THE ECONOMIC RECORD, VOL. 88, NO. 280, MARCH, 2012, 127–147 How Partisan is the Press? Multiple Measures of Media Slant* JOSHUA S. GANS ANDREW LEIGH Rotman School of Management, University of Economics Program, Research School of Toronto, Toronto, Ontario, Canada Social Sciences, Australian National University, ACT, Canberra, Australia We employ several different approaches to estimate the political position of Australian media outlets, relative to federal parliamen- tarians. First, we use parliamentary mentions to code over 100 public intellectuals on a left–right scale. We then estimate slant by using the number of mentions that each public intellectual receives in each media outlet. Second, we have independent raters separately code front-page election stories and headlines. Third, we tabulate the number of electoral endorsements that newspapers give to each side of politics in federal elections. Overall, we find that the Australian media are quite centrist, with very few outlets being statistically distinguishable from the middle of Australian politics. It is possible that this is because of the lack of competi- tion in the Australian media market. To the extent that we can separate content slant from editorial slant, we find some evidence that editors are more partisan than journalists. IIntroduction system.1 It is, therefore, hardly surprising that As the primary means through which politi- journalists and politicians are acutely concerned cians communicate with the electorate, a free and about the political leanings of media outlets. fair media is integral to a healthy democratic Occasionally, media outlets boast of their influ- ence, as with The Sun newspaper claiming the day after the Conservative victory in the 1992 UK * We have received valuable comments from asso- ciate editor Mardi Dungey, two anonymous referees, Tim Groseclose and seminar participants at the Aus- 1 Press freedom is enshrined in Article 19 of the tralian National University and the 2008 Australasian Universal Declaration of Human Rights, which states: Public Choice meetings. Vivienne Groves, Kimberley ‘Everyone has the right to freedom of opinion and Jin, Rachael Meager, Susanne Schmidt, Michelle Tan expression; this right includes freedom to hold opin- and Elena Varganova provided outstanding research ions without interference and to seek, receive, and assistance. We are also grateful to those individuals impart information and ideas through any media who coded newspaper headlines and articles: David regardless of frontiers’. An annual ‘Freedom of the Christie, Chamath De Silva, Robert Macdonald, Chris- Press Index’ compiled by Freedom House tracks the tine Mahoney and Jimmy Phengrasmy. extent to which media outlets are subject to overt JEL classifications: D72, L82 political control, as well as the potential for owner- Correspondence: Joshua S. Gans, Rotman School of ship concentration to lead to bias. In the 2008 report, Management, University of Toronto, Toronto, Ontario, Freedom House rates the freedom of the press in Aus- M5S 3E6, Canada. Email: [email protected] tralia as 35th out of 195 nations. 127 Ó 2011 The Economic Society of Australia doi: 10.1111/j.1475-4932.2011.00782.x 128 ECONOMIC RECORD MARCH election ‘It’s The Sun Wot Won It’. More fre- average difference between a newspaper’s quently, politicians object to perceptions of stance on ballot propositions and the public’s favouritism, as when Barack Obama described stance as expressed through their votes. We then Fox News in 2009 as ‘one television station that is compare this measure with the results from entirely devoted to attacking my administration’. other approaches, such as the political positions In this article, we focus on measuring media of think tanks (as used in Groseclose and Milyo, slant. We define a news outlet as being slanted 2003), or the coding of articles. Another meth- if it gives more favourable coverage to one side odological contribution of our work is to sepa- of politics than the other. While measuring rate the journalistic stance of a media outlet media slant is both important and policy- from its editorial stance. relevant, it is also empirically difficult. For Our second main contribution has to do with example, most media outlets tend to provide a the structure of the media market. In our empiri- greater volume of coverage to the incumbent cal analysis, we focus on Australia, in contrast political party than to opposition political par- with a literature that has previously been heav- ties. We do not regard differences in the volume ily focused on the United States. This has the of coverage in itself as being a form of media advantage that it allows us to see the extent to slant. However, a media outlet that criticised all which US findings can be generalised into other opposition proposals and praised all government contexts, and study a media market that is more announcements would be regarded as slanted. heavily concentrated. In our view, a good measure of media slant Politically, Australia is a bicameral parliamen- ought to reflect the ideological affinity between tary democracy with single-member electorates in a particular media outlet and one side of poli- House of Representatives and multi-member tics. In effect, such a measure plots media out- electorates (with state ⁄ territory boundaries) in lets onto the political spectrum, allowing us to the Senate. There are 150 members of the House answer questions like: ‘If this newspaper were a of Representatives and 76 senators. Voting is politician, how would it vote?’ compulsory (with a fine of A$20, a little less Note that we deliberately use the term ‘media than the median hourly wage), and ballots are slant’ instead of ‘media bias’, for the reason counted using preferential voting (also known as that our measures are relative rather than abso- instant runoff voting in the House of Represen- lute. To see this, suppose that a political party tative and Single Transferrable Vote in the Sen- were to publicly pronounce that the earth is ate). At the national level, there are effectively flat. In this instance, one might expect that two political parties: the left-leaning Australian most – if not all – media outlets would Labor Party, and a right-wing Coalition of the denounce that political party, perhaps making predominantly urban Liberal Party of Australia unkind comments about the intellect and judge- and the rural National Party of Australia. Party ment of the party’s leaders as they did so. If an discipline is strong, and it is extremely rare for election were in the offing, editorials in some members to ‘cross the floor’ and vote with the newspapers might even opine that these pro- opposing party.2 Our analysis focuses on the nouncements made the party unfit to govern. period 1996–2007, when the Coalition held Such an event would not reflect media bias, as office at the federal level. journalists are judging politicians statements Although two-party politics considerably sim- against an absolute standard (scientific truth). plifies our analysis, it is worth noting that it However, it would be captured as a form of may have the effect of collapsing multi-dimen- ‘media slant’. sional differences into a single continuum. Relative to the previous literature, our article Although most of the differences in Australian makes two main advances. The first relates to politics can be mapped onto a standard left– the methodology for estimating media slant. We right spectrum (e.g. size of government, level of introduce a new measure of media slant, based labour market regulation), our approach does on the political positions of public intellectuals, not allow for the possibility of a second axis which is different from those that have previ- ously been used in the literature. For example, rather than using public intellectuals, Puglisi 2 Such strict party discipline means that there would and Snyder (2010) estimated the relative politi- be little point in constructing Poole–Rosenthal-type cal position of a media outlet by analysing the measures of the ideology of Australian legislators. Ó 2011 The Economic Society of Australia 2012 HOW PARTISAN IS THE PRESS? 129 (e.g. authoritarian versus libertarian).3 In prac- range of media slant is more concentrated than tice, we believe that this is unlikely to be a has been observed for the United States. problem, given that Australia has strong party We also examine newspaper article content discipline, and a much lower emphasis than in and find that in reporting the 2004 election, there US politics on issues of personal liberty such as is relatively little bias in that content. Similarly, abortion, gun control or religious education. in absolute terms the same can be said for article To measure media slant, we use three headlines (which are determined at an editorial approaches. Our main approach is to use the level). On both content-coding metrics, only one political positions of ‘public intellectuals’ (i.e. outlet (The Age newspaper) is distinguishable commentators and academics who are regularly from the centre position. The same, however, quoted in both parliament and the press). Based cannot be said for editorial endorsements that, upon positive mentions on the floor of parlia- interestingly, are highly correlated with observed ment, we place each of the public intellectuals bias in article headlines. This suggests that slant on the political spectrum. Based on mentions in is determined at an editorial level rather than the media, we then develop an aggregated index through pressure or article selection by journal- of the political position of each media outlet. ists. We note that previous papers on media Our second measure of media slant relies on slant, such as Larcinese et al. (2011) and Puglisi content analysis. After removing all identifying and Snyder (2010), have not distinguished information (e.g. headline, newspaper name), we between the text and the headline of each article asked a team of people, which we refer to as and, therefore, did not capture this nuance.