Radio Public Service Announcements and Voter Participation Among Native Americans: Evidence from Two Field Experiments

The CPS includes information on turnout and registration, as well as on a number of key socio-demographic characteristics for tens of thousands of households. Two other large (50,000+ respondents) nationally representative data sets we considered are the National Annenberg Election Study (NAES) (2008) and the Cooperative Congressional Election Survey (CCES) (2012). The NAES includes a substantial number of self-identified Native Americans, although less than the CPS, but does not include Alaska and Hawaii. The CCES includes less than half the number of self-identified Native Americans as the CPS at around 400. Of the three data sets, the CPS can be considered the most authoritative on the subject of voter turnout given its higher response rates and the non-political focus of its questionnaire. None of three data sets distinguishes among tribes or reservations, which are thought to have varying turnout rates (Peterson 1997: 326). The American National Election Study (ANES) has relatively small sample sizes and consequently tends to have far fewer than 50 self-identified Native Americans.

Verba et al. (1995) mention group consciousness as another political orientation. We are aware of no electoral survey that measures Native Americans’ sense of linked fate or other feelings of group consciousness.

There are a few exceptions; most notably Senator Barack Obama visited the Crow Reservation in Montana as part of his 2008 primary campaign (McDonald 2010, p. 260).

The only states with an American Indian or Alaska Native population greater than 5 % are: Alaska (16.4 %), New Mexico (9.5 %), South Dakota (8.2 %), Oklahoma (7.8 %) and Montana (6.5 %) (US Census 2000). 36 % of Native Americans and Alaska Natives live in a designated American Indian area (e.g., reservations) (US Census Bureau 2012a).

Another group of explanations underlines the importance of electoral institutions in explaining registration and turnout differences between racial and ethnic groups. With regards to explaining differences in Native American turnout, examples include whether tribal elections are held on the same day as general elections (increasing turnout), the lack of Native American candidates (McCool et al. 2007) and whether electoral materials are available in Native American languages (McDonald 2010, pp. 45–47).

Evaluations of radio messages on non-political topics suggest they “produce positive changes or prevent negative changes in health-related behaviours across large populations” but that success is substantially increased under certain conditions, for instance when targeting one-off rather than ongoing behaviors, or when the relevant key services and products are easily accessible (Wakefield et al. 2010, p. 1268). Such one-off behaviors are arguably analogous to voting.

The data we use here are from the 2004 NAES (Annenberg Public Policy Center 2008). The question on radio usage in the 2008 NAES focused specifically on whether respondents listened to radio as a source of campaign information.

For example, KGVA 88.1 serves around 25,000 people in an area of 9491 square miles encompassing the Fort Belknap Indian Reservation and surrounding areas in Montana. It explicitly aims to disseminate community information. For instance, it broadcasts live the tribal council meetings, and on its website lists jobs and advertises community events such as book buybacks at the college bookstore. Among other things, its program includes Native American news, culture and music. See: http://​www.​kgvafm.​org/​index.​html.

Levo-Henriksson (2007, Chapter 5) reports that listeners emphasize the benefits of a Native American- (in this case, Hopi-) owned radio station, focussing on its importance for local information dissemination and the expression of Hopi culture and identity.

More information on Koahnic Broadcast Corporation and National Native News is available through their websites: http://​www.​knba.​org and http://​www.​nativenews.​net.

These estimates may be used to assess cost-effectiveness. In 2010, the average cost of airing messages was $460 per station. The average number of Native American registered voters in a station zone is 1324. Raising turnout by 2.3 percentage points using this number implies that an additional 30 Native American voters cast ballots in each propagation zone in 2010, yielding a dollar-per-vote figure of approximately $15. The equivalent dollar-per-vote in 2008 was $16.

Randomization inference simulates a large number of random assignments and for each one calculates the test statistic. The resulting distribution of simulated test statistics becomes the reference distribution to which the actual estimate is compared. See Gerber and Green (2012, Chapter 3) for a description of this method.

The Bayesian updating formulas, which are based on normal priors and normal sampling distributions, are described in Gerber and Green (2012, Chapter 11).