. If you don’t, you’ll have to make categories to quantify responses, which can lead to possibilities for more research but also takes more time.
        Selecting the population to study is very important, because the type of listeners who an-swer the questions can skew your results. Casual listeners may respond differently from core listeners. Choose your participants randomly. This helps insure that extreme scores will cancel each other. Draw at least 100 names from the list. By the time you go through the attrition of unre-turned surveys, you may have the bare minimum number of responses required for statistical theory to work. Which statistical method you use to measure your data depends on what kind of data you have. With nominal, the simplest data, you divide things into discrete categories. “When I listen, I listen to: Rock, Country Music, Classical Music.” You can successfully analyze these data by examining frequency counts, percentages, and the most frequent score in a group of answers.
        Ordinal data: “I listen to classical music all the time, sometimes, occasionally, very occasionally, not at all.” To analyze these results, use all the same methods as for nominal data, plus the median, the score smack in the middle of the data.
        Interval data set an arbitrary zero point and are used to measure a degree of strength. “I listen to classical music often, sometimes, seldom, never.” When you use this measure, you can determine how much more some people listen than others. This cannot be done if ordinal data are used, because a zero point is not set. 
        If you do not know how to generate or statis-tically treat data, you can get help. Your colleagues may have taken statistics classes. University and college research departments will lend a hand. Marketing and business students can take on your surveys as class projects. We’ve had good luck with this approach at KSJE. You can also find statistics books written in plain English. Donald Ary and Lucy Chesser Jacobs’ Introduction to Statistics is a good example.
        When you finish a survey, remember that its results will not give a definitive answer to your research questions. They will indicate how things look at that moment. Compare your data to similar surveys, anecdotal evidence, and your own common sense. 
         Over the years, we’ve discovered that our audience doesn’t like 20th Century music but is “nuts” about Medieval and Renaissance sounds. Why?  Well, as every good researcher says, there’s always room for further study. It  might be a question well worth the trouble to read through and develop categories to quantify. 

Connie Gotsch is Program Director at KSJE, San Juan College, Farmington, 
New Mexico.