Market Size, Preference Externalities, and the Availability of Foreign Language Radio Programming in the United States

Abstract

Using data for 320 radio stations operating in the 50 largest Arbitron metro radio markets during 2004 that offered at least some programming in 1 or more of 19 different foreign languages, strongly positive statistical relations were found between the size of foreign language populations in the radio market and the amount, or variety, of radio programming in their respective language that is available. A preference externality effect was also found: consistently negative relations between the variety of foreign language programming available and size of the English language population. Similar results were found for a measure of programming quality: the percentage of news and talk programming that is locally produced. Conventional wisdom that minority populations tend to be “underserved” by media is generally supported.

Notes

¹The Federal Communications Commission has also acted in the past in an attempt to preserve certain radio formats, such as classical music.

²The Bacon Directory collects information by surveying stations over the previous year, generally between February and October in the case of 2004 (private correspondence with a Bacon Directory spokesperson, Spring 2006).

³We report rankings of the Arbitron markets and their consistent counties as of early 2006 in terms of their population aged 12 and over, as reported on the Arbitron Web site (www.arbitron.com). The Arbitron radio market ranking and definitions change very little year by year.

⁴See http://www.radionewsweb.com/2005-02.html

⁵Among the 320 foreign languages stations we studied, 13 were classified by the Bacon Directory as “non-commercial” (6 full-time Spanish stations, 5 part-time Spanish stations, 1 split between Italian and German, and 1 carrying 11 different languages). We include these stations in our statistical analysis, although their objectives are presumably less focused on market forces. First, Berry and Waldfogel (1999) showed in another study that public radio stations tend to have a significant crowding-out effect, which means that their existence tends to substitute for the presence of profit-making stations that would otherwise enter the market. Second, the contributions mechanism presumably insures that viability of these stations depends on their attracting a substantial audience. In any case, exclusion of these stations (which represent > 5% of the total foreign language station group) did not affect our findings.

⁶See Pearlman (2006).

⁷Although the Federal Communications Commission has been vague about the definition of “underservice,” notices of inquiry and notices of proposed rulemaking related to localism and minority ownership have made clear that the commission is concerned that the minority residents who do not share the majority taste are not being served through the market system. Hence, we examine the level of underservice by comparing the number and percentage of foreign language population with the amount and percentage of foreign language radio programming in individual markets. We do not consider the audiences' willingness to pay. Similarly, previous legal scholars have relied on comparing the percentage of minority population and the percentage of minority ownership of radio stations to describe the level of minority ownership (Singleton 1982).

⁸The eight omitted language are Creole, Yiddish, Albanian, Gaelic, Welsh, Gujarati, Tamil, and Amharic. Among these, five had no available census data on population size, and six did not have income and commute data used in our models. The number of stations in our study remained the same at 320 because all 8 of these languages were offered along with 1 of the other 19 languages and, thus, was of evidently minor significance.

⁹We were able to find detailed information at the Web sites of seven such multi-language stations in our sample about the actual proportions of programming hours devoted to the different languages. These few cases varied greatly, indicating no suggested proportional allocation. We know, however, that if a given language has one full-time station and one part-time (e.g., a 2-way language split), for example, the correct number of full-time equivalent stations is > 1 and < 2, so that 1.5 is the best estimate.

¹⁰The FL-COMMUTE and FL-INCOME variables are defined for ethnic, rather than foreign language, groups specifically. The commute and income data are available only for ethnic groups. Although ethnic groups are larger than their corresponding foreign language groups, the income and commute data of ethnic groups are the closest proxy we can find for those of foreign language groups.

¹¹If we assign a value of zero to %LOCAL when there is no full-time FL station in a market, this could confuse with the case when 0% of news and talk is made locally. However, when there are some full-time FL stations in a market, the range for %LOCAL is from 50 to 100. In other words, when some full-time FL stations are present, there is no such case where, on average, 0% of news and talk is made locally for a market. Hence, we assigned a value of zero to %LOCAL when there is no full-time FL station in a market.

*p < .10.

**p < .05.

***p < .01.

*p < .10.

**p < .05.

***p < .01.

*p < .10.

**p < .05.

***p < .01.

*p < .10.

**p < .05.

***p < .01.

¹²The apparent lack of robust significance for cross-effects in the Spanish sub-sample in Table 6 suggests collinearity between Spanish-speaking populations and English-speaking populations. However, the test with the variance inflation factor values rejected the presence of significant multicollinearity.

¹³We used a population that does not speak a specific foreign language (total population – population that speaks a specific foreign language) as an alternative of an English-speaking population to check cross-group effect, and obtained extremely similar estimation results.