Published online by Cambridge University Press: 16 December 2011
This study reassesses the ability of the mass media to influence voter opinions directly. Combining data on media content with individuals’ assessments of British political parties during the 2005 general election campaign allows a test of newspapers’ persuasive influence in a way previously considered a ‘virtual impossibility’. Utilizing repeated measures from the 2005 BES campaign panel, multilevel regression analysis reveals significant impact of partisan slant not just on the evaluation of the party mentioned but also on evaluations of its competitor(s). The strongest evidence of direct media persuasion is provided by the finding that variation in slant over the campaign drives how undecided voters evaluate the incumbent government party, even when controlling for a newspaper's average partisan slant.
1 Finkel, Steven E, ‘Reexamining the Minimal Effects Model in Recent Presidential Campaigns’, Journal of Politics, 55 (1993), 1–21CrossRefGoogle Scholar; Bartels, Larry M., ‘Messages Received: The Political Impact of Media Exposure’, American Political Science Review, 87 (1993), 267–285CrossRefGoogle Scholar; Dalton, Russell J., Beck, Paul A. and Robert Huckfeldt, ‘Partisan Cues and the Media: Information Flows in the 1992 Presidential Election’, American Political Science Review, 92 (1998), 111–126CrossRefGoogle Scholar; Paul Allen Beck, Dalton, Russell J., Greene, Steven and Huckfeldt, Robert, ‘The Social Calculus of Voting: Interpersonal, Media, and Organizational Influences on Presidential Choices’, American Political Science Review, 96 (2002), 57–74Google Scholar; Zaller, John, ‘The Statistical Power of Election Studies to Detect Media Exposure Effects in Political Campaigns’, Electoral Studies, 21 (2002), 297–329CrossRefGoogle Scholar; Peter, Jochen, ‘Our Long Return to the Concept of Powerful Mass Media – A Cross-National Comparative Investigation into the Effects of Consonant Media Coverage’, International Journal of Public Opinion Research, 16 (2004), 144–168CrossRefGoogle Scholar; Druckman, James N. and Michael Parkin, ‘The Impact of Media Bias: How Editorial Slant Affects Voters’, Journal of Politics, 67 (2005), 1030–1049CrossRefGoogle Scholar; Norris, Pippa, ‘Did the Media Matter? Agenda-Setting, Persuasion and Mobilization Effects in the British General Election Campaign’, British Politics, 1 (2006), 195–221CrossRefGoogle Scholar; De Vreese, Claes H. and Boomgarden, Hajo G., ‘Media Message Flow and Interpersonal Communication: The Conditional Nature of Effects on Public Opinion’, Communication Research, 33 (2006), 19–37CrossRefGoogle Scholar; Jonathan McDonald Ladd and Lenz, Gabriel S., ‘Exploiting a Rare Communication Shift to Document the Persuasive Power of the News Media’, American Journal of Political Science, 53 (2009), 394–410Google Scholar.
5 Petty, Richard E., Christian Wheeler and Tormala, Zakary L., ‘Persuasion and Attitude Change’, in Theodore Millon and Melvin J. Lerner, eds, Handbook of Psychology. Volume 5: Personality and Social Psychology (Hoboken, N.J.: Wiley, 2003)Google Scholar.
6 Lazarsfeld, Paul E., Bernard Berelson and Hazel Gaudet, The People's Choice (New York: Duell, Sloan and Pearce, 1944)Google Scholar; Berelson, Bernard R., Lazarfeld, Paul F. and McPhee, William N., Voting: A Study of Opinion Formation in a Presidential Campaign (Chicago: University of Chicago Press, 1954)Google Scholar.
7 Bartels, ‘Messages Received’, p. 267. Indeed, only very few studies into media persuasion exist from the 1970s or 1980s. See, for example: Erikson, Robert S., ‘The Influence of Newspaper Endorsements in Presidential Elections: The Case of 1964’, American Journal of Political Science, 20 (1976), 207–233CrossRefGoogle Scholar; Coombs, Steven L., ‘Editorial Endorsements and Electoral Outcomes’, in Michael B. MacKuen and Steven L. Coombs, More than News: Media Power in Public Affairs (Beverly Hills, Calif.: Sage, 1981)Google Scholar; Wagner, Joseph, ‘Media Do Make a Difference: The Differential Impact of Mass Media in the 1976 Presidential Race’, American Journal of Political Science, 27 (1983), 407–430CrossRefGoogle Scholar.
8 Norris, ‘Did the Media Matter?’
9 Bartels, ‘Messages Received’.
10 John Zaller, ‘The Statistical Power of Election Studies to Detect Media Exposure Effects’.
13 Erikson, , ‘The Influence of Newspaper Endorsements in Presidential Elections’, p. 207Google Scholar.
14 Seymour-Ure, Colin, ‘Editorial Opinion in the National Press’, Parliamentary Affairs, 50 (1997), 586–608CrossRefGoogle Scholar; Wring, Dominic, ‘The Tony press: media coverage of the election campaign’, in A. Geddes and John Tongue, eds, Labour's Second Landslide (Manchester: Manchester University Press, 2001)Google Scholar.
16 Gosnell, Harold F., Machine Politics: Chicago Model, 2nd edn (Chicago: University of Chicago Press, 1968)Google Scholar; Erikson, ‘The Influence of Newspaper Endorsements in Presidential Elections’; Coombs, ‘Editorial Endorsements and Electoral Outcomes’; Wagner, ‘Media Do Make a Difference’; Webber, Richard, ‘The 1992 General Election: Constituency Results and Local Patterns of National Newspaper Readership’, Journal of Elections, Public Opinion & Parties, 3 (1993), 205–215Google Scholar; Curtice, John and Semetko, Holli A., ‘Does It Matter What the Paper Say?’ in Anthony Heath, Roger Jowell and John Curtice, eds, Labour Last Chance? (Aldershot, Surrey: Dartmouth, 1994)Google Scholar; Curtice, John, ‘Is the Sun Shining on Tony Blair? The Electoral Influence of British Newspapers’, Harvard International Journal of Press/Politics, 2 (1997), 9–26CrossRefGoogle Scholar; Newton, Kenneth and Brynin, Malcolm, ‘The National Press and Party Voting in the UK’, Political Studies, 49 (2001), 265–285CrossRefGoogle Scholar; Norris, ‘Did the Media Matter?’; Ladd and Lenz, ‘Exploiting a Rare Communication Shift’.
17 Kim Fridkin Kahn and Kenney, Patrick J., ‘The Slant of the News: How Editorial Endorsements Influence Campaign Coverage and Citizens’ Views of Candidates’, American Political Science Review, 96 (2002), 381–394Google Scholar.
18 Deacon, David and Wring, Dominic, ‘Partisan Dealignment and the British Press’, in John Bartle, Simon Atkinson and Roger Mortimore, eds, Political Communications: The General Election Campaign of 2001 (London: Frank Cass, 2002), pp. 197–214Google Scholar.
19 The use of an overly simplistic dichotomy that divides the press into Conservative v. Labour papers may explain why an earlier study of the 2005 British general election found only limited persuasion effects on party and leader evaluations or vote choice, with some of the significant estimates even pointing in the wrong direction; see Norris, , ‘Did the Media Matter?’, pp. 210–213Google Scholar.
22 Page, Benjamin I., Shapiro, Robert Y. and Dempsey, Glenn R., ‘What Moves Public Opinion?’ American Political Science Review, 81 (1987), 23–43CrossRefGoogle Scholar; Fan, David P. and Tims, Albert R., ‘The Impact of the News Media on Public Opinion: American Presidential Election 1987–1988’, International Journal of Public Opinion Research, 1 (1989), 151–163CrossRefGoogle Scholar; Dalton et al., ‘Partisan Cues and the Media’; Gavin, Neil T. and David Sanders, ‘The Press and Its Influence on British Political Attitudes under New Labour’, Political Studies, 51 (2003), 573–591CrossRefGoogle Scholar; Dobrzynska, Agnieszka, André Blais and Nadeau, Richard, ‘Do the Media have a Direct Impact on the Vote? The Case of the 1997 Canadian Election’, International Journal of Public Opinion Research, 15 (2003), 27–43CrossRefGoogle Scholar; Peter, ‘Our Long Return to the Concept of Powerful Mass Media’; Druckman and Parkin, ‘The Impact of Media Bias’; De Vreese and Boomgarden, ‘Media Message Flow’.
24 Fan and Tims, ‘The Impact of the News Media on Public Opinion’; Kleinnijenhuis, Jan and Fan, David P., ‘Media Coverage and the Flow of Voters in Multiparty Systems: The 1994 National Elections in Holland and Germany’, International Journal of Public Opinion Research, 11 (1999), 233–256CrossRefGoogle Scholar; Dobrzynska et al., ‘Do the Media have a Direct Impact on the Vote?’
26 De Vreese and Boomgarden, ‘Media Message Flow’.
27 Dobrzynska et al. , ‘Do the Media have a Direct Impact on the Vote?’, p. 31Google Scholar; see also Lazarsfeld et al., The People's Choice; Kleinnijenhuis et al., ‘Issues and Personalities in German and Dutch Television News’.
28 O'Keefe, Daniel J., Persuasion: Theory and Research, 2nd edn (Newbury Park, Calif.: Sage, 2002), p. 3Google Scholar. See also Walton, Douglas, ‘What is Propaganda, and What Exactly Is Wrong with It?’ Public Affairs Quarterly, 11(1997), 383–413Google Scholar, p. 394; and Garsten, Bryan, Saving Persuasion: A Defense of Rhetoric and Judgment (Cambridge, Mass.: Harvard University Press, 2006)CrossRefGoogle Scholar, pp. 2ff.
30 O'Keefe, ‘Persuasion’, p. 4.
31 O'Keefe, ‘Persuasion’, p. 5.
32 We selected seven of the largest and most prominent newspapers to provide a sample of the British media landscape that accounts for variation in newspaper type (broadsheets, red-tops and black-tops) and partisan tendencies (Labour and Tory papers as well as newspapers with more ambiguous endorsements like the Independent).
34 Elliott, Larry, ‘Why Alastair Campbell made Blair eat humble pie: Polls proved that only Gordon Brown could deliver a third Labour term’, Guardian, 21 April 2005, p. 26Google Scholar.
35 Brandenburg, ‘Party Strategy and Media Bias’.
36 See Kahn and Kenney, ‘The Slant of the News’.
37 We standardized text lines because columns width tends to vary within and between newspapers. Our standardization procedure consisted of multiplying the number of text lines of any data entry with its observed column width and then dividing by the average column width across the entire dataset.
38 This design also allows us to assess the impact of the sheer volume of positive or negative evaluations, or the trade off between positive and negative reporting (in other words, whether, e.g., 20 positive versus 20 negative statements have a comparable effect to that of two or three positive and negative points). These measures were included in initial analysis but require further exploration and are not included in the analyses presented here.
39 The use of internet panels remains subject to criticism because of limitations in the sampling and selection procedures of some of these panels, which it is said might result in violations of analytical assumptions or possible bias in the findings. However, in an extensive survey comparison, the team of researchers that conducted the 2001 and 2005 British Election Study addressed most of the contentious issues. They argued that, while on the one hand sampling problems remain with internet surveys because internet access is not randomly distributed, internet surveys on the other hand address some core problems with face-to-face surveys, for example by eliminating interviewer effects. They also found that while marginal distributions of key variables differ across both types of survey, when estimating parameters in vote choice and turnout models, face-to-face and internet surveys yield remarkably similar results. Also, issues of over- or under-representation of social groups appear to be dealt with effectively through weighting procedures in internet surveys carried out by Yougov. See Sanders, David, Clarke, Harold D., Stewart, Marianne C. and Whiteley, Paul, ‘Does Mode Matter for Modeling Political Choice? Evidence from the 2005 British Election Study’, Political Analysis, 15 (2007), 257–285CrossRefGoogle Scholar. Most importantly, however, our aim is to assess possible attitude change through media effects, which does not require our sample of newspaper readers to be nationally representative. For this reason, the graphs presented below, as well as subsequent regression models, are based on data that are not weighted to reflect nationally representative readership figures.
40 Because of an error in the programming of the questionnaire for Wave 1, respondents were not actually able to give parties a score of 10, but instead could only give a maximum score of 9. In our analysis, this is addressed by applying a scale factor to the Wave 1 scores.
41 Data included in our analysis comprise content analysis of op-ed pages only, since only these articles were coded for slant. The entire 2005 newspaper dataset is much larger and also contains content analysis of all news stories about the campaign (see Brandenburg, ‘Party Strategy and Media Bias’ for a description of the entire dataset).
42 Daily average tone scores were calculated from the coded text for each party in each newspaper (n = 26). Accordingly, the boxplots indicate the mean and quartile range calculated from twenty-six individual scores per party.
43 van der Brug, Wouter and van der Eijk, Cees, ‘De campagne deed er toe, mediagebruik niet’, in Philip van Praag Jr and Kees Brants, eds, Tussen Beeld en Inhoud. Politiek en media in de verkiezingen van 1998 (Amsterdam: Het Spinhuis, 2000), pp. 214–242Google Scholar; van der Brug, Wouter and van der Eijk, Cees, ‘Welke effecten hadden de campagnes nu eigenlijk?’ in Kees Brants and Philip van Praag, eds, Politiek en Media in Verwarring. De verkiezingscampagnes in het lange jaar 2002 (Amsterdam: Spinhuis, 2005), pp. 244–267Google Scholar.
44 In our content analysis of media channels, we also included evening news broadcasts on ITV and BBC. Since news broadcasts prove inherently more neutral than newspaper op-ed articles, the coding of television news differs from newspaper coding. In newspaper coding, any opinion voiced by anyone other than the author was not coded. However, television newscasts largely broadcast only such statements from party officials, experts or interviewed citizens, and therefore these statements were coded for partisan bias.
45 Note that in a fixed effects model the first media indicator is automatically created, and is redundant. The second media indicator is hence most reflective of the media indicator variable in a fixed effects model. In multilevel literature, the procedure is often referred to as group-mean centering. Cf. Allison, Paul D., Effects Regression Methods for Longitudinal Data Using SAS (Cary, N.C.: SAS Institute, 2005)Google Scholar.
46 For respondents interviewed during the first six days of the campaign, the average was based on the number of days in the campaign so far.
47 These multicollinearity issues stem largely from the fact that a newspaper treatment of a party over the course of the campaign tends to be related to its treatment of the other parties.
48 For Labour and the Conservatives, tone on these two parties has been included. For the Liberal Democrats, separate models including tone on Labour or the Conservatives were run. Note that exclusion of tone on a third party does not reduce the explanatory power of the model.
50 Zaller, ‘Origins and Nature of Mass Opinion’.
51 At face value, it might seem desirable also to include non-readers as a control group whose opinion change over the campaign could serve as a baseline against which to assess media effects. Not only would this complicate the model specification but also, and more fundamentally, we are not proposing that readers as a whole are pushed in a direction different from that of non-readers. Rather, our models explain patterns of opinion change that result from specific media inputs – whether readers are actually persuaded by what their papers say. For that purpose, what matters are the differences in the messages readers are exposed to, and not any differences between the exposed and the unexposed.
53 Both the random and fixed effects estimates were obtained using STATA's xtreg.
54 Table 2 presents the model for Liberal Democrats with media tone on Conservatives as the second party. The model with media tone on the Labour party is presented in the Appendix.
55 We imputed the mean income for those respondents who failed to report their income. Since this estimate is not significant as well, these respondents show no significant difference in evaluation of the various parties.