![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
Over 10% of the American electorate lives in counties served by out-of-state media because of the mismatch between media markets and state boundaries. Frequently, these “orphan” counties face a different information environment than others in their home state: they receive no news coverage and political advertising for their own statewide races, irrelevant information pertaining to candidates in the neighboring state who will not appear on their ballots, or both. With a combination of county-level, individual-level, and political advertising data, our analysis evaluates the effect of orphan county residency and irrelevant political information on political participation. Results indicate that orphan counties have lower turnout rates than non-orphan counties and that this difference is explained by lower levels of interest in the campaign stemming from exposure to irrelevant information.
The authors would like to thank Paul Freedman, Julie George, Grigore Pop-Eleches, John Sides, and the anonymous reviewers at Political Communication for their helpful comments and suggestions.
Notes
1. The authors acknowledge that media markets frequently cross state boundaries (Althaus & Trautman, p. 828), but their analysis does not address the issue.
2. CMAG did run a two-state pilot study in 1998. It has not released the 2006 data yet.
3. We refer to our units as counties even though we are considering counties, parishes, and noncounty areas common in Virginia (though they exist elsewhere) that exist beyond counties. Hawaii and Alaska are omitted.
4. We classified the “native” portion of media markets based on population. Whichever state within a media market held the plurality of the voting age population was considered the “native” state.
5. Following others (CitationAbramowitz, 1988; CitationShaw, 2006), we transform our measure of campaign advertising into logged advertising to acknowledge the diminishing nature of its assumed effect at higher values (i.e., the 500th ad is thought to have less impact than the first). We do not differentiate between candidate, party, and interest group ads since our main concern is distinguishing between those ads that are relevant for a voter's electoral decision and those that are not. Also, because the natural logarithm is undefined at the value of zero, we added one to the number of ad airings before the algorithm was taken.
6. We also tested 3-, 7-, and 10-day measures. In line with the CitationHill et al. (2008) analysis, we found that the 3- and 10-day measures were weaker than the 5-day measure. The 7-day measure yielded results that were virtually identical to the 5-day measure.
7. These distributions are for orphan and non-orphan counties in states with Senate and/or gubernatorial races.
8. The coefficient indicating the effect of relevant ad airings in non-orphan counties is negative (−0.2) but not statistically significant.
9. One reviewer suggested that the effect of irrelevant information, as well as the orphan county effect that we have identified here, might capture a general sense of alienation that residents of such locations feel because they live so far from their state center. To test this alternative hypothesis, we created a dummy variable to identify counties that lie along a state or international border. As CitationCho and Nicley (2008) point out, borders are not only “sites of division but also of interaction” (p. 3), which means residents of border counties are more likely to shop, work, and socialize in a neighboring state. As a result, residents of such border counties are more likely to integrate elements of another state's (or country's) identity and to feel less like they belong to their own state. When this border variable was included in the models discussed here, it did not significantly change the results.
10. Previous research has found logged measures of advertising are also appropriate for individual-level analyses (CitationStevens, 2008).
11. The two variables are highly associated (χ2 = 649.5, p ≤. 001), and we combine them because of their inherent comparability and for purposes of simplifying the presentation of our inferences.