SELECTIVE EXPOSURE IN POLITICAL WEB BROWSING

Abstract

The effect of Internet use on political information exposure is investigated using representative survey data from Japan and the USA. Internet users can simply choose political information that is consistent with their political attitudes. This selectivity in information exposure via the Internet might have serious consequences on the democratic social system, such as fragmentation of shared information and a decrease in political tolerance. Three research questions were empirically investigated as to the presence of selective exposure in political web browsing, the effect of political web browsing on political tolerance, and the contingencies on which selective exposure occurs. Multivariate quantitative analyses show that web browsing, as a form of Internet use for seeking political information, facilitates exposure to arguments that are consistent with one's attitudes. However, selective avoidance, which suppresses exposure to heterogeneous arguments, is not empirically supported. Moreover, although Internet use as a source of information facilitates exposure to homogeneous arguments under certain conditions, it does not have a negative effect on political tolerance. This is because selective exposure to homogeneous arguments takes place if and only if the perceived issue's importance is high. That is, even if selective exposure has an effect on a few issues perceived as highly important, there is no such bias in other less-important issues, which attenuates the effect of selective exposure on the homogeneity of the information environment surrounding each person. It is concluded that the fear of a fragmented society due to selectivity in using the Internet seems to be empirically groundless.

Keywords:

• Internet use
• information environment
• selective exposure
• cyber-balkanization
• political tolerance

Notes

Japan Election Study 3 was supported by Grants-in-Aid for Scientific Research of the Japan Society for the Promotion of Science (JSPS) and conducted by Ken'ichi Ikeda, Yoshiaki Kobayashi, and Hiroshi Hirano. The authors would like to express their greatest thanks for permission to use the data.

http://www.pewinternet.org/PPF/r/141/report_display.asp. Refer to Horrigan et al. (2004) for the report of this survey.

The reason why the number of negative arguments that were not selected by respondents was not tallied was that the non-selection includes the possibility of DK/NA and non-attitudes on the arguments.

Nevertheless, the partial correlation between the number of homogeneous and heterogeneous arguments on postal reform is −0.74 after controlling the effect of the total number of arguments, which indicates that redundancy is still high. Therefore, the main focus of the analysis is on the number of homogeneous arguments on postal reform, regarding the analysis on the number of heterogeneous arguments as supplementary. Consequently, an excessive generalization of its result is avoided.

Among pro-Bush respondents, the reliability coefficient of a scale (the number of homogeneous arguments) composed of pro-Bush arguments and against-Kerry arguments is 0.65, and that of a scale (the number of heterogeneous arguments) composed of against-Bush arguments and pro-Kerry arguments is 0.23. On the other hand, among pro-Kerry respondents, the reliability coefficient of a scale (the number of homogeneous arguments) composed of against-Bush arguments and pro-Kerry arguments is 0.66, and that of a scale (the number of heterogeneous arguments) composed of pro-Bush arguments and against-Kerry arguments is 0.32.

A variable of political knowledge cannot be used because no corresponding variable is available in the US data. However, considering the null effect of political knowledge on the number of arguments (Tables 1 and 2), it does not seem to cause serious analytical problems.

Strictly speaking, it is not desirable to measure as ‘the Internet and e-mail’ which includes e-mailing for personal communication because the focal point here is political web browsing. However, the proportion of e-mailing for personal communication such as exchange of evaluative opinions or mobilization for voting can be assumed to be small because, in the leading sentence, the context of the question is restricted to ‘Where have you gotten MOST of your news and information about the presidential election campaigns? (underline added)’ and thus the emphasis is on the aspect of information exposure of Internet use, not the personally communicative aspect.

The summed scale was broken down because the scale reliability is low (α = 0.31). After breaking it down, the numbers of respondents in each category are high (19.7 per cent), middle (53.1 per cent), and low (27.3 per cent).

For more detailed information about the items, see the questionnaire of the survey. http://www.pewinternet.org/pdfs/PIP_Polinfo_questions.pdf Detailed information about how measurements were constructed is available upon request for authors.