Searching for the Next U.S. President: Differences in Search Engine Results for the 2008 U.S. Presidential Candidates

ABSTRACT

Studies have examined online campaigns, but none have researched the role of search engines during elections. This study analyzes search engine results from the 2008 U.S. presidential election. Analyses show that the sources and content generated by three search engines differed among the search engines. The top-three results from each search engine, however, originated from large, well-trafficked sites. The results indicate that search engines act according to a market model by providing access to mainstream sites rather than diverse, less familiar information. These results allow scholars to understand the types of political information accessible to citizens during elections.

Keywords:

• Content analysis
• election campaign
• Internet
• politics
• search engines

Acknowledgments

The research presented in this article was conducted as part of the author's M.A. thesis. The author thanks Allan Louden for his guidance in conducting the research. She also thanks Sean Luechtefeld for his help with the content analysis and Natalie Jomini Stroud for her assistance in preparing the research for publication.

Notes

1. The study is open to replication. The Web pages studied were publicly available during the 2008 general election. The SPSS dataset used in this study, including the URLs of the Web pages that were studied and the coding variables for each Web page, will be uploaded to the JITP Dataverse.

2. The chi-square analyses indicated that the “Commercial” category for <Obama> and the “e-Government” category for <McCain> had expected values of less than five for each search engine, which violates an assumption of the chi-square statistic (Field, 2005). To avoid violating the assumption, I combined the “Commercial” with the “e-Government” categories for both Obama and McCain to raise the expected cell counts to acceptable levels for chi-square analyses.

3. These categories were collapsed for the <Obama> data, but not for the <McCain> data. Searches for <McCain> generated enough Web pages for each category that the expected frequencies for the categories were higher than five for each search engine, and thus did not violate the assumption of the chi-square statistic (Field, 2005). In order to capture nuanced differences among all four categories for the <McCain> results, I did not collapse the categories as I did for the <Obama> results. The differences among the search engines, were still significant if I did collapse the <McCain> results (χ² (2, n = 240) = 10.67, p < 0.05).