Abstract
A knowledgeable electorate is one of the fundamental assumptions of a healthy democracy, and yet studies consistently find the public underperforms in tests of political knowledge. In addition, television news exposure is often poorly associated with political knowledge. Explanations for this include the kinds of knowledge measured. Using an experiment nested in a national survey, this study finds television news exposure is a significant predictor of recognition knowledge while exposure to news via the Internet is a predictor of recall knowledge.
Notes
Notes
1 The seminal work on political knowledge by CitationDelli Carpini & Keeter (1996) includes a large number of survey items over the decades with the question wording, but not the response alternatives, provided, thus it is difficult to tell which items were recall, recognition, or some other format.
2 Online news, of course, includes a number of features that blend print, audio, still photography, and video. Nevertheless, print remains a dominant feature of online news, particularly compared to the traditional broadcast approaches.
3 See the study page for additional details: http://www.electionstudies.org/studypages/2010_2012EGSS/2010_2012EGSS.htm
4 For Vice President, the response alternatives were Joe Biden, Hillary Clinton, William Daley, and Leon Panetta. For Chief Justice the response alternatives were John Roberts, David Cole, Anthony Kennedy, and Larry Thompson. For Prime Minister, the response alternatives were David Cameron, Nick Clegg, Tony Hayward, and Richard Branson. The order of the questions, and the response alternatives, were randomized to control for any order effect.
6 Political interest was measured on a 1–4 (high) scale. Partisanship was created from two questions asking respondents to identify their party preference or, if they considered themselves independents, which party they leaned toward. From this, a 1–4 folded scale was created with a 4 representing strong Democrats and Republicans, a 3 representing “leaners,” a 2 representing soft “leaners” and a 1 representing those who, after the questions, continued to consider themselves not aligned with either party. The voter registration item was a “yes-no” response to whether respondents were registered to vote.
7 Cronbach's alpha for the recall index was .72 and was .60 for the recognition index.
8 Data for these and subsequent analyses were weighted to represent the population at large. The weighting variable in all subsequent analyses is C3_weight from the ANES data.
9 These zero-order Pearson product moment correlations among the media variables ranged from a low of .18 (Internet news and paper newspapers) to .45 (paper newspapers and television), all significant at p < .001. A complete report of the correlations is available from the author.
10 Examination of various tests found no serious collinearity issues in these models.
11 The regression models, obviously, excluded education. Another approach was to create interaction terms of the four media variables with education. This approach, however, resulted in high multicollinearity statistics and was therefore discarded.
12 The two unstandardized regression coefficients differ significantly from one another by z-test (Z = 2.19, p < .05).
13 This is not to suggest that the Internet in all its forms is a print medium, but that for many it is supplanting print newspapers as a source of print news, thus leading to different results for the traditional newspaper exposure measure seen in previous studies.
Additional information
Notes on contributors
Barry A. Hollander
Barry A. Hollander (Ph.D., University of Florida) is a professor in the Department of Journalism at the University of Georgia. His research interests include political knowledge, the effects of emerging media, and the role of religion in politics.